From abeec4b63029c2c4151a78fc395d312113881845 Mon Sep 17 00:00:00 2001 From: captin411 Date: Wed, 19 Oct 2022 03:18:26 -0700 Subject: Add auto focal point cropping to Preprocess images This algorithm plots a bunch of points of interest on the source image and averages their locations to find a center. Most points come from OpenCV. One point comes from an entropy model. OpenCV points account for 50% of the weight and the entropy based point is the other 50%. The center of all weighted points is calculated and a bounding box is drawn as close to centered over that point as possible. --- modules/textual_inversion/preprocess.py | 151 ++++++++++++++++++++++++++++++-- 1 file changed, 146 insertions(+), 5 deletions(-) (limited to 'modules/textual_inversion/preprocess.py') diff --git a/modules/textual_inversion/preprocess.py b/modules/textual_inversion/preprocess.py index 886cf0c3..168bfb09 100644 --- a/modules/textual_inversion/preprocess.py +++ b/modules/textual_inversion/preprocess.py @@ -1,5 +1,7 @@ import os -from PIL import Image, ImageOps +import cv2 +import numpy as np +from PIL import Image, ImageOps, ImageDraw import platform import sys import tqdm @@ -11,7 +13,7 @@ if cmd_opts.deepdanbooru: import modules.deepbooru as deepbooru -def preprocess(process_src, process_dst, process_width, process_height, process_flip, process_split, process_caption, process_caption_deepbooru=False): +def preprocess(process_src, process_dst, process_width, process_height, process_flip, process_split, process_caption, process_caption_deepbooru=False, process_entropy_focus=False): try: if process_caption: shared.interrogator.load() @@ -21,7 +23,7 @@ def preprocess(process_src, process_dst, process_width, process_height, process_ db_opts[deepbooru.OPT_INCLUDE_RANKS] = False deepbooru.create_deepbooru_process(opts.interrogate_deepbooru_score_threshold, db_opts) - preprocess_work(process_src, process_dst, process_width, process_height, process_flip, process_split, process_caption, process_caption_deepbooru) + preprocess_work(process_src, process_dst, process_width, process_height, process_flip, process_split, process_caption, process_caption_deepbooru, process_entropy_focus) finally: @@ -33,7 +35,7 @@ def preprocess(process_src, process_dst, process_width, process_height, process_ -def preprocess_work(process_src, process_dst, process_width, process_height, process_flip, process_split, process_caption, process_caption_deepbooru=False): +def preprocess_work(process_src, process_dst, process_width, process_height, process_flip, process_split, process_caption, process_caption_deepbooru=False, process_entropy_focus=False): width = process_width height = process_height src = os.path.abspath(process_src) @@ -93,6 +95,8 @@ def preprocess_work(process_src, process_dst, process_width, process_height, pro is_tall = ratio > 1.35 is_wide = ratio < 1 / 1.35 + processing_option_ran = False + if process_split and is_tall: img = img.resize((width, height * img.height // img.width)) @@ -101,6 +105,8 @@ def preprocess_work(process_src, process_dst, process_width, process_height, pro bot = img.crop((0, img.height - height, width, img.height)) save_pic(bot, index) + + processing_option_ran = True elif process_split and is_wide: img = img.resize((width * img.width // img.height, height)) @@ -109,8 +115,143 @@ def preprocess_work(process_src, process_dst, process_width, process_height, pro right = img.crop((img.width - width, 0, img.width, height)) save_pic(right, index) - else: + + processing_option_ran = True + + if process_entropy_focus and (is_tall or is_wide): + if is_tall: + img = img.resize((width, height * img.height // img.width)) + else: + img = img.resize((width * img.width // img.height, height)) + + x_focal_center, y_focal_center = image_central_focal_point(img, width, height) + + # take the focal point and turn it into crop coordinates that try to center over the focal + # point but then get adjusted back into the frame + y_half = int(height / 2) + x_half = int(width / 2) + + x1 = x_focal_center - x_half + if x1 < 0: + x1 = 0 + elif x1 + width > img.width: + x1 = img.width - width + + y1 = y_focal_center - y_half + if y1 < 0: + y1 = 0 + elif y1 + height > img.height: + y1 = img.height - height + + x2 = x1 + width + y2 = y1 + height + + crop = [x1, y1, x2, y2] + + focal = img.crop(tuple(crop)) + save_pic(focal, index) + + processing_option_ran = True + + if not processing_option_ran: img = images.resize_image(1, img, width, height) save_pic(img, index) shared.state.nextjob() + + +def image_central_focal_point(im, target_width, target_height): + focal_points = [] + + focal_points.extend( + image_focal_points(im) + ) + + fp_entropy = image_entropy_point(im, target_width, target_height) + fp_entropy['weight'] = len(focal_points) + 1 # about half of the weight to entropy + + focal_points.append(fp_entropy) + + weight = 0.0 + x = 0.0 + y = 0.0 + for focal_point in focal_points: + weight += focal_point['weight'] + x += focal_point['x'] * focal_point['weight'] + y += focal_point['y'] * focal_point['weight'] + avg_x = round(x // weight) + avg_y = round(y // weight) + + return avg_x, avg_y + + +def image_focal_points(im): + grayscale = im.convert("L") + + # naive attempt at preventing focal points from collecting at watermarks near the bottom + gd = ImageDraw.Draw(grayscale) + gd.rectangle([0, im.height*.9, im.width, im.height], fill="#999") + + np_im = np.array(grayscale) + + points = cv2.goodFeaturesToTrack( + np_im, + maxCorners=50, + qualityLevel=0.04, + minDistance=min(grayscale.width, grayscale.height)*0.05, + useHarrisDetector=False, + ) + + if points is None: + return [] + + focal_points = [] + for point in points: + x, y = point.ravel() + focal_points.append({ + 'x': x, + 'y': y, + 'weight': 1.0 + }) + + return focal_points + + +def image_entropy_point(im, crop_width, crop_height): + img = im.copy() + # just make it easier to slide the test crop with images oriented the same way + if (img.size[0] < img.size[1]): + portrait = True + img = img.rotate(90, expand=1) + + e_max = 0 + crop_current = [0, 0, crop_width, crop_height] + crop_best = crop_current + while crop_current[2] < img.size[0]: + crop = img.crop(tuple(crop_current)) + e = image_entropy(crop) + + if (e_max < e): + e_max = e + crop_best = list(crop_current) + + crop_current[0] += 4 + crop_current[2] += 4 + + x_mid = int((crop_best[2] - crop_best[0])/2) + y_mid = int((crop_best[3] - crop_best[1])/2) + + return { + 'x': x_mid, + 'y': y_mid, + 'weight': 1.0 + } + + +def image_entropy(im): + # greyscale image entropy + band = np.asarray(im.convert("L")) + hist, _ = np.histogram(band, bins=range(0, 256)) + hist = hist[hist > 0] + return -np.log2(hist / hist.sum()).sum() + -- cgit v1.2.3 From 41e3877be2c667316515c86037413763eb0ba4da Mon Sep 17 00:00:00 2001 From: captin411 Date: Wed, 19 Oct 2022 13:44:59 -0700 Subject: fix entropy point calculation --- modules/textual_inversion/preprocess.py | 34 ++++++++++++++++++--------------- 1 file changed, 19 insertions(+), 15 deletions(-) (limited to 'modules/textual_inversion/preprocess.py') diff --git a/modules/textual_inversion/preprocess.py b/modules/textual_inversion/preprocess.py index 168bfb09..7c1a594e 100644 --- a/modules/textual_inversion/preprocess.py +++ b/modules/textual_inversion/preprocess.py @@ -196,9 +196,9 @@ def image_focal_points(im): points = cv2.goodFeaturesToTrack( np_im, - maxCorners=50, + maxCorners=100, qualityLevel=0.04, - minDistance=min(grayscale.width, grayscale.height)*0.05, + minDistance=min(grayscale.width, grayscale.height)*0.07, useHarrisDetector=False, ) @@ -218,28 +218,32 @@ def image_focal_points(im): def image_entropy_point(im, crop_width, crop_height): - img = im.copy() - # just make it easier to slide the test crop with images oriented the same way - if (img.size[0] < img.size[1]): - portrait = True - img = img.rotate(90, expand=1) + landscape = im.height < im.width + portrait = im.height > im.width + if landscape: + move_idx = [0, 2] + move_max = im.size[0] + elif portrait: + move_idx = [1, 3] + move_max = im.size[1] e_max = 0 crop_current = [0, 0, crop_width, crop_height] crop_best = crop_current - while crop_current[2] < img.size[0]: - crop = img.crop(tuple(crop_current)) + while crop_current[move_idx[1]] < move_max: + crop = im.crop(tuple(crop_current)) e = image_entropy(crop) - if (e_max < e): + if (e > e_max): e_max = e crop_best = list(crop_current) - crop_current[0] += 4 - crop_current[2] += 4 + crop_current[move_idx[0]] += 4 + crop_current[move_idx[1]] += 4 + + x_mid = int(crop_best[0] + crop_width/2) + y_mid = int(crop_best[1] + crop_height/2) - x_mid = int((crop_best[2] - crop_best[0])/2) - y_mid = int((crop_best[3] - crop_best[1])/2) return { 'x': x_mid, @@ -250,7 +254,7 @@ def image_entropy_point(im, crop_width, crop_height): def image_entropy(im): # greyscale image entropy - band = np.asarray(im.convert("L")) + band = np.asarray(im.convert("1")) hist, _ = np.histogram(band, bins=range(0, 256)) hist = hist[hist > 0] return -np.log2(hist / hist.sum()).sum() -- cgit v1.2.3 From fbcce66601994f6ed370db36d9c238840fed6bd2 Mon Sep 17 00:00:00 2001 From: DepFA <35278260+dfaker@users.noreply.github.com> Date: Thu, 20 Oct 2022 00:46:54 +0100 Subject: add existing caption file handling --- modules/textual_inversion/preprocess.py | 32 ++++++++++++++++++++++++-------- 1 file changed, 24 insertions(+), 8 deletions(-) (limited to 'modules/textual_inversion/preprocess.py') diff --git a/modules/textual_inversion/preprocess.py b/modules/textual_inversion/preprocess.py index 886cf0c3..5c43fe13 100644 --- a/modules/textual_inversion/preprocess.py +++ b/modules/textual_inversion/preprocess.py @@ -48,7 +48,7 @@ def preprocess_work(process_src, process_dst, process_width, process_height, pro shared.state.textinfo = "Preprocessing..." shared.state.job_count = len(files) - def save_pic_with_caption(image, index): + def save_pic_with_caption(image, index, existing_caption=None): caption = "" if process_caption: @@ -66,17 +66,26 @@ def preprocess_work(process_src, process_dst, process_width, process_height, pro basename = f"{index:05}-{subindex[0]}-{filename_part}" image.save(os.path.join(dst, f"{basename}.png")) + if preprocess_txt_action == 'prepend' and existing_caption: + caption = existing_caption + ' ' + caption + elif preprocess_txt_action == 'append' and existing_caption: + caption = caption + ' ' + existing_caption + elif preprocess_txt_action == 'copy' and existing_caption: + caption = existing_caption + + caption = caption.strip() + if len(caption) > 0: with open(os.path.join(dst, f"{basename}.txt"), "w", encoding="utf8") as file: file.write(caption) subindex[0] += 1 - def save_pic(image, index): + def save_pic(image, index, existing_caption=None): save_pic_with_caption(image, index) if process_flip: - save_pic_with_caption(ImageOps.mirror(image), index) + save_pic_with_caption(ImageOps.mirror(image), index, existing_caption=existing_caption) for index, imagefile in enumerate(tqdm.tqdm(files)): subindex = [0] @@ -86,6 +95,13 @@ def preprocess_work(process_src, process_dst, process_width, process_height, pro except Exception: continue + existing_caption = None + + try: + existing_caption = open(os.path.splitext(filename)[0] + '.txt', 'r').read() + except Exception as e: + print(e) + if shared.state.interrupted: break @@ -97,20 +113,20 @@ def preprocess_work(process_src, process_dst, process_width, process_height, pro img = img.resize((width, height * img.height // img.width)) top = img.crop((0, 0, width, height)) - save_pic(top, index) + save_pic(top, index, existing_caption=existing_caption) bot = img.crop((0, img.height - height, width, img.height)) - save_pic(bot, index) + save_pic(bot, index, existing_caption=existing_caption) elif process_split and is_wide: img = img.resize((width * img.width // img.height, height)) left = img.crop((0, 0, width, height)) - save_pic(left, index) + save_pic(left, index, existing_caption=existing_caption) right = img.crop((img.width - width, 0, img.width, height)) - save_pic(right, index) + save_pic(right, index, existing_caption=existing_caption) else: img = images.resize_image(1, img, width, height) - save_pic(img, index) + save_pic(img, index, existing_caption=existing_caption) shared.state.nextjob() -- cgit v1.2.3 From 9b65c4ecf4f8eb6187ee721918adebe68e9bc631 Mon Sep 17 00:00:00 2001 From: DepFA <35278260+dfaker@users.noreply.github.com> Date: Thu, 20 Oct 2022 00:49:23 +0100 Subject: pass preprocess_txt_action param --- modules/textual_inversion/preprocess.py | 6 +++--- 1 file changed, 3 insertions(+), 3 deletions(-) (limited to 'modules/textual_inversion/preprocess.py') diff --git a/modules/textual_inversion/preprocess.py b/modules/textual_inversion/preprocess.py index 5c43fe13..3713bc89 100644 --- a/modules/textual_inversion/preprocess.py +++ b/modules/textual_inversion/preprocess.py @@ -11,7 +11,7 @@ if cmd_opts.deepdanbooru: import modules.deepbooru as deepbooru -def preprocess(process_src, process_dst, process_width, process_height, process_flip, process_split, process_caption, process_caption_deepbooru=False): +def preprocess(process_src, process_dst, process_width, process_height, preprocess_txt_action, process_flip, process_split, process_caption, process_caption_deepbooru=False): try: if process_caption: shared.interrogator.load() @@ -21,7 +21,7 @@ def preprocess(process_src, process_dst, process_width, process_height, process_ db_opts[deepbooru.OPT_INCLUDE_RANKS] = False deepbooru.create_deepbooru_process(opts.interrogate_deepbooru_score_threshold, db_opts) - preprocess_work(process_src, process_dst, process_width, process_height, process_flip, process_split, process_caption, process_caption_deepbooru) + preprocess_work(process_src, process_dst, process_width, process_height, preprocess_txt_action, process_flip, process_split, process_caption, process_caption_deepbooru) finally: @@ -33,7 +33,7 @@ def preprocess(process_src, process_dst, process_width, process_height, process_ -def preprocess_work(process_src, process_dst, process_width, process_height, process_flip, process_split, process_caption, process_caption_deepbooru=False): +def preprocess_work(process_src, process_dst, process_width, process_height, preprocess_txt_action, process_flip, process_split, process_caption, process_caption_deepbooru=False): width = process_width height = process_height src = os.path.abspath(process_src) -- cgit v1.2.3 From 59ed74438318af893d2cba552b0e28dbc2a9266c Mon Sep 17 00:00:00 2001 From: captin411 Date: Wed, 19 Oct 2022 17:19:02 -0700 Subject: face detection algo, configurability, reusability Try to move the crop in the direction of a face if it is present More internal configuration options for choosing weights of each of the algorithm's findings Move logic into its module --- modules/textual_inversion/autocrop.py | 216 ++++++++++++++++++++++++++++++++ modules/textual_inversion/preprocess.py | 150 +++------------------- 2 files changed, 230 insertions(+), 136 deletions(-) create mode 100644 modules/textual_inversion/autocrop.py (limited to 'modules/textual_inversion/preprocess.py') diff --git a/modules/textual_inversion/autocrop.py b/modules/textual_inversion/autocrop.py new file mode 100644 index 00000000..f858a958 --- /dev/null +++ b/modules/textual_inversion/autocrop.py @@ -0,0 +1,216 @@ +import cv2 +from collections import defaultdict +from math import log, sqrt +import numpy as np +from PIL import Image, ImageDraw + +GREEN = "#0F0" +BLUE = "#00F" +RED = "#F00" + +def crop_image(im, settings): + """ Intelligently crop an image to the subject matter """ + if im.height > im.width: + im = im.resize((settings.crop_width, settings.crop_height * im.height // im.width)) + else: + im = im.resize((settings.crop_width * im.width // im.height, settings.crop_height)) + + focus = focal_point(im, settings) + + # take the focal point and turn it into crop coordinates that try to center over the focal + # point but then get adjusted back into the frame + y_half = int(settings.crop_height / 2) + x_half = int(settings.crop_width / 2) + + x1 = focus.x - x_half + if x1 < 0: + x1 = 0 + elif x1 + settings.crop_width > im.width: + x1 = im.width - settings.crop_width + + y1 = focus.y - y_half + if y1 < 0: + y1 = 0 + elif y1 + settings.crop_height > im.height: + y1 = im.height - settings.crop_height + + x2 = x1 + settings.crop_width + y2 = y1 + settings.crop_height + + crop = [x1, y1, x2, y2] + + if settings.annotate_image: + d = ImageDraw.Draw(im) + rect = list(crop) + rect[2] -= 1 + rect[3] -= 1 + d.rectangle(rect, outline=GREEN) + if settings.destop_view_image: + im.show() + + return im.crop(tuple(crop)) + +def focal_point(im, settings): + corner_points = image_corner_points(im, settings) + entropy_points = image_entropy_points(im, settings) + face_points = image_face_points(im, settings) + + total_points = len(corner_points) + len(entropy_points) + len(face_points) + + corner_weight = settings.corner_points_weight + entropy_weight = settings.entropy_points_weight + face_weight = settings.face_points_weight + + weight_pref_total = corner_weight + entropy_weight + face_weight + + # weight things + pois = [] + if weight_pref_total == 0 or total_points == 0: + return pois + + pois.extend( + [ PointOfInterest( p.x, p.y, weight=p.weight * ( (corner_weight/weight_pref_total) / (len(corner_points)/total_points) )) for p in corner_points ] + ) + pois.extend( + [ PointOfInterest( p.x, p.y, weight=p.weight * ( (entropy_weight/weight_pref_total) / (len(entropy_points)/total_points) )) for p in entropy_points ] + ) + pois.extend( + [ PointOfInterest( p.x, p.y, weight=p.weight * ( (face_weight/weight_pref_total) / (len(face_points)/total_points) )) for p in face_points ] + ) + + if settings.annotate_image: + d = ImageDraw.Draw(im) + + average_point = poi_average(pois, settings, im=im) + + if settings.annotate_image: + d.ellipse([average_point.x - 25, average_point.y - 25, average_point.x + 25, average_point.y + 25], outline=GREEN) + + return average_point + + +def image_face_points(im, settings): + np_im = np.array(im) + gray = cv2.cvtColor(np_im, cv2.COLOR_BGR2GRAY) + classifier = cv2.CascadeClassifier(f'{cv2.data.haarcascades}haarcascade_frontalface_default.xml') + + minsize = int(min(im.width, im.height) * 0.15) # at least N percent of the smallest side + faces = classifier.detectMultiScale(gray, scaleFactor=1.05, + minNeighbors=5, minSize=(minsize, minsize), flags=cv2.CASCADE_SCALE_IMAGE) + + if len(faces) == 0: + return [] + + rects = [[f[0], f[1], f[0] + f[2], f[1] + f[3]] for f in faces] + if settings.annotate_image: + for f in rects: + d = ImageDraw.Draw(im) + d.rectangle(f, outline=RED) + + return [PointOfInterest((r[0] +r[2]) // 2, (r[1] + r[3]) // 2) for r in rects] + + +def image_corner_points(im, settings): + grayscale = im.convert("L") + + # naive attempt at preventing focal points from collecting at watermarks near the bottom + gd = ImageDraw.Draw(grayscale) + gd.rectangle([0, im.height*.9, im.width, im.height], fill="#999") + + np_im = np.array(grayscale) + + points = cv2.goodFeaturesToTrack( + np_im, + maxCorners=100, + qualityLevel=0.04, + minDistance=min(grayscale.width, grayscale.height)*0.07, + useHarrisDetector=False, + ) + + if points is None: + return [] + + focal_points = [] + for point in points: + x, y = point.ravel() + focal_points.append(PointOfInterest(x, y)) + + return focal_points + + +def image_entropy_points(im, settings): + landscape = im.height < im.width + portrait = im.height > im.width + if landscape: + move_idx = [0, 2] + move_max = im.size[0] + elif portrait: + move_idx = [1, 3] + move_max = im.size[1] + else: + return [] + + e_max = 0 + crop_current = [0, 0, settings.crop_width, settings.crop_height] + crop_best = crop_current + while crop_current[move_idx[1]] < move_max: + crop = im.crop(tuple(crop_current)) + e = image_entropy(crop) + + if (e > e_max): + e_max = e + crop_best = list(crop_current) + + crop_current[move_idx[0]] += 4 + crop_current[move_idx[1]] += 4 + + x_mid = int(crop_best[0] + settings.crop_width/2) + y_mid = int(crop_best[1] + settings.crop_height/2) + + return [PointOfInterest(x_mid, y_mid)] + + +def image_entropy(im): + # greyscale image entropy + band = np.asarray(im.convert("1")) + hist, _ = np.histogram(band, bins=range(0, 256)) + hist = hist[hist > 0] + return -np.log2(hist / hist.sum()).sum() + + +def poi_average(pois, settings, im=None): + weight = 0.0 + x = 0.0 + y = 0.0 + for pois in pois: + if settings.annotate_image and im is not None: + w = 4 * 0.5 * sqrt(pois.weight) + d = ImageDraw.Draw(im) + d.ellipse([ + pois.x - w, pois.y - w, + pois.x + w, pois.y + w ], fill=BLUE) + weight += pois.weight + x += pois.x * pois.weight + y += pois.y * pois.weight + avg_x = round(x / weight) + avg_y = round(y / weight) + + return PointOfInterest(avg_x, avg_y) + + +class PointOfInterest: + def __init__(self, x, y, weight=1.0): + self.x = x + self.y = y + self.weight = weight + + +class Settings: + def __init__(self, crop_width=512, crop_height=512, corner_points_weight=0.5, entropy_points_weight=0.5, face_points_weight=0.5, annotate_image=False): + self.crop_width = crop_width + self.crop_height = crop_height + self.corner_points_weight = corner_points_weight + self.entropy_points_weight = entropy_points_weight + self.face_points_weight = entropy_points_weight + self.annotate_image = annotate_image + self.destop_view_image = False \ No newline at end of file diff --git a/modules/textual_inversion/preprocess.py b/modules/textual_inversion/preprocess.py index 7c1a594e..0c79f012 100644 --- a/modules/textual_inversion/preprocess.py +++ b/modules/textual_inversion/preprocess.py @@ -1,7 +1,5 @@ import os -import cv2 -import numpy as np -from PIL import Image, ImageOps, ImageDraw +from PIL import Image, ImageOps import platform import sys import tqdm @@ -9,6 +7,7 @@ import time from modules import shared, images from modules.shared import opts, cmd_opts +from modules.textual_inversion import autocrop if cmd_opts.deepdanbooru: import modules.deepbooru as deepbooru @@ -80,6 +79,7 @@ def preprocess_work(process_src, process_dst, process_width, process_height, pro if process_flip: save_pic_with_caption(ImageOps.mirror(image), index) + for index, imagefile in enumerate(tqdm.tqdm(files)): subindex = [0] filename = os.path.join(src, imagefile) @@ -118,37 +118,16 @@ def preprocess_work(process_src, process_dst, process_width, process_height, pro processing_option_ran = True - if process_entropy_focus and (is_tall or is_wide): - if is_tall: - img = img.resize((width, height * img.height // img.width)) - else: - img = img.resize((width * img.width // img.height, height)) - - x_focal_center, y_focal_center = image_central_focal_point(img, width, height) - - # take the focal point and turn it into crop coordinates that try to center over the focal - # point but then get adjusted back into the frame - y_half = int(height / 2) - x_half = int(width / 2) - - x1 = x_focal_center - x_half - if x1 < 0: - x1 = 0 - elif x1 + width > img.width: - x1 = img.width - width - - y1 = y_focal_center - y_half - if y1 < 0: - y1 = 0 - elif y1 + height > img.height: - y1 = img.height - height - - x2 = x1 + width - y2 = y1 + height - - crop = [x1, y1, x2, y2] - - focal = img.crop(tuple(crop)) + if process_entropy_focus and img.height != img.width: + autocrop_settings = autocrop.Settings( + crop_width = width, + crop_height = height, + face_points_weight = 0.9, + entropy_points_weight = 0.7, + corner_points_weight = 0.5, + annotate_image = False + ) + focal = autocrop.crop_image(img, autocrop_settings) save_pic(focal, index) processing_option_ran = True @@ -157,105 +136,4 @@ def preprocess_work(process_src, process_dst, process_width, process_height, pro img = images.resize_image(1, img, width, height) save_pic(img, index) - shared.state.nextjob() - - -def image_central_focal_point(im, target_width, target_height): - focal_points = [] - - focal_points.extend( - image_focal_points(im) - ) - - fp_entropy = image_entropy_point(im, target_width, target_height) - fp_entropy['weight'] = len(focal_points) + 1 # about half of the weight to entropy - - focal_points.append(fp_entropy) - - weight = 0.0 - x = 0.0 - y = 0.0 - for focal_point in focal_points: - weight += focal_point['weight'] - x += focal_point['x'] * focal_point['weight'] - y += focal_point['y'] * focal_point['weight'] - avg_x = round(x // weight) - avg_y = round(y // weight) - - return avg_x, avg_y - - -def image_focal_points(im): - grayscale = im.convert("L") - - # naive attempt at preventing focal points from collecting at watermarks near the bottom - gd = ImageDraw.Draw(grayscale) - gd.rectangle([0, im.height*.9, im.width, im.height], fill="#999") - - np_im = np.array(grayscale) - - points = cv2.goodFeaturesToTrack( - np_im, - maxCorners=100, - qualityLevel=0.04, - minDistance=min(grayscale.width, grayscale.height)*0.07, - useHarrisDetector=False, - ) - - if points is None: - return [] - - focal_points = [] - for point in points: - x, y = point.ravel() - focal_points.append({ - 'x': x, - 'y': y, - 'weight': 1.0 - }) - - return focal_points - - -def image_entropy_point(im, crop_width, crop_height): - landscape = im.height < im.width - portrait = im.height > im.width - if landscape: - move_idx = [0, 2] - move_max = im.size[0] - elif portrait: - move_idx = [1, 3] - move_max = im.size[1] - - e_max = 0 - crop_current = [0, 0, crop_width, crop_height] - crop_best = crop_current - while crop_current[move_idx[1]] < move_max: - crop = im.crop(tuple(crop_current)) - e = image_entropy(crop) - - if (e > e_max): - e_max = e - crop_best = list(crop_current) - - crop_current[move_idx[0]] += 4 - crop_current[move_idx[1]] += 4 - - x_mid = int(crop_best[0] + crop_width/2) - y_mid = int(crop_best[1] + crop_height/2) - - - return { - 'x': x_mid, - 'y': y_mid, - 'weight': 1.0 - } - - -def image_entropy(im): - # greyscale image entropy - band = np.asarray(im.convert("1")) - hist, _ = np.histogram(band, bins=range(0, 256)) - hist = hist[hist > 0] - return -np.log2(hist / hist.sum()).sum() - + shared.state.nextjob() \ No newline at end of file -- cgit v1.2.3 From 858462f719c22ca9f24b94a41699653c34b5f4fb Mon Sep 17 00:00:00 2001 From: DepFA <35278260+dfaker@users.noreply.github.com> Date: Thu, 20 Oct 2022 02:57:18 +0100 Subject: do caption copy for both flips --- modules/textual_inversion/preprocess.py | 2 +- 1 file changed, 1 insertion(+), 1 deletion(-) (limited to 'modules/textual_inversion/preprocess.py') diff --git a/modules/textual_inversion/preprocess.py b/modules/textual_inversion/preprocess.py index 3713bc89..6bba3852 100644 --- a/modules/textual_inversion/preprocess.py +++ b/modules/textual_inversion/preprocess.py @@ -82,7 +82,7 @@ def preprocess_work(process_src, process_dst, process_width, process_height, pre subindex[0] += 1 def save_pic(image, index, existing_caption=None): - save_pic_with_caption(image, index) + save_pic_with_caption(image, index, existing_caption=existing_caption) if process_flip: save_pic_with_caption(ImageOps.mirror(image), index, existing_caption=existing_caption) -- cgit v1.2.3 From 9681419e422515e42444e0174355b760645a846f Mon Sep 17 00:00:00 2001 From: Milly Date: Thu, 20 Oct 2022 16:53:46 +0900 Subject: train: fixed preprocess image ratio --- modules/textual_inversion/preprocess.py | 54 +++++++++++++++++++++------------ 1 file changed, 35 insertions(+), 19 deletions(-) (limited to 'modules/textual_inversion/preprocess.py') diff --git a/modules/textual_inversion/preprocess.py b/modules/textual_inversion/preprocess.py index 886cf0c3..2743bdeb 100644 --- a/modules/textual_inversion/preprocess.py +++ b/modules/textual_inversion/preprocess.py @@ -1,5 +1,6 @@ import os from PIL import Image, ImageOps +import math import platform import sys import tqdm @@ -38,6 +39,8 @@ def preprocess_work(process_src, process_dst, process_width, process_height, pro height = process_height src = os.path.abspath(process_src) dst = os.path.abspath(process_dst) + split_threshold = 0.5 + overlap_ratio = 0.2 assert src != dst, 'same directory specified as source and destination' @@ -78,6 +81,29 @@ def preprocess_work(process_src, process_dst, process_width, process_height, pro if process_flip: save_pic_with_caption(ImageOps.mirror(image), index) + def split_pic(image, inverse_xy): + if inverse_xy: + from_w, from_h = image.height, image.width + to_w, to_h = height, width + else: + from_w, from_h = image.width, image.height + to_w, to_h = width, height + h = from_h * to_w // from_w + if inverse_xy: + image = image.resize((h, to_w)) + else: + image = image.resize((to_w, h)) + + split_count = math.ceil((h - to_h * overlap_ratio) / (to_h * (1.0 - overlap_ratio))) + y_step = (h - to_h) / (split_count - 1) + for i in range(split_count): + y = int(y_step * i) + if inverse_xy: + splitted = image.crop((y, 0, y + to_h, to_w)) + else: + splitted = image.crop((0, y, to_w, y + to_h)) + yield splitted + for index, imagefile in enumerate(tqdm.tqdm(files)): subindex = [0] filename = os.path.join(src, imagefile) @@ -89,26 +115,16 @@ def preprocess_work(process_src, process_dst, process_width, process_height, pro if shared.state.interrupted: break - ratio = img.height / img.width - is_tall = ratio > 1.35 - is_wide = ratio < 1 / 1.35 - - if process_split and is_tall: - img = img.resize((width, height * img.height // img.width)) - - top = img.crop((0, 0, width, height)) - save_pic(top, index) - - bot = img.crop((0, img.height - height, width, img.height)) - save_pic(bot, index) - elif process_split and is_wide: - img = img.resize((width * img.width // img.height, height)) - - left = img.crop((0, 0, width, height)) - save_pic(left, index) + if img.height > img.width: + ratio = (img.width * height) / (img.height * width) + inverse_xy = False + else: + ratio = (img.height * width) / (img.width * height) + inverse_xy = True - right = img.crop((img.width - width, 0, img.width, height)) - save_pic(right, index) + if process_split and ratio < 1.0 and ratio <= split_threshold: + for splitted in split_pic(img, inverse_xy): + save_pic(splitted, index) else: img = images.resize_image(1, img, width, height) save_pic(img, index) -- cgit v1.2.3 From 85dd62c4c7635b8e21a75f140d093036069e97a1 Mon Sep 17 00:00:00 2001 From: Milly Date: Thu, 20 Oct 2022 22:56:45 +0900 Subject: train: ui: added `Split image threshold` and `Split image overlap ratio` to preprocess --- modules/textual_inversion/preprocess.py | 10 +++++----- modules/ui.py | 16 ++++++++++++++-- 2 files changed, 19 insertions(+), 7 deletions(-) (limited to 'modules/textual_inversion/preprocess.py') diff --git a/modules/textual_inversion/preprocess.py b/modules/textual_inversion/preprocess.py index 2743bdeb..c8df8aa0 100644 --- a/modules/textual_inversion/preprocess.py +++ b/modules/textual_inversion/preprocess.py @@ -12,7 +12,7 @@ if cmd_opts.deepdanbooru: import modules.deepbooru as deepbooru -def preprocess(process_src, process_dst, process_width, process_height, process_flip, process_split, process_caption, process_caption_deepbooru=False): +def preprocess(process_src, process_dst, process_width, process_height, process_flip, process_split, process_caption, process_caption_deepbooru=False, split_threshold=0.5, overlap_ratio=0.2): try: if process_caption: shared.interrogator.load() @@ -22,7 +22,7 @@ def preprocess(process_src, process_dst, process_width, process_height, process_ db_opts[deepbooru.OPT_INCLUDE_RANKS] = False deepbooru.create_deepbooru_process(opts.interrogate_deepbooru_score_threshold, db_opts) - preprocess_work(process_src, process_dst, process_width, process_height, process_flip, process_split, process_caption, process_caption_deepbooru) + preprocess_work(process_src, process_dst, process_width, process_height, process_flip, process_split, process_caption, process_caption_deepbooru, split_threshold, overlap_ratio) finally: @@ -34,13 +34,13 @@ def preprocess(process_src, process_dst, process_width, process_height, process_ -def preprocess_work(process_src, process_dst, process_width, process_height, process_flip, process_split, process_caption, process_caption_deepbooru=False): +def preprocess_work(process_src, process_dst, process_width, process_height, process_flip, process_split, process_caption, process_caption_deepbooru=False, split_threshold=0.5, overlap_ratio=0.2): width = process_width height = process_height src = os.path.abspath(process_src) dst = os.path.abspath(process_dst) - split_threshold = 0.5 - overlap_ratio = 0.2 + split_threshold = max(0.0, min(1.0, split_threshold)) + overlap_ratio = max(0.0, min(0.9, overlap_ratio)) assert src != dst, 'same directory specified as source and destination' diff --git a/modules/ui.py b/modules/ui.py index a2dbd41e..bc7f3330 100644 --- a/modules/ui.py +++ b/modules/ui.py @@ -1240,10 +1240,14 @@ def create_ui(wrap_gradio_gpu_call): with gr.Row(): process_flip = gr.Checkbox(label='Create flipped copies') - process_split = gr.Checkbox(label='Split oversized images into two') + process_split = gr.Checkbox(label='Split oversized images') process_caption = gr.Checkbox(label='Use BLIP for caption') process_caption_deepbooru = gr.Checkbox(label='Use deepbooru for caption', visible=True if cmd_opts.deepdanbooru else False) + with gr.Row(visible=False) as process_split_extra_row: + process_split_threshold = gr.Slider(label='Split image threshold', value=0.5, minimum=0.0, maximum=1.0, step=0.05) + process_overlap_ratio = gr.Slider(label='Split image overlap ratio', value=0.2, minimum=0.0, maximum=0.9, step=0.05) + with gr.Row(): with gr.Column(scale=3): gr.HTML(value="") @@ -1251,6 +1255,12 @@ def create_ui(wrap_gradio_gpu_call): with gr.Column(): run_preprocess = gr.Button(value="Preprocess", variant='primary') + process_split.change( + fn=lambda show: gr_show(show), + inputs=[process_split], + outputs=[process_split_extra_row], + ) + with gr.Tab(label="Train"): gr.HTML(value="

Train an embedding; must specify a directory with a set of 1:1 ratio images

") with gr.Row(): @@ -1327,7 +1337,9 @@ def create_ui(wrap_gradio_gpu_call): process_flip, process_split, process_caption, - process_caption_deepbooru + process_caption_deepbooru, + process_split_threshold, + process_overlap_ratio, ], outputs=[ ti_output, -- cgit v1.2.3 From f49c08ea566385db339c6628f65c3a121033f67c Mon Sep 17 00:00:00 2001 From: AUTOMATIC <16777216c@gmail.com> Date: Fri, 21 Oct 2022 18:46:02 +0300 Subject: prevent error spam when processing images without txt files for captions --- modules/textual_inversion/preprocess.py | 9 ++++----- 1 file changed, 4 insertions(+), 5 deletions(-) (limited to 'modules/textual_inversion/preprocess.py') diff --git a/modules/textual_inversion/preprocess.py b/modules/textual_inversion/preprocess.py index 17e4ddc1..33eaddb6 100644 --- a/modules/textual_inversion/preprocess.py +++ b/modules/textual_inversion/preprocess.py @@ -122,11 +122,10 @@ def preprocess_work(process_src, process_dst, process_width, process_height, pre continue existing_caption = None - - try: - existing_caption = open(os.path.splitext(filename)[0] + '.txt', 'r').read() - except Exception as e: - print(e) + existing_caption_filename = os.path.splitext(filename)[0] + '.txt' + if os.path.exists(existing_caption_filename): + with open(existing_caption_filename, 'r', encoding="utf8") as file: + existing_caption = file.read() if shared.state.interrupted: break -- cgit v1.2.3 From db8ed5fe5cd6e967d12d43d96b7f83083e58626c Mon Sep 17 00:00:00 2001 From: captin411 Date: Tue, 25 Oct 2022 15:22:29 -0700 Subject: Focal crop UI elements --- modules/textual_inversion/preprocess.py | 26 +++++++++++++------------- modules/ui.py | 20 ++++++++++++++++++-- 2 files changed, 31 insertions(+), 15 deletions(-) (limited to 'modules/textual_inversion/preprocess.py') diff --git a/modules/textual_inversion/preprocess.py b/modules/textual_inversion/preprocess.py index a8c17c6f..1e4d4de8 100644 --- a/modules/textual_inversion/preprocess.py +++ b/modules/textual_inversion/preprocess.py @@ -13,7 +13,7 @@ if cmd_opts.deepdanbooru: import modules.deepbooru as deepbooru -def preprocess(process_src, process_dst, process_width, process_height, preprocess_txt_action, process_flip, process_split, process_caption, process_caption_deepbooru=False, split_threshold=0.5, overlap_ratio=0.2, process_entropy_focus=False): +def preprocess(process_src, process_dst, process_width, process_height, preprocess_txt_action, process_flip, process_split, process_caption, process_caption_deepbooru=False, split_threshold=0.5, overlap_ratio=0.2, process_focal_crop=False, process_focal_crop_face_weight=0.9, process_focal_crop_entropy_weight=0.3, process_focal_crop_edges_weight=0.5, process_focal_crop_debug=False): try: if process_caption: shared.interrogator.load() @@ -23,7 +23,7 @@ def preprocess(process_src, process_dst, process_width, process_height, preproce db_opts[deepbooru.OPT_INCLUDE_RANKS] = False deepbooru.create_deepbooru_process(opts.interrogate_deepbooru_score_threshold, db_opts) - preprocess_work(process_src, process_dst, process_width, process_height, preprocess_txt_action, process_flip, process_split, process_caption, process_caption_deepbooru, split_threshold, overlap_ratio, process_entropy_focus) + preprocess_work(process_src, process_dst, process_width, process_height, preprocess_txt_action, process_flip, process_split, process_caption, process_caption_deepbooru, split_threshold, overlap_ratio, process_focal_crop, process_focal_crop_face_weight, process_focal_crop_entropy_weight, process_focal_crop_edges_weight, process_focal_crop_debug) finally: @@ -35,7 +35,7 @@ def preprocess(process_src, process_dst, process_width, process_height, preproce -def preprocess_work(process_src, process_dst, process_width, process_height, preprocess_txt_action, process_flip, process_split, process_caption, process_caption_deepbooru=False, split_threshold=0.5, overlap_ratio=0.2, process_entropy_focus=False): +def preprocess_work(process_src, process_dst, process_width, process_height, preprocess_txt_action, process_flip, process_split, process_caption, process_caption_deepbooru=False, split_threshold=0.5, overlap_ratio=0.2, process_focal_crop=False, process_focal_crop_face_weight=0.9, process_focal_crop_entropy_weight=0.3, process_focal_crop_edges_weight=0.5, process_focal_crop_debug=False): width = process_width height = process_height src = os.path.abspath(process_src) @@ -139,27 +139,27 @@ def preprocess_work(process_src, process_dst, process_width, process_height, pre ratio = (img.height * width) / (img.width * height) inverse_xy = True - processing_option_ran = False + process_default_resize = True if process_split and ratio < 1.0 and ratio <= split_threshold: for splitted in split_pic(img, inverse_xy): save_pic(splitted, index, existing_caption=existing_caption) - processing_option_ran = True + process_default_resize = False if process_entropy_focus and img.height != img.width: autocrop_settings = autocrop.Settings( crop_width = width, crop_height = height, - face_points_weight = 0.9, - entropy_points_weight = 0.7, - corner_points_weight = 0.5, - annotate_image = False + face_points_weight = process_focal_crop_face_weight, + entropy_points_weight = process_focal_crop_entropy_weight, + corner_points_weight = process_focal_crop_edges_weight, + annotate_image = process_focal_crop_debug ) - focal = autocrop.crop_image(img, autocrop_settings) - save_pic(focal, index, existing_caption=existing_caption) - processing_option_ran = True + for focal in autocrop.crop_image(img, autocrop_settings): + save_pic(focal, index, existing_caption=existing_caption) + process_default_resize = False - if not processing_option_ran: + if process_default_resize: img = images.resize_image(1, img, width, height) save_pic(img, index, existing_caption=existing_caption) diff --git a/modules/ui.py b/modules/ui.py index 028eb4e5..95b9c703 100644 --- a/modules/ui.py +++ b/modules/ui.py @@ -1260,7 +1260,7 @@ def create_ui(wrap_gradio_gpu_call): with gr.Row(): process_flip = gr.Checkbox(label='Create flipped copies') process_split = gr.Checkbox(label='Split oversized images') - process_entropy_focus = gr.Checkbox(label='Create auto focal point crop') + process_focal_crop = gr.Checkbox(label='Auto focal point crop') process_caption = gr.Checkbox(label='Use BLIP for caption') process_caption_deepbooru = gr.Checkbox(label='Use deepbooru for caption', visible=True if cmd_opts.deepdanbooru else False) @@ -1268,6 +1268,12 @@ def create_ui(wrap_gradio_gpu_call): process_split_threshold = gr.Slider(label='Split image threshold', value=0.5, minimum=0.0, maximum=1.0, step=0.05) process_overlap_ratio = gr.Slider(label='Split image overlap ratio', value=0.2, minimum=0.0, maximum=0.9, step=0.05) + with gr.Row(visible=False) as process_focal_crop_row: + process_focal_crop_face_weight = gr.Slider(label='Focal point face weight', value=0.9, minimum=0.0, maximum=1.0, step=0.05) + process_focal_crop_entropy_weight = gr.Slider(label='Focal point entropy weight', value=0.3, minimum=0.0, maximum=1.0, step=0.05) + process_focal_crop_edges_weight = gr.Slider(label='Focal point edges weight', value=0.5, minimum=0.0, maximum=1.0, step=0.05) + process_focal_crop_debug = gr.Checkbox(label='Create debug image') + with gr.Row(): with gr.Column(scale=3): gr.HTML(value="") @@ -1281,6 +1287,12 @@ def create_ui(wrap_gradio_gpu_call): outputs=[process_split_extra_row], ) + process_focal_crop.change( + fn=lambda show: gr_show(show), + inputs=[process_focal_crop], + outputs=[process_focal_crop_row], + ) + with gr.Tab(label="Train"): gr.HTML(value="

Train an embedding or Hypernetwork; you must specify a directory with a set of 1:1 ratio images [wiki]

") with gr.Row(): @@ -1368,7 +1380,11 @@ def create_ui(wrap_gradio_gpu_call): process_caption_deepbooru, process_split_threshold, process_overlap_ratio, - process_entropy_focus, + process_focal_crop, + process_focal_crop_face_weight, + process_focal_crop_entropy_weight, + process_focal_crop_edges_weight, + process_focal_crop_debug, ], outputs=[ ti_output, -- cgit v1.2.3 From 54f0c1482427a5b3f2248b97be55878e742cbcb1 Mon Sep 17 00:00:00 2001 From: captin411 Date: Tue, 25 Oct 2022 16:14:13 -0700 Subject: download better face detection module dynamically --- modules/textual_inversion/autocrop.py | 20 ++++++++++++++++++++ modules/textual_inversion/preprocess.py | 13 +++++++++++-- requirements.txt | 2 ++ 3 files changed, 33 insertions(+), 2 deletions(-) (limited to 'modules/textual_inversion/preprocess.py') diff --git a/modules/textual_inversion/autocrop.py b/modules/textual_inversion/autocrop.py index caaf18c8..01a92b12 100644 --- a/modules/textual_inversion/autocrop.py +++ b/modules/textual_inversion/autocrop.py @@ -1,4 +1,5 @@ import cv2 +import requests import os from collections import defaultdict from math import log, sqrt @@ -293,6 +294,25 @@ def is_square(w, h): return w == h +def download_and_cache_models(dirname): + download_url = 'https://github.com/opencv/opencv_zoo/blob/91fb0290f50896f38a0ab1e558b74b16bc009428/models/face_detection_yunet/face_detection_yunet_2022mar.onnx?raw=true' + model_file_name = 'face_detection_yunet.onnx' + + if not os.path.exists(dirname): + os.makedirs(dirname) + + cache_file = os.path.join(dirname, model_file_name) + if not os.path.exists(cache_file): + print(f"downloading face detection model from '{download_url}' to '{cache_file}'") + response = requests.get(download_url) + with open(cache_file, "wb") as f: + f.write(response.content) + + if os.path.exists(cache_file): + return cache_file + return None + + class PointOfInterest: def __init__(self, x, y, weight=1.0, size=10): self.x = x diff --git a/modules/textual_inversion/preprocess.py b/modules/textual_inversion/preprocess.py index 1e4d4de8..e13b1894 100644 --- a/modules/textual_inversion/preprocess.py +++ b/modules/textual_inversion/preprocess.py @@ -7,6 +7,7 @@ import tqdm import time from modules import shared, images +from modules.paths import models_path from modules.shared import opts, cmd_opts from modules.textual_inversion import autocrop if cmd_opts.deepdanbooru: @@ -146,14 +147,22 @@ def preprocess_work(process_src, process_dst, process_width, process_height, pre save_pic(splitted, index, existing_caption=existing_caption) process_default_resize = False - if process_entropy_focus and img.height != img.width: + if process_focal_crop and img.height != img.width: + + dnn_model_path = None + try: + dnn_model_path = autocrop.download_and_cache_models(os.path.join(models_path, "opencv")) + except Exception as e: + print("Unable to load face detection model for auto crop selection. Falling back to lower quality haar method.", e) + autocrop_settings = autocrop.Settings( crop_width = width, crop_height = height, face_points_weight = process_focal_crop_face_weight, entropy_points_weight = process_focal_crop_entropy_weight, corner_points_weight = process_focal_crop_edges_weight, - annotate_image = process_focal_crop_debug + annotate_image = process_focal_crop_debug, + dnn_model_path = dnn_model_path, ) for focal in autocrop.crop_image(img, autocrop_settings): save_pic(focal, index, existing_caption=existing_caption) diff --git a/requirements.txt b/requirements.txt index da1969cf..75b37c4f 100644 --- a/requirements.txt +++ b/requirements.txt @@ -8,6 +8,8 @@ gradio==3.5 invisible-watermark numpy omegaconf +opencv-python +requests piexif Pillow pytorch_lightning -- cgit v1.2.3 From 8011be33c36eb7aa9e9498fc714614034e07f67a Mon Sep 17 00:00:00 2001 From: AUTOMATIC <16777216c@gmail.com> Date: Tue, 8 Nov 2022 08:37:05 +0300 Subject: move functions out of main body for image preprocessing for easier hijacking --- modules/textual_inversion/preprocess.py | 162 ++++++++++++++++++-------------- 1 file changed, 93 insertions(+), 69 deletions(-) (limited to 'modules/textual_inversion/preprocess.py') diff --git a/modules/textual_inversion/preprocess.py b/modules/textual_inversion/preprocess.py index e13b1894..488aa5b5 100644 --- a/modules/textual_inversion/preprocess.py +++ b/modules/textual_inversion/preprocess.py @@ -35,6 +35,84 @@ def preprocess(process_src, process_dst, process_width, process_height, preproce deepbooru.release_process() +def listfiles(dirname): + return os.listdir(dirname) + + +class PreprocessParams: + src = None + dstdir = None + subindex = 0 + flip = False + process_caption = False + process_caption_deepbooru = False + preprocess_txt_action = None + + +def save_pic_with_caption(image, index, params: PreprocessParams, existing_caption=None): + caption = "" + + if params.process_caption: + caption += shared.interrogator.generate_caption(image) + + if params.process_caption_deepbooru: + if len(caption) > 0: + caption += ", " + caption += deepbooru.get_tags_from_process(image) + + filename_part = params.src + filename_part = os.path.splitext(filename_part)[0] + filename_part = os.path.basename(filename_part) + + basename = f"{index:05}-{params.subindex}-{filename_part}" + image.save(os.path.join(params.dstdir, f"{basename}.png")) + + if params.preprocess_txt_action == 'prepend' and existing_caption: + caption = existing_caption + ' ' + caption + elif params.preprocess_txt_action == 'append' and existing_caption: + caption = caption + ' ' + existing_caption + elif params.preprocess_txt_action == 'copy' and existing_caption: + caption = existing_caption + + caption = caption.strip() + + if len(caption) > 0: + with open(os.path.join(params.dstdir, f"{basename}.txt"), "w", encoding="utf8") as file: + file.write(caption) + + params.subindex += 1 + + +def save_pic(image, index, params, existing_caption=None): + save_pic_with_caption(image, index, params, existing_caption=existing_caption) + + if params.flip: + save_pic_with_caption(ImageOps.mirror(image), index, params, existing_caption=existing_caption) + + +def split_pic(image, inverse_xy, width, height, overlap_ratio): + if inverse_xy: + from_w, from_h = image.height, image.width + to_w, to_h = height, width + else: + from_w, from_h = image.width, image.height + to_w, to_h = width, height + h = from_h * to_w // from_w + if inverse_xy: + image = image.resize((h, to_w)) + else: + image = image.resize((to_w, h)) + + split_count = math.ceil((h - to_h * overlap_ratio) / (to_h * (1.0 - overlap_ratio))) + y_step = (h - to_h) / (split_count - 1) + for i in range(split_count): + y = int(y_step * i) + if inverse_xy: + splitted = image.crop((y, 0, y + to_h, to_w)) + else: + splitted = image.crop((0, y, to_w, y + to_h)) + yield splitted + def preprocess_work(process_src, process_dst, process_width, process_height, preprocess_txt_action, process_flip, process_split, process_caption, process_caption_deepbooru=False, split_threshold=0.5, overlap_ratio=0.2, process_focal_crop=False, process_focal_crop_face_weight=0.9, process_focal_crop_entropy_weight=0.3, process_focal_crop_edges_weight=0.5, process_focal_crop_debug=False): width = process_width @@ -48,82 +126,28 @@ def preprocess_work(process_src, process_dst, process_width, process_height, pre os.makedirs(dst, exist_ok=True) - files = os.listdir(src) + files = listfiles(src) shared.state.textinfo = "Preprocessing..." shared.state.job_count = len(files) - def save_pic_with_caption(image, index, existing_caption=None): - caption = "" - - if process_caption: - caption += shared.interrogator.generate_caption(image) - - if process_caption_deepbooru: - if len(caption) > 0: - caption += ", " - caption += deepbooru.get_tags_from_process(image) - - filename_part = filename - filename_part = os.path.splitext(filename_part)[0] - filename_part = os.path.basename(filename_part) - - basename = f"{index:05}-{subindex[0]}-{filename_part}" - image.save(os.path.join(dst, f"{basename}.png")) - - if preprocess_txt_action == 'prepend' and existing_caption: - caption = existing_caption + ' ' + caption - elif preprocess_txt_action == 'append' and existing_caption: - caption = caption + ' ' + existing_caption - elif preprocess_txt_action == 'copy' and existing_caption: - caption = existing_caption - - caption = caption.strip() - - if len(caption) > 0: - with open(os.path.join(dst, f"{basename}.txt"), "w", encoding="utf8") as file: - file.write(caption) - - subindex[0] += 1 - - def save_pic(image, index, existing_caption=None): - save_pic_with_caption(image, index, existing_caption=existing_caption) - - if process_flip: - save_pic_with_caption(ImageOps.mirror(image), index, existing_caption=existing_caption) - - def split_pic(image, inverse_xy): - if inverse_xy: - from_w, from_h = image.height, image.width - to_w, to_h = height, width - else: - from_w, from_h = image.width, image.height - to_w, to_h = width, height - h = from_h * to_w // from_w - if inverse_xy: - image = image.resize((h, to_w)) - else: - image = image.resize((to_w, h)) - - split_count = math.ceil((h - to_h * overlap_ratio) / (to_h * (1.0 - overlap_ratio))) - y_step = (h - to_h) / (split_count - 1) - for i in range(split_count): - y = int(y_step * i) - if inverse_xy: - splitted = image.crop((y, 0, y + to_h, to_w)) - else: - splitted = image.crop((0, y, to_w, y + to_h)) - yield splitted - + params = PreprocessParams() + params.dstdir = dst + params.flip = process_flip + params.process_caption = process_caption + params.process_caption_deepbooru = process_caption_deepbooru + params.preprocess_txt_action = preprocess_txt_action for index, imagefile in enumerate(tqdm.tqdm(files)): - subindex = [0] + params.subindex = 0 filename = os.path.join(src, imagefile) try: img = Image.open(filename).convert("RGB") except Exception: continue + params.src = filename + existing_caption = None existing_caption_filename = os.path.splitext(filename)[0] + '.txt' if os.path.exists(existing_caption_filename): @@ -143,8 +167,8 @@ def preprocess_work(process_src, process_dst, process_width, process_height, pre process_default_resize = True if process_split and ratio < 1.0 and ratio <= split_threshold: - for splitted in split_pic(img, inverse_xy): - save_pic(splitted, index, existing_caption=existing_caption) + for splitted in split_pic(img, inverse_xy, width, height, overlap_ratio): + save_pic(splitted, index, params, existing_caption=existing_caption) process_default_resize = False if process_focal_crop and img.height != img.width: @@ -165,11 +189,11 @@ def preprocess_work(process_src, process_dst, process_width, process_height, pre dnn_model_path = dnn_model_path, ) for focal in autocrop.crop_image(img, autocrop_settings): - save_pic(focal, index, existing_caption=existing_caption) + save_pic(focal, index, params, existing_caption=existing_caption) process_default_resize = False if process_default_resize: img = images.resize_image(1, img, width, height) - save_pic(img, index, existing_caption=existing_caption) + save_pic(img, index, params, existing_caption=existing_caption) - shared.state.nextjob() \ No newline at end of file + shared.state.nextjob() -- cgit v1.2.3 From c81d440d876dfd2ab3560410f37442ef56fc6632 Mon Sep 17 00:00:00 2001 From: AUTOMATIC <16777216c@gmail.com> Date: Sun, 20 Nov 2022 16:39:20 +0300 Subject: moved deepdanbooru to pure pytorch implementation --- README.md | 2 +- launch.py | 5 - modules/api/api.py | 10 +- modules/deepbooru.py | 258 +++++------- modules/deepbooru_model.py | 676 ++++++++++++++++++++++++++++++++ modules/shared.py | 2 +- modules/textual_inversion/preprocess.py | 12 +- modules/ui.py | 7 +- 8 files changed, 777 insertions(+), 195 deletions(-) create mode 100644 modules/deepbooru_model.py (limited to 'modules/textual_inversion/preprocess.py') diff --git a/README.md b/README.md index 33508f31..5f5ab3aa 100644 --- a/README.md +++ b/README.md @@ -70,7 +70,7 @@ Check the [custom scripts](https://github.com/AUTOMATIC1111/stable-diffusion-web - separate prompts using uppercase `AND` - also supports weights for prompts: `a cat :1.2 AND a dog AND a penguin :2.2` - No token limit for prompts (original stable diffusion lets you use up to 75 tokens) -- DeepDanbooru integration, creates danbooru style tags for anime prompts (add --deepdanbooru to commandline args) +- DeepDanbooru integration, creates danbooru style tags for anime prompts - [xformers](https://github.com/AUTOMATIC1111/stable-diffusion-webui/wiki/Xformers), major speed increase for select cards: (add --xformers to commandline args) - via extension: [History tab](https://github.com/yfszzx/stable-diffusion-webui-images-browser): view, direct and delete images conveniently within the UI - Generate forever option diff --git a/launch.py b/launch.py index 0f84b5d1..d2f1055c 100644 --- a/launch.py +++ b/launch.py @@ -134,7 +134,6 @@ def prepare_enviroment(): gfpgan_package = os.environ.get('GFPGAN_PACKAGE', "git+https://github.com/TencentARC/GFPGAN.git@8d2447a2d918f8eba5a4a01463fd48e45126a379") clip_package = os.environ.get('CLIP_PACKAGE', "git+https://github.com/openai/CLIP.git@d50d76daa670286dd6cacf3bcd80b5e4823fc8e1") - deepdanbooru_package = os.environ.get('DEEPDANBOORU_PACKAGE', "git+https://github.com/KichangKim/DeepDanbooru.git@d91a2963bf87c6a770d74894667e9ffa9f6de7ff") xformers_windows_package = os.environ.get('XFORMERS_WINDOWS_PACKAGE', 'https://github.com/C43H66N12O12S2/stable-diffusion-webui/releases/download/f/xformers-0.0.14.dev0-cp310-cp310-win_amd64.whl') @@ -158,7 +157,6 @@ def prepare_enviroment(): sys.argv, update_check = extract_arg(sys.argv, '--update-check') sys.argv, run_tests = extract_arg(sys.argv, '--tests') xformers = '--xformers' in sys.argv - deepdanbooru = '--deepdanbooru' in sys.argv ngrok = '--ngrok' in sys.argv try: @@ -193,9 +191,6 @@ def prepare_enviroment(): elif platform.system() == "Linux": run_pip("install xformers", "xformers") - if not is_installed("deepdanbooru") and deepdanbooru: - run_pip(f"install {deepdanbooru_package}#egg=deepdanbooru[tensorflow] tensorflow==2.10.0 tensorflow-io==0.27.0", "deepdanbooru") - if not is_installed("pyngrok") and ngrok: run_pip("install pyngrok", "ngrok") diff --git a/modules/api/api.py b/modules/api/api.py index 79b2c818..7a567be3 100644 --- a/modules/api/api.py +++ b/modules/api/api.py @@ -9,7 +9,7 @@ from fastapi.security import HTTPBasic, HTTPBasicCredentials from secrets import compare_digest import modules.shared as shared -from modules import sd_samplers +from modules import sd_samplers, deepbooru from modules.api.models import * from modules.processing import StableDiffusionProcessingTxt2Img, StableDiffusionProcessingImg2Img, process_images from modules.extras import run_extras, run_pnginfo @@ -18,9 +18,6 @@ from modules.sd_models import checkpoints_list from modules.realesrgan_model import get_realesrgan_models from typing import List -if shared.cmd_opts.deepdanbooru: - from modules.deepbooru import get_deepbooru_tags - def upscaler_to_index(name: str): try: return [x.name.lower() for x in shared.sd_upscalers].index(name.lower()) @@ -245,10 +242,7 @@ class Api: if interrogatereq.model == "clip": processed = shared.interrogator.interrogate(img) elif interrogatereq.model == "deepdanbooru": - if shared.cmd_opts.deepdanbooru: - processed = get_deepbooru_tags(img) - else: - raise HTTPException(status_code=404, detail="Model not found. Add --deepdanbooru when launching for using the model.") + processed = deepbooru.model.tag(img) else: raise HTTPException(status_code=404, detail="Model not found") diff --git a/modules/deepbooru.py b/modules/deepbooru.py index 8bbc90a4..b9066d81 100644 --- a/modules/deepbooru.py +++ b/modules/deepbooru.py @@ -1,173 +1,97 @@ -import os.path -from concurrent.futures import ProcessPoolExecutor -import multiprocessing -import time +import os import re +import torch +from PIL import Image +import numpy as np + +from modules import modelloader, paths, deepbooru_model, devices, images, shared + re_special = re.compile(r'([\\()])') -def get_deepbooru_tags(pil_image): - """ - This method is for running only one image at a time for simple use. Used to the img2img interrogate. - """ - from modules import shared # prevents circular reference - - try: - create_deepbooru_process(shared.opts.interrogate_deepbooru_score_threshold, create_deepbooru_opts()) - return get_tags_from_process(pil_image) - finally: - release_process() - - -OPT_INCLUDE_RANKS = "include_ranks" -def create_deepbooru_opts(): - from modules import shared - - return { - "use_spaces": shared.opts.deepbooru_use_spaces, - "use_escape": shared.opts.deepbooru_escape, - "alpha_sort": shared.opts.deepbooru_sort_alpha, - OPT_INCLUDE_RANKS: shared.opts.interrogate_return_ranks, - } - - -def deepbooru_process(queue, deepbooru_process_return, threshold, deepbooru_opts): - model, tags = get_deepbooru_tags_model() - while True: # while process is running, keep monitoring queue for new image - pil_image = queue.get() - if pil_image == "QUIT": - break - else: - deepbooru_process_return["value"] = get_deepbooru_tags_from_model(model, tags, pil_image, threshold, deepbooru_opts) - - -def create_deepbooru_process(threshold, deepbooru_opts): - """ - Creates deepbooru process. A queue is created to send images into the process. This enables multiple images - to be processed in a row without reloading the model or creating a new process. To return the data, a shared - dictionary is created to hold the tags created. To wait for tags to be returned, a value of -1 is assigned - to the dictionary and the method adding the image to the queue should wait for this value to be updated with - the tags. - """ - from modules import shared # prevents circular reference - context = multiprocessing.get_context("spawn") - shared.deepbooru_process_manager = context.Manager() - shared.deepbooru_process_queue = shared.deepbooru_process_manager.Queue() - shared.deepbooru_process_return = shared.deepbooru_process_manager.dict() - shared.deepbooru_process_return["value"] = -1 - shared.deepbooru_process = context.Process(target=deepbooru_process, args=(shared.deepbooru_process_queue, shared.deepbooru_process_return, threshold, deepbooru_opts)) - shared.deepbooru_process.start() - - -def get_tags_from_process(image): - from modules import shared - - shared.deepbooru_process_return["value"] = -1 - shared.deepbooru_process_queue.put(image) - while shared.deepbooru_process_return["value"] == -1: - time.sleep(0.2) - caption = shared.deepbooru_process_return["value"] - shared.deepbooru_process_return["value"] = -1 - - return caption - - -def release_process(): - """ - Stops the deepbooru process to return used memory - """ - from modules import shared # prevents circular reference - shared.deepbooru_process_queue.put("QUIT") - shared.deepbooru_process.join() - shared.deepbooru_process_queue = None - shared.deepbooru_process = None - shared.deepbooru_process_return = None - shared.deepbooru_process_manager = None - -def get_deepbooru_tags_model(): - import deepdanbooru as dd - import tensorflow as tf - import numpy as np - this_folder = os.path.dirname(__file__) - model_path = os.path.abspath(os.path.join(this_folder, '..', 'models', 'deepbooru')) - if not os.path.exists(os.path.join(model_path, 'project.json')): - # there is no point importing these every time - import zipfile - from basicsr.utils.download_util import load_file_from_url - load_file_from_url( - r"https://github.com/KichangKim/DeepDanbooru/releases/download/v3-20211112-sgd-e28/deepdanbooru-v3-20211112-sgd-e28.zip", - model_path) - with zipfile.ZipFile(os.path.join(model_path, "deepdanbooru-v3-20211112-sgd-e28.zip"), "r") as zip_ref: - zip_ref.extractall(model_path) - os.remove(os.path.join(model_path, "deepdanbooru-v3-20211112-sgd-e28.zip")) - - tags = dd.project.load_tags_from_project(model_path) - model = dd.project.load_model_from_project( - model_path, compile_model=False - ) - return model, tags - - -def get_deepbooru_tags_from_model(model, tags, pil_image, threshold, deepbooru_opts): - import deepdanbooru as dd - import tensorflow as tf - import numpy as np - - alpha_sort = deepbooru_opts['alpha_sort'] - use_spaces = deepbooru_opts['use_spaces'] - use_escape = deepbooru_opts['use_escape'] - include_ranks = deepbooru_opts['include_ranks'] - - width = model.input_shape[2] - height = model.input_shape[1] - image = np.array(pil_image) - image = tf.image.resize( - image, - size=(height, width), - method=tf.image.ResizeMethod.AREA, - preserve_aspect_ratio=True, - ) - image = image.numpy() # EagerTensor to np.array - image = dd.image.transform_and_pad_image(image, width, height) - image = image / 255.0 - image_shape = image.shape - image = image.reshape((1, image_shape[0], image_shape[1], image_shape[2])) - - y = model.predict(image)[0] - - result_dict = {} - - for i, tag in enumerate(tags): - result_dict[tag] = y[i] - - unsorted_tags_in_theshold = [] - result_tags_print = [] - for tag in tags: - if result_dict[tag] >= threshold: + +class DeepDanbooru: + def __init__(self): + self.model = None + + def load(self): + if self.model is not None: + return + + files = modelloader.load_models( + model_path=os.path.join(paths.models_path, "torch_deepdanbooru"), + model_url='https://github.com/AUTOMATIC1111/TorchDeepDanbooru/releases/download/v1/model-resnet_custom_v3.pt', + ext_filter=".pt", + download_name='model-resnet_custom_v3.pt', + ) + + self.model = deepbooru_model.DeepDanbooruModel() + self.model.load_state_dict(torch.load(files[0], map_location="cpu")) + + self.model.eval() + self.model.to(devices.cpu, devices.dtype) + + def start(self): + self.load() + self.model.to(devices.device) + + def stop(self): + if not shared.opts.interrogate_keep_models_in_memory: + self.model.to(devices.cpu) + devices.torch_gc() + + def tag(self, pil_image): + self.start() + res = self.tag_multi(pil_image) + self.stop() + + return res + + def tag_multi(self, pil_image, force_disable_ranks=False): + threshold = shared.opts.interrogate_deepbooru_score_threshold + use_spaces = shared.opts.deepbooru_use_spaces + use_escape = shared.opts.deepbooru_escape + alpha_sort = shared.opts.deepbooru_sort_alpha + include_ranks = shared.opts.interrogate_return_ranks and not force_disable_ranks + + pic = images.resize_image(2, pil_image.convert("RGB"), 512, 512) + a = np.expand_dims(np.array(pic, dtype=np.float32), 0) / 255 + + with torch.no_grad(), devices.autocast(): + x = torch.from_numpy(a).cuda() + y = self.model(x)[0].detach().cpu().numpy() + + probability_dict = {} + + for tag, probability in zip(self.model.tags, y): + if probability < threshold: + continue + if tag.startswith("rating:"): continue - unsorted_tags_in_theshold.append((result_dict[tag], tag)) - result_tags_print.append(f'{result_dict[tag]} {tag}') - - # sort tags - result_tags_out = [] - sort_ndx = 0 - if alpha_sort: - sort_ndx = 1 - - # sort by reverse by likelihood and normal for alpha, and format tag text as requested - unsorted_tags_in_theshold.sort(key=lambda y: y[sort_ndx], reverse=(not alpha_sort)) - for weight, tag in unsorted_tags_in_theshold: - tag_outformat = tag - if use_spaces: - tag_outformat = tag_outformat.replace('_', ' ') - if use_escape: - tag_outformat = re.sub(re_special, r'\\\1', tag_outformat) - if include_ranks: - tag_outformat = f"({tag_outformat}:{weight:.3f})" - - result_tags_out.append(tag_outformat) - - print('\n'.join(sorted(result_tags_print, reverse=True))) - - return ', '.join(result_tags_out) + + probability_dict[tag] = probability + + if alpha_sort: + tags = sorted(probability_dict) + else: + tags = [tag for tag, _ in sorted(probability_dict.items(), key=lambda x: -x[1])] + + res = [] + + for tag in tags: + probability = probability_dict[tag] + tag_outformat = tag + if use_spaces: + tag_outformat = tag_outformat.replace('_', ' ') + if use_escape: + tag_outformat = re.sub(re_special, r'\\\1', tag_outformat) + if include_ranks: + tag_outformat = f"({tag_outformat}:{probability:.3f})" + + res.append(tag_outformat) + + return ", ".join(res) + + +model = DeepDanbooru() diff --git a/modules/deepbooru_model.py b/modules/deepbooru_model.py new file mode 100644 index 00000000..edd40c81 --- /dev/null +++ b/modules/deepbooru_model.py @@ -0,0 +1,676 @@ +import torch +import torch.nn as nn +import torch.nn.functional as F + +# see https://github.com/AUTOMATIC1111/TorchDeepDanbooru for more + + +class DeepDanbooruModel(nn.Module): + def __init__(self): + super(DeepDanbooruModel, self).__init__() + + self.tags = [] + + self.n_Conv_0 = nn.Conv2d(kernel_size=(7, 7), in_channels=3, out_channels=64, stride=(2, 2)) + self.n_MaxPool_0 = nn.MaxPool2d(kernel_size=(3, 3), stride=(2, 2)) + self.n_Conv_1 = nn.Conv2d(kernel_size=(1, 1), in_channels=64, out_channels=256) + self.n_Conv_2 = nn.Conv2d(kernel_size=(1, 1), in_channels=64, out_channels=64) + self.n_Conv_3 = nn.Conv2d(kernel_size=(3, 3), in_channels=64, out_channels=64) + self.n_Conv_4 = nn.Conv2d(kernel_size=(1, 1), in_channels=64, out_channels=256) + self.n_Conv_5 = nn.Conv2d(kernel_size=(1, 1), in_channels=256, out_channels=64) + self.n_Conv_6 = nn.Conv2d(kernel_size=(3, 3), in_channels=64, out_channels=64) + self.n_Conv_7 = nn.Conv2d(kernel_size=(1, 1), in_channels=64, out_channels=256) + self.n_Conv_8 = nn.Conv2d(kernel_size=(1, 1), in_channels=256, out_channels=64) + self.n_Conv_9 = nn.Conv2d(kernel_size=(3, 3), in_channels=64, out_channels=64) + self.n_Conv_10 = nn.Conv2d(kernel_size=(1, 1), in_channels=64, out_channels=256) + self.n_Conv_11 = nn.Conv2d(kernel_size=(1, 1), in_channels=256, out_channels=512, stride=(2, 2)) + self.n_Conv_12 = nn.Conv2d(kernel_size=(1, 1), in_channels=256, out_channels=128) + self.n_Conv_13 = nn.Conv2d(kernel_size=(3, 3), in_channels=128, out_channels=128, stride=(2, 2)) + self.n_Conv_14 = nn.Conv2d(kernel_size=(1, 1), in_channels=128, out_channels=512) + self.n_Conv_15 = nn.Conv2d(kernel_size=(1, 1), in_channels=512, out_channels=128) + self.n_Conv_16 = nn.Conv2d(kernel_size=(3, 3), in_channels=128, out_channels=128) + self.n_Conv_17 = nn.Conv2d(kernel_size=(1, 1), in_channels=128, out_channels=512) + self.n_Conv_18 = nn.Conv2d(kernel_size=(1, 1), in_channels=512, out_channels=128) + self.n_Conv_19 = nn.Conv2d(kernel_size=(3, 3), in_channels=128, out_channels=128) + self.n_Conv_20 = nn.Conv2d(kernel_size=(1, 1), in_channels=128, out_channels=512) + self.n_Conv_21 = nn.Conv2d(kernel_size=(1, 1), in_channels=512, out_channels=128) + self.n_Conv_22 = nn.Conv2d(kernel_size=(3, 3), in_channels=128, out_channels=128) + self.n_Conv_23 = nn.Conv2d(kernel_size=(1, 1), in_channels=128, out_channels=512) + self.n_Conv_24 = nn.Conv2d(kernel_size=(1, 1), in_channels=512, out_channels=128) + self.n_Conv_25 = nn.Conv2d(kernel_size=(3, 3), in_channels=128, out_channels=128) + self.n_Conv_26 = nn.Conv2d(kernel_size=(1, 1), in_channels=128, out_channels=512) + self.n_Conv_27 = nn.Conv2d(kernel_size=(1, 1), in_channels=512, out_channels=128) + self.n_Conv_28 = nn.Conv2d(kernel_size=(3, 3), in_channels=128, out_channels=128) + self.n_Conv_29 = nn.Conv2d(kernel_size=(1, 1), in_channels=128, out_channels=512) + self.n_Conv_30 = nn.Conv2d(kernel_size=(1, 1), in_channels=512, out_channels=128) + self.n_Conv_31 = nn.Conv2d(kernel_size=(3, 3), in_channels=128, out_channels=128) + self.n_Conv_32 = nn.Conv2d(kernel_size=(1, 1), in_channels=128, out_channels=512) + self.n_Conv_33 = nn.Conv2d(kernel_size=(1, 1), in_channels=512, out_channels=128) + self.n_Conv_34 = nn.Conv2d(kernel_size=(3, 3), in_channels=128, out_channels=128) + self.n_Conv_35 = nn.Conv2d(kernel_size=(1, 1), in_channels=128, out_channels=512) + self.n_Conv_36 = nn.Conv2d(kernel_size=(1, 1), in_channels=512, out_channels=1024, stride=(2, 2)) + self.n_Conv_37 = nn.Conv2d(kernel_size=(1, 1), in_channels=512, out_channels=256) + self.n_Conv_38 = nn.Conv2d(kernel_size=(3, 3), in_channels=256, out_channels=256, stride=(2, 2)) + self.n_Conv_39 = nn.Conv2d(kernel_size=(1, 1), in_channels=256, out_channels=1024) + self.n_Conv_40 = nn.Conv2d(kernel_size=(1, 1), in_channels=1024, out_channels=256) + self.n_Conv_41 = nn.Conv2d(kernel_size=(3, 3), in_channels=256, out_channels=256) + self.n_Conv_42 = nn.Conv2d(kernel_size=(1, 1), in_channels=256, out_channels=1024) + self.n_Conv_43 = nn.Conv2d(kernel_size=(1, 1), in_channels=1024, out_channels=256) + self.n_Conv_44 = nn.Conv2d(kernel_size=(3, 3), in_channels=256, out_channels=256) + self.n_Conv_45 = nn.Conv2d(kernel_size=(1, 1), in_channels=256, out_channels=1024) + self.n_Conv_46 = nn.Conv2d(kernel_size=(1, 1), in_channels=1024, out_channels=256) + self.n_Conv_47 = nn.Conv2d(kernel_size=(3, 3), in_channels=256, out_channels=256) + self.n_Conv_48 = nn.Conv2d(kernel_size=(1, 1), in_channels=256, out_channels=1024) + self.n_Conv_49 = nn.Conv2d(kernel_size=(1, 1), in_channels=1024, out_channels=256) + self.n_Conv_50 = nn.Conv2d(kernel_size=(3, 3), in_channels=256, out_channels=256) + self.n_Conv_51 = nn.Conv2d(kernel_size=(1, 1), in_channels=256, out_channels=1024) + self.n_Conv_52 = nn.Conv2d(kernel_size=(1, 1), in_channels=1024, out_channels=256) + self.n_Conv_53 = nn.Conv2d(kernel_size=(3, 3), in_channels=256, out_channels=256) + self.n_Conv_54 = nn.Conv2d(kernel_size=(1, 1), in_channels=256, out_channels=1024) + self.n_Conv_55 = nn.Conv2d(kernel_size=(1, 1), in_channels=1024, out_channels=256) + self.n_Conv_56 = nn.Conv2d(kernel_size=(3, 3), in_channels=256, out_channels=256) + self.n_Conv_57 = nn.Conv2d(kernel_size=(1, 1), in_channels=256, out_channels=1024) + self.n_Conv_58 = nn.Conv2d(kernel_size=(1, 1), in_channels=1024, out_channels=256) + self.n_Conv_59 = nn.Conv2d(kernel_size=(3, 3), in_channels=256, out_channels=256) + self.n_Conv_60 = nn.Conv2d(kernel_size=(1, 1), in_channels=256, out_channels=1024) + self.n_Conv_61 = nn.Conv2d(kernel_size=(1, 1), in_channels=1024, out_channels=256) + self.n_Conv_62 = nn.Conv2d(kernel_size=(3, 3), in_channels=256, out_channels=256) + self.n_Conv_63 = nn.Conv2d(kernel_size=(1, 1), in_channels=256, out_channels=1024) + self.n_Conv_64 = nn.Conv2d(kernel_size=(1, 1), in_channels=1024, out_channels=256) + self.n_Conv_65 = nn.Conv2d(kernel_size=(3, 3), in_channels=256, out_channels=256) + self.n_Conv_66 = nn.Conv2d(kernel_size=(1, 1), in_channels=256, out_channels=1024) + self.n_Conv_67 = nn.Conv2d(kernel_size=(1, 1), in_channels=1024, out_channels=256) + self.n_Conv_68 = nn.Conv2d(kernel_size=(3, 3), in_channels=256, out_channels=256) + self.n_Conv_69 = nn.Conv2d(kernel_size=(1, 1), in_channels=256, out_channels=1024) + self.n_Conv_70 = nn.Conv2d(kernel_size=(1, 1), in_channels=1024, out_channels=256) + self.n_Conv_71 = nn.Conv2d(kernel_size=(3, 3), in_channels=256, out_channels=256) + self.n_Conv_72 = nn.Conv2d(kernel_size=(1, 1), in_channels=256, out_channels=1024) + self.n_Conv_73 = nn.Conv2d(kernel_size=(1, 1), in_channels=1024, out_channels=256) + self.n_Conv_74 = nn.Conv2d(kernel_size=(3, 3), in_channels=256, out_channels=256) + self.n_Conv_75 = nn.Conv2d(kernel_size=(1, 1), in_channels=256, out_channels=1024) + self.n_Conv_76 = nn.Conv2d(kernel_size=(1, 1), in_channels=1024, out_channels=256) + self.n_Conv_77 = nn.Conv2d(kernel_size=(3, 3), in_channels=256, out_channels=256) + self.n_Conv_78 = nn.Conv2d(kernel_size=(1, 1), in_channels=256, out_channels=1024) + self.n_Conv_79 = nn.Conv2d(kernel_size=(1, 1), in_channels=1024, out_channels=256) + self.n_Conv_80 = nn.Conv2d(kernel_size=(3, 3), in_channels=256, out_channels=256) + self.n_Conv_81 = nn.Conv2d(kernel_size=(1, 1), in_channels=256, out_channels=1024) + self.n_Conv_82 = nn.Conv2d(kernel_size=(1, 1), in_channels=1024, out_channels=256) + self.n_Conv_83 = nn.Conv2d(kernel_size=(3, 3), in_channels=256, out_channels=256) + self.n_Conv_84 = nn.Conv2d(kernel_size=(1, 1), in_channels=256, out_channels=1024) + self.n_Conv_85 = nn.Conv2d(kernel_size=(1, 1), in_channels=1024, out_channels=256) + self.n_Conv_86 = nn.Conv2d(kernel_size=(3, 3), in_channels=256, out_channels=256) + self.n_Conv_87 = nn.Conv2d(kernel_size=(1, 1), in_channels=256, out_channels=1024) + self.n_Conv_88 = nn.Conv2d(kernel_size=(1, 1), in_channels=1024, out_channels=256) + self.n_Conv_89 = nn.Conv2d(kernel_size=(3, 3), in_channels=256, out_channels=256) + self.n_Conv_90 = nn.Conv2d(kernel_size=(1, 1), in_channels=256, out_channels=1024) + self.n_Conv_91 = nn.Conv2d(kernel_size=(1, 1), in_channels=1024, out_channels=256) + self.n_Conv_92 = nn.Conv2d(kernel_size=(3, 3), in_channels=256, out_channels=256) + self.n_Conv_93 = nn.Conv2d(kernel_size=(1, 1), in_channels=256, out_channels=1024) + self.n_Conv_94 = nn.Conv2d(kernel_size=(1, 1), in_channels=1024, out_channels=256) + self.n_Conv_95 = nn.Conv2d(kernel_size=(3, 3), in_channels=256, out_channels=256) + self.n_Conv_96 = nn.Conv2d(kernel_size=(1, 1), in_channels=256, out_channels=1024) + self.n_Conv_97 = nn.Conv2d(kernel_size=(1, 1), in_channels=1024, out_channels=256) + self.n_Conv_98 = nn.Conv2d(kernel_size=(3, 3), in_channels=256, out_channels=256, stride=(2, 2)) + self.n_Conv_99 = nn.Conv2d(kernel_size=(1, 1), in_channels=256, out_channels=1024) + self.n_Conv_100 = nn.Conv2d(kernel_size=(1, 1), in_channels=1024, out_channels=1024, stride=(2, 2)) + self.n_Conv_101 = nn.Conv2d(kernel_size=(1, 1), in_channels=1024, out_channels=256) + self.n_Conv_102 = nn.Conv2d(kernel_size=(3, 3), in_channels=256, out_channels=256) + self.n_Conv_103 = nn.Conv2d(kernel_size=(1, 1), in_channels=256, out_channels=1024) + self.n_Conv_104 = nn.Conv2d(kernel_size=(1, 1), in_channels=1024, out_channels=256) + self.n_Conv_105 = nn.Conv2d(kernel_size=(3, 3), in_channels=256, out_channels=256) + self.n_Conv_106 = nn.Conv2d(kernel_size=(1, 1), in_channels=256, out_channels=1024) + self.n_Conv_107 = nn.Conv2d(kernel_size=(1, 1), in_channels=1024, out_channels=256) + self.n_Conv_108 = nn.Conv2d(kernel_size=(3, 3), in_channels=256, out_channels=256) + self.n_Conv_109 = nn.Conv2d(kernel_size=(1, 1), in_channels=256, out_channels=1024) + self.n_Conv_110 = nn.Conv2d(kernel_size=(1, 1), in_channels=1024, out_channels=256) + self.n_Conv_111 = nn.Conv2d(kernel_size=(3, 3), in_channels=256, out_channels=256) + self.n_Conv_112 = nn.Conv2d(kernel_size=(1, 1), in_channels=256, out_channels=1024) + self.n_Conv_113 = nn.Conv2d(kernel_size=(1, 1), in_channels=1024, out_channels=256) + self.n_Conv_114 = nn.Conv2d(kernel_size=(3, 3), in_channels=256, out_channels=256) + self.n_Conv_115 = nn.Conv2d(kernel_size=(1, 1), in_channels=256, out_channels=1024) + self.n_Conv_116 = nn.Conv2d(kernel_size=(1, 1), in_channels=1024, out_channels=256) + self.n_Conv_117 = nn.Conv2d(kernel_size=(3, 3), in_channels=256, out_channels=256) + self.n_Conv_118 = nn.Conv2d(kernel_size=(1, 1), in_channels=256, out_channels=1024) + self.n_Conv_119 = nn.Conv2d(kernel_size=(1, 1), in_channels=1024, out_channels=256) + self.n_Conv_120 = nn.Conv2d(kernel_size=(3, 3), in_channels=256, out_channels=256) + self.n_Conv_121 = nn.Conv2d(kernel_size=(1, 1), in_channels=256, out_channels=1024) + self.n_Conv_122 = nn.Conv2d(kernel_size=(1, 1), in_channels=1024, out_channels=256) + self.n_Conv_123 = nn.Conv2d(kernel_size=(3, 3), in_channels=256, out_channels=256) + self.n_Conv_124 = nn.Conv2d(kernel_size=(1, 1), in_channels=256, out_channels=1024) + self.n_Conv_125 = nn.Conv2d(kernel_size=(1, 1), in_channels=1024, out_channels=256) + self.n_Conv_126 = nn.Conv2d(kernel_size=(3, 3), in_channels=256, out_channels=256) + self.n_Conv_127 = nn.Conv2d(kernel_size=(1, 1), in_channels=256, out_channels=1024) + self.n_Conv_128 = nn.Conv2d(kernel_size=(1, 1), in_channels=1024, out_channels=256) + self.n_Conv_129 = nn.Conv2d(kernel_size=(3, 3), in_channels=256, out_channels=256) + self.n_Conv_130 = nn.Conv2d(kernel_size=(1, 1), in_channels=256, out_channels=1024) + self.n_Conv_131 = nn.Conv2d(kernel_size=(1, 1), in_channels=1024, out_channels=256) + self.n_Conv_132 = nn.Conv2d(kernel_size=(3, 3), in_channels=256, out_channels=256) + self.n_Conv_133 = nn.Conv2d(kernel_size=(1, 1), in_channels=256, out_channels=1024) + self.n_Conv_134 = nn.Conv2d(kernel_size=(1, 1), in_channels=1024, out_channels=256) + self.n_Conv_135 = nn.Conv2d(kernel_size=(3, 3), in_channels=256, out_channels=256) + self.n_Conv_136 = nn.Conv2d(kernel_size=(1, 1), in_channels=256, out_channels=1024) + self.n_Conv_137 = nn.Conv2d(kernel_size=(1, 1), in_channels=1024, out_channels=256) + self.n_Conv_138 = nn.Conv2d(kernel_size=(3, 3), in_channels=256, out_channels=256) + self.n_Conv_139 = nn.Conv2d(kernel_size=(1, 1), in_channels=256, out_channels=1024) + self.n_Conv_140 = nn.Conv2d(kernel_size=(1, 1), in_channels=1024, out_channels=256) + self.n_Conv_141 = nn.Conv2d(kernel_size=(3, 3), in_channels=256, out_channels=256) + self.n_Conv_142 = nn.Conv2d(kernel_size=(1, 1), in_channels=256, out_channels=1024) + self.n_Conv_143 = nn.Conv2d(kernel_size=(1, 1), in_channels=1024, out_channels=256) + self.n_Conv_144 = nn.Conv2d(kernel_size=(3, 3), in_channels=256, out_channels=256) + self.n_Conv_145 = nn.Conv2d(kernel_size=(1, 1), in_channels=256, out_channels=1024) + self.n_Conv_146 = nn.Conv2d(kernel_size=(1, 1), in_channels=1024, out_channels=256) + self.n_Conv_147 = nn.Conv2d(kernel_size=(3, 3), in_channels=256, out_channels=256) + self.n_Conv_148 = nn.Conv2d(kernel_size=(1, 1), in_channels=256, out_channels=1024) + self.n_Conv_149 = nn.Conv2d(kernel_size=(1, 1), in_channels=1024, out_channels=256) + self.n_Conv_150 = nn.Conv2d(kernel_size=(3, 3), in_channels=256, out_channels=256) + self.n_Conv_151 = nn.Conv2d(kernel_size=(1, 1), in_channels=256, out_channels=1024) + self.n_Conv_152 = nn.Conv2d(kernel_size=(1, 1), in_channels=1024, out_channels=256) + self.n_Conv_153 = nn.Conv2d(kernel_size=(3, 3), in_channels=256, out_channels=256) + self.n_Conv_154 = nn.Conv2d(kernel_size=(1, 1), in_channels=256, out_channels=1024) + self.n_Conv_155 = nn.Conv2d(kernel_size=(1, 1), in_channels=1024, out_channels=256) + self.n_Conv_156 = nn.Conv2d(kernel_size=(3, 3), in_channels=256, out_channels=256) + self.n_Conv_157 = nn.Conv2d(kernel_size=(1, 1), in_channels=256, out_channels=1024) + self.n_Conv_158 = nn.Conv2d(kernel_size=(1, 1), in_channels=1024, out_channels=2048, stride=(2, 2)) + self.n_Conv_159 = nn.Conv2d(kernel_size=(1, 1), in_channels=1024, out_channels=512) + self.n_Conv_160 = nn.Conv2d(kernel_size=(3, 3), in_channels=512, out_channels=512, stride=(2, 2)) + self.n_Conv_161 = nn.Conv2d(kernel_size=(1, 1), in_channels=512, out_channels=2048) + self.n_Conv_162 = nn.Conv2d(kernel_size=(1, 1), in_channels=2048, out_channels=512) + self.n_Conv_163 = nn.Conv2d(kernel_size=(3, 3), in_channels=512, out_channels=512) + self.n_Conv_164 = nn.Conv2d(kernel_size=(1, 1), in_channels=512, out_channels=2048) + self.n_Conv_165 = nn.Conv2d(kernel_size=(1, 1), in_channels=2048, out_channels=512) + self.n_Conv_166 = nn.Conv2d(kernel_size=(3, 3), in_channels=512, out_channels=512) + self.n_Conv_167 = nn.Conv2d(kernel_size=(1, 1), in_channels=512, out_channels=2048) + self.n_Conv_168 = nn.Conv2d(kernel_size=(1, 1), in_channels=2048, out_channels=4096, stride=(2, 2)) + self.n_Conv_169 = nn.Conv2d(kernel_size=(1, 1), in_channels=2048, out_channels=1024) + self.n_Conv_170 = nn.Conv2d(kernel_size=(3, 3), in_channels=1024, out_channels=1024, stride=(2, 2)) + self.n_Conv_171 = nn.Conv2d(kernel_size=(1, 1), in_channels=1024, out_channels=4096) + self.n_Conv_172 = nn.Conv2d(kernel_size=(1, 1), in_channels=4096, out_channels=1024) + self.n_Conv_173 = nn.Conv2d(kernel_size=(3, 3), in_channels=1024, out_channels=1024) + self.n_Conv_174 = nn.Conv2d(kernel_size=(1, 1), in_channels=1024, out_channels=4096) + self.n_Conv_175 = nn.Conv2d(kernel_size=(1, 1), in_channels=4096, out_channels=1024) + self.n_Conv_176 = nn.Conv2d(kernel_size=(3, 3), in_channels=1024, out_channels=1024) + self.n_Conv_177 = nn.Conv2d(kernel_size=(1, 1), in_channels=1024, out_channels=4096) + self.n_Conv_178 = nn.Conv2d(kernel_size=(1, 1), in_channels=4096, out_channels=9176, bias=False) + + def forward(self, *inputs): + t_358, = inputs + t_359 = t_358.permute(*[0, 3, 1, 2]) + t_359_padded = F.pad(t_359, [2, 3, 2, 3], value=0) + t_360 = self.n_Conv_0(t_359_padded) + t_361 = F.relu(t_360) + t_361 = F.pad(t_361, [0, 1, 0, 1], value=float('-inf')) + t_362 = self.n_MaxPool_0(t_361) + t_363 = self.n_Conv_1(t_362) + t_364 = self.n_Conv_2(t_362) + t_365 = F.relu(t_364) + t_365_padded = F.pad(t_365, [1, 1, 1, 1], value=0) + t_366 = self.n_Conv_3(t_365_padded) + t_367 = F.relu(t_366) + t_368 = self.n_Conv_4(t_367) + t_369 = torch.add(t_368, t_363) + t_370 = F.relu(t_369) + t_371 = self.n_Conv_5(t_370) + t_372 = F.relu(t_371) + t_372_padded = F.pad(t_372, [1, 1, 1, 1], value=0) + t_373 = self.n_Conv_6(t_372_padded) + t_374 = F.relu(t_373) + t_375 = self.n_Conv_7(t_374) + t_376 = torch.add(t_375, t_370) + t_377 = F.relu(t_376) + t_378 = self.n_Conv_8(t_377) + t_379 = F.relu(t_378) + t_379_padded = F.pad(t_379, [1, 1, 1, 1], value=0) + t_380 = self.n_Conv_9(t_379_padded) + t_381 = F.relu(t_380) + t_382 = self.n_Conv_10(t_381) + t_383 = torch.add(t_382, t_377) + t_384 = F.relu(t_383) + t_385 = self.n_Conv_11(t_384) + t_386 = self.n_Conv_12(t_384) + t_387 = F.relu(t_386) + t_387_padded = F.pad(t_387, [0, 1, 0, 1], value=0) + t_388 = self.n_Conv_13(t_387_padded) + t_389 = F.relu(t_388) + t_390 = self.n_Conv_14(t_389) + t_391 = torch.add(t_390, t_385) + t_392 = F.relu(t_391) + t_393 = self.n_Conv_15(t_392) + t_394 = F.relu(t_393) + t_394_padded = F.pad(t_394, [1, 1, 1, 1], value=0) + t_395 = self.n_Conv_16(t_394_padded) + t_396 = F.relu(t_395) + t_397 = self.n_Conv_17(t_396) + t_398 = torch.add(t_397, t_392) + t_399 = F.relu(t_398) + t_400 = self.n_Conv_18(t_399) + t_401 = F.relu(t_400) + t_401_padded = F.pad(t_401, [1, 1, 1, 1], value=0) + t_402 = self.n_Conv_19(t_401_padded) + t_403 = F.relu(t_402) + t_404 = self.n_Conv_20(t_403) + t_405 = torch.add(t_404, t_399) + t_406 = F.relu(t_405) + t_407 = self.n_Conv_21(t_406) + t_408 = F.relu(t_407) + t_408_padded = F.pad(t_408, [1, 1, 1, 1], value=0) + t_409 = self.n_Conv_22(t_408_padded) + t_410 = F.relu(t_409) + t_411 = self.n_Conv_23(t_410) + t_412 = torch.add(t_411, t_406) + t_413 = F.relu(t_412) + t_414 = self.n_Conv_24(t_413) + t_415 = F.relu(t_414) + t_415_padded = F.pad(t_415, [1, 1, 1, 1], value=0) + t_416 = self.n_Conv_25(t_415_padded) + t_417 = F.relu(t_416) + t_418 = self.n_Conv_26(t_417) + t_419 = torch.add(t_418, t_413) + t_420 = F.relu(t_419) + t_421 = self.n_Conv_27(t_420) + t_422 = F.relu(t_421) + t_422_padded = F.pad(t_422, [1, 1, 1, 1], value=0) + t_423 = self.n_Conv_28(t_422_padded) + t_424 = F.relu(t_423) + t_425 = self.n_Conv_29(t_424) + t_426 = torch.add(t_425, t_420) + t_427 = F.relu(t_426) + t_428 = self.n_Conv_30(t_427) + t_429 = F.relu(t_428) + t_429_padded = F.pad(t_429, [1, 1, 1, 1], value=0) + t_430 = self.n_Conv_31(t_429_padded) + t_431 = F.relu(t_430) + t_432 = self.n_Conv_32(t_431) + t_433 = torch.add(t_432, t_427) + t_434 = F.relu(t_433) + t_435 = self.n_Conv_33(t_434) + t_436 = F.relu(t_435) + t_436_padded = F.pad(t_436, [1, 1, 1, 1], value=0) + t_437 = self.n_Conv_34(t_436_padded) + t_438 = F.relu(t_437) + t_439 = self.n_Conv_35(t_438) + t_440 = torch.add(t_439, t_434) + t_441 = F.relu(t_440) + t_442 = self.n_Conv_36(t_441) + t_443 = self.n_Conv_37(t_441) + t_444 = F.relu(t_443) + t_444_padded = F.pad(t_444, [0, 1, 0, 1], value=0) + t_445 = self.n_Conv_38(t_444_padded) + t_446 = F.relu(t_445) + t_447 = self.n_Conv_39(t_446) + t_448 = torch.add(t_447, t_442) + t_449 = F.relu(t_448) + t_450 = self.n_Conv_40(t_449) + t_451 = F.relu(t_450) + t_451_padded = F.pad(t_451, [1, 1, 1, 1], value=0) + t_452 = self.n_Conv_41(t_451_padded) + t_453 = F.relu(t_452) + t_454 = self.n_Conv_42(t_453) + t_455 = torch.add(t_454, t_449) + t_456 = F.relu(t_455) + t_457 = self.n_Conv_43(t_456) + t_458 = F.relu(t_457) + t_458_padded = F.pad(t_458, [1, 1, 1, 1], value=0) + t_459 = self.n_Conv_44(t_458_padded) + t_460 = F.relu(t_459) + t_461 = self.n_Conv_45(t_460) + t_462 = torch.add(t_461, t_456) + t_463 = F.relu(t_462) + t_464 = self.n_Conv_46(t_463) + t_465 = F.relu(t_464) + t_465_padded = F.pad(t_465, [1, 1, 1, 1], value=0) + t_466 = self.n_Conv_47(t_465_padded) + t_467 = F.relu(t_466) + t_468 = self.n_Conv_48(t_467) + t_469 = torch.add(t_468, t_463) + t_470 = F.relu(t_469) + t_471 = self.n_Conv_49(t_470) + t_472 = F.relu(t_471) + t_472_padded = F.pad(t_472, [1, 1, 1, 1], value=0) + t_473 = self.n_Conv_50(t_472_padded) + t_474 = F.relu(t_473) + t_475 = self.n_Conv_51(t_474) + t_476 = torch.add(t_475, t_470) + t_477 = F.relu(t_476) + t_478 = self.n_Conv_52(t_477) + t_479 = F.relu(t_478) + t_479_padded = F.pad(t_479, [1, 1, 1, 1], value=0) + t_480 = self.n_Conv_53(t_479_padded) + t_481 = F.relu(t_480) + t_482 = self.n_Conv_54(t_481) + t_483 = torch.add(t_482, t_477) + t_484 = F.relu(t_483) + t_485 = self.n_Conv_55(t_484) + t_486 = F.relu(t_485) + t_486_padded = F.pad(t_486, [1, 1, 1, 1], value=0) + t_487 = self.n_Conv_56(t_486_padded) + t_488 = F.relu(t_487) + t_489 = self.n_Conv_57(t_488) + t_490 = torch.add(t_489, t_484) + t_491 = F.relu(t_490) + t_492 = self.n_Conv_58(t_491) + t_493 = F.relu(t_492) + t_493_padded = F.pad(t_493, [1, 1, 1, 1], value=0) + t_494 = self.n_Conv_59(t_493_padded) + t_495 = F.relu(t_494) + t_496 = self.n_Conv_60(t_495) + t_497 = torch.add(t_496, t_491) + t_498 = F.relu(t_497) + t_499 = self.n_Conv_61(t_498) + t_500 = F.relu(t_499) + t_500_padded = F.pad(t_500, [1, 1, 1, 1], value=0) + t_501 = self.n_Conv_62(t_500_padded) + t_502 = F.relu(t_501) + t_503 = self.n_Conv_63(t_502) + t_504 = torch.add(t_503, t_498) + t_505 = F.relu(t_504) + t_506 = self.n_Conv_64(t_505) + t_507 = F.relu(t_506) + t_507_padded = F.pad(t_507, [1, 1, 1, 1], value=0) + t_508 = self.n_Conv_65(t_507_padded) + t_509 = F.relu(t_508) + t_510 = self.n_Conv_66(t_509) + t_511 = torch.add(t_510, t_505) + t_512 = F.relu(t_511) + t_513 = self.n_Conv_67(t_512) + t_514 = F.relu(t_513) + t_514_padded = F.pad(t_514, [1, 1, 1, 1], value=0) + t_515 = self.n_Conv_68(t_514_padded) + t_516 = F.relu(t_515) + t_517 = self.n_Conv_69(t_516) + t_518 = torch.add(t_517, t_512) + t_519 = F.relu(t_518) + t_520 = self.n_Conv_70(t_519) + t_521 = F.relu(t_520) + t_521_padded = F.pad(t_521, [1, 1, 1, 1], value=0) + t_522 = self.n_Conv_71(t_521_padded) + t_523 = F.relu(t_522) + t_524 = self.n_Conv_72(t_523) + t_525 = torch.add(t_524, t_519) + t_526 = F.relu(t_525) + t_527 = self.n_Conv_73(t_526) + t_528 = F.relu(t_527) + t_528_padded = F.pad(t_528, [1, 1, 1, 1], value=0) + t_529 = self.n_Conv_74(t_528_padded) + t_530 = F.relu(t_529) + t_531 = self.n_Conv_75(t_530) + t_532 = torch.add(t_531, t_526) + t_533 = F.relu(t_532) + t_534 = self.n_Conv_76(t_533) + t_535 = F.relu(t_534) + t_535_padded = F.pad(t_535, [1, 1, 1, 1], value=0) + t_536 = self.n_Conv_77(t_535_padded) + t_537 = F.relu(t_536) + t_538 = self.n_Conv_78(t_537) + t_539 = torch.add(t_538, t_533) + t_540 = F.relu(t_539) + t_541 = self.n_Conv_79(t_540) + t_542 = F.relu(t_541) + t_542_padded = F.pad(t_542, [1, 1, 1, 1], value=0) + t_543 = self.n_Conv_80(t_542_padded) + t_544 = F.relu(t_543) + t_545 = self.n_Conv_81(t_544) + t_546 = torch.add(t_545, t_540) + t_547 = F.relu(t_546) + t_548 = self.n_Conv_82(t_547) + t_549 = F.relu(t_548) + t_549_padded = F.pad(t_549, [1, 1, 1, 1], value=0) + t_550 = self.n_Conv_83(t_549_padded) + t_551 = F.relu(t_550) + t_552 = self.n_Conv_84(t_551) + t_553 = torch.add(t_552, t_547) + t_554 = F.relu(t_553) + t_555 = self.n_Conv_85(t_554) + t_556 = F.relu(t_555) + t_556_padded = F.pad(t_556, [1, 1, 1, 1], value=0) + t_557 = self.n_Conv_86(t_556_padded) + t_558 = F.relu(t_557) + t_559 = self.n_Conv_87(t_558) + t_560 = torch.add(t_559, t_554) + t_561 = F.relu(t_560) + t_562 = self.n_Conv_88(t_561) + t_563 = F.relu(t_562) + t_563_padded = F.pad(t_563, [1, 1, 1, 1], value=0) + t_564 = self.n_Conv_89(t_563_padded) + t_565 = F.relu(t_564) + t_566 = self.n_Conv_90(t_565) + t_567 = torch.add(t_566, t_561) + t_568 = F.relu(t_567) + t_569 = self.n_Conv_91(t_568) + t_570 = F.relu(t_569) + t_570_padded = F.pad(t_570, [1, 1, 1, 1], value=0) + t_571 = self.n_Conv_92(t_570_padded) + t_572 = F.relu(t_571) + t_573 = self.n_Conv_93(t_572) + t_574 = torch.add(t_573, t_568) + t_575 = F.relu(t_574) + t_576 = self.n_Conv_94(t_575) + t_577 = F.relu(t_576) + t_577_padded = F.pad(t_577, [1, 1, 1, 1], value=0) + t_578 = self.n_Conv_95(t_577_padded) + t_579 = F.relu(t_578) + t_580 = self.n_Conv_96(t_579) + t_581 = torch.add(t_580, t_575) + t_582 = F.relu(t_581) + t_583 = self.n_Conv_97(t_582) + t_584 = F.relu(t_583) + t_584_padded = F.pad(t_584, [0, 1, 0, 1], value=0) + t_585 = self.n_Conv_98(t_584_padded) + t_586 = F.relu(t_585) + t_587 = self.n_Conv_99(t_586) + t_588 = self.n_Conv_100(t_582) + t_589 = torch.add(t_587, t_588) + t_590 = F.relu(t_589) + t_591 = self.n_Conv_101(t_590) + t_592 = F.relu(t_591) + t_592_padded = F.pad(t_592, [1, 1, 1, 1], value=0) + t_593 = self.n_Conv_102(t_592_padded) + t_594 = F.relu(t_593) + t_595 = self.n_Conv_103(t_594) + t_596 = torch.add(t_595, t_590) + t_597 = F.relu(t_596) + t_598 = self.n_Conv_104(t_597) + t_599 = F.relu(t_598) + t_599_padded = F.pad(t_599, [1, 1, 1, 1], value=0) + t_600 = self.n_Conv_105(t_599_padded) + t_601 = F.relu(t_600) + t_602 = self.n_Conv_106(t_601) + t_603 = torch.add(t_602, t_597) + t_604 = F.relu(t_603) + t_605 = self.n_Conv_107(t_604) + t_606 = F.relu(t_605) + t_606_padded = F.pad(t_606, [1, 1, 1, 1], value=0) + t_607 = self.n_Conv_108(t_606_padded) + t_608 = F.relu(t_607) + t_609 = self.n_Conv_109(t_608) + t_610 = torch.add(t_609, t_604) + t_611 = F.relu(t_610) + t_612 = self.n_Conv_110(t_611) + t_613 = F.relu(t_612) + t_613_padded = F.pad(t_613, [1, 1, 1, 1], value=0) + t_614 = self.n_Conv_111(t_613_padded) + t_615 = F.relu(t_614) + t_616 = self.n_Conv_112(t_615) + t_617 = torch.add(t_616, t_611) + t_618 = F.relu(t_617) + t_619 = self.n_Conv_113(t_618) + t_620 = F.relu(t_619) + t_620_padded = F.pad(t_620, [1, 1, 1, 1], value=0) + t_621 = self.n_Conv_114(t_620_padded) + t_622 = F.relu(t_621) + t_623 = self.n_Conv_115(t_622) + t_624 = torch.add(t_623, t_618) + t_625 = F.relu(t_624) + t_626 = self.n_Conv_116(t_625) + t_627 = F.relu(t_626) + t_627_padded = F.pad(t_627, [1, 1, 1, 1], value=0) + t_628 = self.n_Conv_117(t_627_padded) + t_629 = F.relu(t_628) + t_630 = self.n_Conv_118(t_629) + t_631 = torch.add(t_630, t_625) + t_632 = F.relu(t_631) + t_633 = self.n_Conv_119(t_632) + t_634 = F.relu(t_633) + t_634_padded = F.pad(t_634, [1, 1, 1, 1], value=0) + t_635 = self.n_Conv_120(t_634_padded) + t_636 = F.relu(t_635) + t_637 = self.n_Conv_121(t_636) + t_638 = torch.add(t_637, t_632) + t_639 = F.relu(t_638) + t_640 = self.n_Conv_122(t_639) + t_641 = F.relu(t_640) + t_641_padded = F.pad(t_641, [1, 1, 1, 1], value=0) + t_642 = self.n_Conv_123(t_641_padded) + t_643 = F.relu(t_642) + t_644 = self.n_Conv_124(t_643) + t_645 = torch.add(t_644, t_639) + t_646 = F.relu(t_645) + t_647 = self.n_Conv_125(t_646) + t_648 = F.relu(t_647) + t_648_padded = F.pad(t_648, [1, 1, 1, 1], value=0) + t_649 = self.n_Conv_126(t_648_padded) + t_650 = F.relu(t_649) + t_651 = self.n_Conv_127(t_650) + t_652 = torch.add(t_651, t_646) + t_653 = F.relu(t_652) + t_654 = self.n_Conv_128(t_653) + t_655 = F.relu(t_654) + t_655_padded = F.pad(t_655, [1, 1, 1, 1], value=0) + t_656 = self.n_Conv_129(t_655_padded) + t_657 = F.relu(t_656) + t_658 = self.n_Conv_130(t_657) + t_659 = torch.add(t_658, t_653) + t_660 = F.relu(t_659) + t_661 = self.n_Conv_131(t_660) + t_662 = F.relu(t_661) + t_662_padded = F.pad(t_662, [1, 1, 1, 1], value=0) + t_663 = self.n_Conv_132(t_662_padded) + t_664 = F.relu(t_663) + t_665 = self.n_Conv_133(t_664) + t_666 = torch.add(t_665, t_660) + t_667 = F.relu(t_666) + t_668 = self.n_Conv_134(t_667) + t_669 = F.relu(t_668) + t_669_padded = F.pad(t_669, [1, 1, 1, 1], value=0) + t_670 = self.n_Conv_135(t_669_padded) + t_671 = F.relu(t_670) + t_672 = self.n_Conv_136(t_671) + t_673 = torch.add(t_672, t_667) + t_674 = F.relu(t_673) + t_675 = self.n_Conv_137(t_674) + t_676 = F.relu(t_675) + t_676_padded = F.pad(t_676, [1, 1, 1, 1], value=0) + t_677 = self.n_Conv_138(t_676_padded) + t_678 = F.relu(t_677) + t_679 = self.n_Conv_139(t_678) + t_680 = torch.add(t_679, t_674) + t_681 = F.relu(t_680) + t_682 = self.n_Conv_140(t_681) + t_683 = F.relu(t_682) + t_683_padded = F.pad(t_683, [1, 1, 1, 1], value=0) + t_684 = self.n_Conv_141(t_683_padded) + t_685 = F.relu(t_684) + t_686 = self.n_Conv_142(t_685) + t_687 = torch.add(t_686, t_681) + t_688 = F.relu(t_687) + t_689 = self.n_Conv_143(t_688) + t_690 = F.relu(t_689) + t_690_padded = F.pad(t_690, [1, 1, 1, 1], value=0) + t_691 = self.n_Conv_144(t_690_padded) + t_692 = F.relu(t_691) + t_693 = self.n_Conv_145(t_692) + t_694 = torch.add(t_693, t_688) + t_695 = F.relu(t_694) + t_696 = self.n_Conv_146(t_695) + t_697 = F.relu(t_696) + t_697_padded = F.pad(t_697, [1, 1, 1, 1], value=0) + t_698 = self.n_Conv_147(t_697_padded) + t_699 = F.relu(t_698) + t_700 = self.n_Conv_148(t_699) + t_701 = torch.add(t_700, t_695) + t_702 = F.relu(t_701) + t_703 = self.n_Conv_149(t_702) + t_704 = F.relu(t_703) + t_704_padded = F.pad(t_704, [1, 1, 1, 1], value=0) + t_705 = self.n_Conv_150(t_704_padded) + t_706 = F.relu(t_705) + t_707 = self.n_Conv_151(t_706) + t_708 = torch.add(t_707, t_702) + t_709 = F.relu(t_708) + t_710 = self.n_Conv_152(t_709) + t_711 = F.relu(t_710) + t_711_padded = F.pad(t_711, [1, 1, 1, 1], value=0) + t_712 = self.n_Conv_153(t_711_padded) + t_713 = F.relu(t_712) + t_714 = self.n_Conv_154(t_713) + t_715 = torch.add(t_714, t_709) + t_716 = F.relu(t_715) + t_717 = self.n_Conv_155(t_716) + t_718 = F.relu(t_717) + t_718_padded = F.pad(t_718, [1, 1, 1, 1], value=0) + t_719 = self.n_Conv_156(t_718_padded) + t_720 = F.relu(t_719) + t_721 = self.n_Conv_157(t_720) + t_722 = torch.add(t_721, t_716) + t_723 = F.relu(t_722) + t_724 = self.n_Conv_158(t_723) + t_725 = self.n_Conv_159(t_723) + t_726 = F.relu(t_725) + t_726_padded = F.pad(t_726, [0, 1, 0, 1], value=0) + t_727 = self.n_Conv_160(t_726_padded) + t_728 = F.relu(t_727) + t_729 = self.n_Conv_161(t_728) + t_730 = torch.add(t_729, t_724) + t_731 = F.relu(t_730) + t_732 = self.n_Conv_162(t_731) + t_733 = F.relu(t_732) + t_733_padded = F.pad(t_733, [1, 1, 1, 1], value=0) + t_734 = self.n_Conv_163(t_733_padded) + t_735 = F.relu(t_734) + t_736 = self.n_Conv_164(t_735) + t_737 = torch.add(t_736, t_731) + t_738 = F.relu(t_737) + t_739 = self.n_Conv_165(t_738) + t_740 = F.relu(t_739) + t_740_padded = F.pad(t_740, [1, 1, 1, 1], value=0) + t_741 = self.n_Conv_166(t_740_padded) + t_742 = F.relu(t_741) + t_743 = self.n_Conv_167(t_742) + t_744 = torch.add(t_743, t_738) + t_745 = F.relu(t_744) + t_746 = self.n_Conv_168(t_745) + t_747 = self.n_Conv_169(t_745) + t_748 = F.relu(t_747) + t_748_padded = F.pad(t_748, [0, 1, 0, 1], value=0) + t_749 = self.n_Conv_170(t_748_padded) + t_750 = F.relu(t_749) + t_751 = self.n_Conv_171(t_750) + t_752 = torch.add(t_751, t_746) + t_753 = F.relu(t_752) + t_754 = self.n_Conv_172(t_753) + t_755 = F.relu(t_754) + t_755_padded = F.pad(t_755, [1, 1, 1, 1], value=0) + t_756 = self.n_Conv_173(t_755_padded) + t_757 = F.relu(t_756) + t_758 = self.n_Conv_174(t_757) + t_759 = torch.add(t_758, t_753) + t_760 = F.relu(t_759) + t_761 = self.n_Conv_175(t_760) + t_762 = F.relu(t_761) + t_762_padded = F.pad(t_762, [1, 1, 1, 1], value=0) + t_763 = self.n_Conv_176(t_762_padded) + t_764 = F.relu(t_763) + t_765 = self.n_Conv_177(t_764) + t_766 = torch.add(t_765, t_760) + t_767 = F.relu(t_766) + t_768 = self.n_Conv_178(t_767) + t_769 = F.avg_pool2d(t_768, kernel_size=t_768.shape[-2:]) + t_770 = torch.squeeze(t_769, 3) + t_770 = torch.squeeze(t_770, 2) + t_771 = torch.sigmoid(t_770) + return t_771 + + def load_state_dict(self, state_dict, **kwargs): + self.tags = state_dict.get('tags', []) + + super(DeepDanbooruModel, self).load_state_dict({k: v for k, v in state_dict.items() if k != 'tags'}) + diff --git a/modules/shared.py b/modules/shared.py index a4457305..c93ae2a3 100644 --- a/modules/shared.py +++ b/modules/shared.py @@ -55,7 +55,7 @@ parser.add_argument("--ldsr-models-path", type=str, help="Path to directory with parser.add_argument("--clip-models-path", type=str, help="Path to directory with CLIP model file(s).", default=None) parser.add_argument("--xformers", action='store_true', help="enable xformers for cross attention layers") parser.add_argument("--force-enable-xformers", action='store_true', help="enable xformers for cross attention layers regardless of whether the checking code thinks you can run it; do not make bug reports if this fails to work") -parser.add_argument("--deepdanbooru", action='store_true', help="enable deepdanbooru interrogator") +parser.add_argument("--deepdanbooru", action='store_true', help="does not do anything") parser.add_argument("--opt-split-attention", action='store_true', help="force-enables Doggettx's cross-attention layer optimization. By default, it's on for torch cuda.") parser.add_argument("--opt-split-attention-invokeai", action='store_true', help="force-enables InvokeAI's cross-attention layer optimization. By default, it's on when cuda is unavailable.") parser.add_argument("--opt-split-attention-v1", action='store_true', help="enable older version of split attention optimization that does not consume all the VRAM it can find") diff --git a/modules/textual_inversion/preprocess.py b/modules/textual_inversion/preprocess.py index 488aa5b5..56b9b2eb 100644 --- a/modules/textual_inversion/preprocess.py +++ b/modules/textual_inversion/preprocess.py @@ -6,12 +6,10 @@ import sys import tqdm import time -from modules import shared, images +from modules import shared, images, deepbooru from modules.paths import models_path from modules.shared import opts, cmd_opts from modules.textual_inversion import autocrop -if cmd_opts.deepdanbooru: - import modules.deepbooru as deepbooru def preprocess(process_src, process_dst, process_width, process_height, preprocess_txt_action, process_flip, process_split, process_caption, process_caption_deepbooru=False, split_threshold=0.5, overlap_ratio=0.2, process_focal_crop=False, process_focal_crop_face_weight=0.9, process_focal_crop_entropy_weight=0.3, process_focal_crop_edges_weight=0.5, process_focal_crop_debug=False): @@ -20,9 +18,7 @@ def preprocess(process_src, process_dst, process_width, process_height, preproce shared.interrogator.load() if process_caption_deepbooru: - db_opts = deepbooru.create_deepbooru_opts() - db_opts[deepbooru.OPT_INCLUDE_RANKS] = False - deepbooru.create_deepbooru_process(opts.interrogate_deepbooru_score_threshold, db_opts) + deepbooru.model.start() preprocess_work(process_src, process_dst, process_width, process_height, preprocess_txt_action, process_flip, process_split, process_caption, process_caption_deepbooru, split_threshold, overlap_ratio, process_focal_crop, process_focal_crop_face_weight, process_focal_crop_entropy_weight, process_focal_crop_edges_weight, process_focal_crop_debug) @@ -32,7 +28,7 @@ def preprocess(process_src, process_dst, process_width, process_height, preproce shared.interrogator.send_blip_to_ram() if process_caption_deepbooru: - deepbooru.release_process() + deepbooru.model.stop() def listfiles(dirname): @@ -58,7 +54,7 @@ def save_pic_with_caption(image, index, params: PreprocessParams, existing_capti if params.process_caption_deepbooru: if len(caption) > 0: caption += ", " - caption += deepbooru.get_tags_from_process(image) + caption += deepbooru.model.tag_multi(image) filename_part = params.src filename_part = os.path.splitext(filename_part)[0] diff --git a/modules/ui.py b/modules/ui.py index a5953fce..e6da1b2a 100644 --- a/modules/ui.py +++ b/modules/ui.py @@ -19,14 +19,11 @@ import numpy as np from PIL import Image, PngImagePlugin -from modules import sd_hijack, sd_models, localization, script_callbacks, ui_extensions +from modules import sd_hijack, sd_models, localization, script_callbacks, ui_extensions, deepbooru from modules.paths import script_path from modules.shared import opts, cmd_opts, restricted_opts -if cmd_opts.deepdanbooru: - from modules.deepbooru import get_deepbooru_tags - import modules.codeformer_model import modules.generation_parameters_copypaste as parameters_copypaste import modules.gfpgan_model @@ -352,7 +349,7 @@ def interrogate(image): def interrogate_deepbooru(image): - prompt = get_deepbooru_tags(image) + prompt = deepbooru.model.tag(image) return gr_show(True) if prompt is None else prompt -- cgit v1.2.3 From 192ddc04d6de0d780f73aa5fbaa8c66cd4642e1c Mon Sep 17 00:00:00 2001 From: Vladimir Mandic Date: Tue, 3 Jan 2023 10:34:51 -0500 Subject: add job info to modules --- modules/extras.py | 17 +++++++++++++---- modules/hypernetworks/hypernetwork.py | 1 + modules/textual_inversion/preprocess.py | 1 + modules/textual_inversion/textual_inversion.py | 1 + 4 files changed, 16 insertions(+), 4 deletions(-) (limited to 'modules/textual_inversion/preprocess.py') diff --git a/modules/extras.py b/modules/extras.py index 7e222313..d665440a 100644 --- a/modules/extras.py +++ b/modules/extras.py @@ -58,6 +58,9 @@ cached_images: LruCache = LruCache(max_size=5) def run_extras(extras_mode, resize_mode, image, image_folder, input_dir, output_dir, show_extras_results, gfpgan_visibility, codeformer_visibility, codeformer_weight, upscaling_resize, upscaling_resize_w, upscaling_resize_h, upscaling_crop, extras_upscaler_1, extras_upscaler_2, extras_upscaler_2_visibility, upscale_first: bool, save_output: bool = True): devices.torch_gc() + shared.state.begin() + shared.state.job = 'extras' + imageArr = [] # Also keep track of original file names imageNameArr = [] @@ -94,6 +97,7 @@ def run_extras(extras_mode, resize_mode, image, image_folder, input_dir, output_ # Extra operation definitions def run_gfpgan(image: Image.Image, info: str) -> Tuple[Image.Image, str]: + shared.state.job = 'extras-gfpgan' restored_img = modules.gfpgan_model.gfpgan_fix_faces(np.array(image, dtype=np.uint8)) res = Image.fromarray(restored_img) @@ -104,6 +108,7 @@ def run_extras(extras_mode, resize_mode, image, image_folder, input_dir, output_ return (res, info) def run_codeformer(image: Image.Image, info: str) -> Tuple[Image.Image, str]: + shared.state.job = 'extras-codeformer' restored_img = modules.codeformer_model.codeformer.restore(np.array(image, dtype=np.uint8), w=codeformer_weight) res = Image.fromarray(restored_img) @@ -114,6 +119,7 @@ def run_extras(extras_mode, resize_mode, image, image_folder, input_dir, output_ return (res, info) def upscale(image, scaler_index, resize, mode, resize_w, resize_h, crop): + shared.state.job = 'extras-upscale' upscaler = shared.sd_upscalers[scaler_index] res = upscaler.scaler.upscale(image, resize, upscaler.data_path) if mode == 1 and crop: @@ -180,6 +186,9 @@ def run_extras(extras_mode, resize_mode, image, image_folder, input_dir, output_ for image, image_name in zip(imageArr, imageNameArr): if image is None: return outputs, "Please select an input image.", '' + + shared.state.textinfo = f'Processing image {image_name}' + existing_pnginfo = image.info or {} image = image.convert("RGB") @@ -193,6 +202,10 @@ def run_extras(extras_mode, resize_mode, image, image_folder, input_dir, output_ else: basename = '' + if opts.enable_pnginfo: # append info before save + image.info = existing_pnginfo + image.info["extras"] = info + if save_output: # Add upscaler name as a suffix. suffix = f"-{shared.sd_upscalers[extras_upscaler_1].name}" if shared.opts.use_upscaler_name_as_suffix else "" @@ -203,10 +216,6 @@ def run_extras(extras_mode, resize_mode, image, image_folder, input_dir, output_ images.save_image(image, path=outpath, basename=basename, seed=None, prompt=None, extension=opts.samples_format, info=info, short_filename=True, no_prompt=True, grid=False, pnginfo_section_name="extras", existing_info=existing_pnginfo, forced_filename=None, suffix=suffix) - if opts.enable_pnginfo: - image.info = existing_pnginfo - image.info["extras"] = info - if extras_mode != 2 or show_extras_results : outputs.append(image) diff --git a/modules/hypernetworks/hypernetwork.py b/modules/hypernetworks/hypernetwork.py index 109e8078..450fecac 100644 --- a/modules/hypernetworks/hypernetwork.py +++ b/modules/hypernetworks/hypernetwork.py @@ -417,6 +417,7 @@ def train_hypernetwork(hypernetwork_name, learn_rate, batch_size, gradient_step, shared.loaded_hypernetwork = Hypernetwork() shared.loaded_hypernetwork.load(path) + shared.state.job = "train-hypernetwork" shared.state.textinfo = "Initializing hypernetwork training..." shared.state.job_count = steps diff --git a/modules/textual_inversion/preprocess.py b/modules/textual_inversion/preprocess.py index 56b9b2eb..feb876c6 100644 --- a/modules/textual_inversion/preprocess.py +++ b/modules/textual_inversion/preprocess.py @@ -124,6 +124,7 @@ def preprocess_work(process_src, process_dst, process_width, process_height, pre files = listfiles(src) + shared.state.job = "preprocess" shared.state.textinfo = "Preprocessing..." shared.state.job_count = len(files) diff --git a/modules/textual_inversion/textual_inversion.py b/modules/textual_inversion/textual_inversion.py index fd253477..2c1251d6 100644 --- a/modules/textual_inversion/textual_inversion.py +++ b/modules/textual_inversion/textual_inversion.py @@ -245,6 +245,7 @@ def train_embedding(embedding_name, learn_rate, batch_size, gradient_step, data_ create_image_every = create_image_every or 0 validate_train_inputs(embedding_name, learn_rate, batch_size, gradient_step, data_root, template_file, steps, save_embedding_every, create_image_every, log_directory, name="embedding") + shared.state.job = "train-embedding" shared.state.textinfo = "Initializing textual inversion training..." shared.state.job_count = steps -- cgit v1.2.3 From 3f43d8a966ba8462ba019a5ad573f94508cd45f8 Mon Sep 17 00:00:00 2001 From: Vladimir Mandic Date: Wed, 11 Jan 2023 10:28:55 -0500 Subject: set descriptions --- modules/hypernetworks/hypernetwork.py | 4 +++- modules/textual_inversion/preprocess.py | 7 ++++++- modules/textual_inversion/textual_inversion.py | 4 +++- 3 files changed, 12 insertions(+), 3 deletions(-) (limited to 'modules/textual_inversion/preprocess.py') diff --git a/modules/hypernetworks/hypernetwork.py b/modules/hypernetworks/hypernetwork.py index 300d3975..194679e8 100644 --- a/modules/hypernetworks/hypernetwork.py +++ b/modules/hypernetworks/hypernetwork.py @@ -619,7 +619,9 @@ def train_hypernetwork(hypernetwork_name, learn_rate, batch_size, gradient_step, epoch_num = hypernetwork.step // steps_per_epoch epoch_step = hypernetwork.step % steps_per_epoch - pbar.set_description(f"[Epoch {epoch_num}: {epoch_step+1}/{steps_per_epoch}]loss: {loss_step:.7f}") + description = f"Training hypernetwork [Epoch {epoch_num}: {epoch_step+1}/{steps_per_epoch}]loss: {loss_step:.7f}" + pbar.set_description(description) + shared.state.textinfo = description if hypernetwork_dir is not None and steps_done % save_hypernetwork_every == 0: # Before saving, change name to match current checkpoint. hypernetwork_name_every = f'{hypernetwork_name}-{steps_done}' diff --git a/modules/textual_inversion/preprocess.py b/modules/textual_inversion/preprocess.py index feb876c6..3c1042ad 100644 --- a/modules/textual_inversion/preprocess.py +++ b/modules/textual_inversion/preprocess.py @@ -135,7 +135,8 @@ def preprocess_work(process_src, process_dst, process_width, process_height, pre params.process_caption_deepbooru = process_caption_deepbooru params.preprocess_txt_action = preprocess_txt_action - for index, imagefile in enumerate(tqdm.tqdm(files)): + pbar = tqdm.tqdm(files) + for index, imagefile in enumerate(pbar): params.subindex = 0 filename = os.path.join(src, imagefile) try: @@ -143,6 +144,10 @@ def preprocess_work(process_src, process_dst, process_width, process_height, pre except Exception: continue + description = f"Preprocessing [Image {index}/{len(files)}]" + pbar.set_description(description) + shared.state.textinfo = description + params.src = filename existing_caption = None diff --git a/modules/textual_inversion/textual_inversion.py b/modules/textual_inversion/textual_inversion.py index 3866c154..b915b091 100644 --- a/modules/textual_inversion/textual_inversion.py +++ b/modules/textual_inversion/textual_inversion.py @@ -476,7 +476,9 @@ def train_embedding(embedding_name, learn_rate, batch_size, gradient_step, data_ epoch_num = embedding.step // steps_per_epoch epoch_step = embedding.step % steps_per_epoch - pbar.set_description(f"[Epoch {epoch_num}: {epoch_step+1}/{steps_per_epoch}]loss: {loss_step:.7f}") + description = f"Training textual inversion [Epoch {epoch_num}: {epoch_step+1}/{steps_per_epoch}]loss: {loss_step:.7f}" + pbar.set_description(description) + shared.state.textinfo = description if embedding_dir is not None and steps_done % save_embedding_every == 0: # Before saving, change name to match current checkpoint. embedding_name_every = f'{embedding_name}-{steps_done}' -- cgit v1.2.3