diff options
Diffstat (limited to 'modules/textual_inversion')
| -rw-r--r-- | modules/textual_inversion/dataset.py | 4 | ||||
| -rw-r--r-- | modules/textual_inversion/image_embedding.py | 5 | ||||
| -rw-r--r-- | modules/textual_inversion/preprocess.py | 85 | ||||
| -rw-r--r-- | modules/textual_inversion/textual_inversion.py | 6 | ||||
| -rw-r--r-- | modules/textual_inversion/ui.py | 4 |
5 files changed, 69 insertions, 35 deletions
diff --git a/modules/textual_inversion/dataset.py b/modules/textual_inversion/dataset.py index 23bb4b6a..5b1c5002 100644 --- a/modules/textual_inversion/dataset.py +++ b/modules/textual_inversion/dataset.py @@ -83,7 +83,7 @@ class PersonalizedBase(Dataset): self.dataset.append(entry)
- assert len(self.dataset) > 1, "No images have been found in the dataset."
+ assert len(self.dataset) > 0, "No images have been found in the dataset."
self.length = len(self.dataset) * repeats // batch_size
self.initial_indexes = np.arange(len(self.dataset))
@@ -91,7 +91,7 @@ class PersonalizedBase(Dataset): self.shuffle()
def shuffle(self):
- self.indexes = self.initial_indexes[torch.randperm(self.initial_indexes.shape[0])]
+ self.indexes = self.initial_indexes[torch.randperm(self.initial_indexes.shape[0]).numpy()]
def create_text(self, filename_text):
text = random.choice(self.lines)
diff --git a/modules/textual_inversion/image_embedding.py b/modules/textual_inversion/image_embedding.py index 898ce3b3..ea653806 100644 --- a/modules/textual_inversion/image_embedding.py +++ b/modules/textual_inversion/image_embedding.py @@ -5,6 +5,7 @@ import zlib from PIL import Image, PngImagePlugin, ImageDraw, ImageFont
from fonts.ttf import Roboto
import torch
+from modules.shared import opts
class EmbeddingEncoder(json.JSONEncoder):
@@ -133,7 +134,7 @@ def caption_image_overlay(srcimage, title, footerLeft, footerMid, footerRight, t from math import cos
image = srcimage.copy()
-
+ fontsize = 32
if textfont is None:
try:
textfont = ImageFont.truetype(opts.font or Roboto, fontsize)
@@ -150,7 +151,7 @@ def caption_image_overlay(srcimage, title, footerLeft, footerMid, footerRight, t image = Image.alpha_composite(image.convert('RGBA'), gradient.resize(image.size))
draw = ImageDraw.Draw(image)
- fontsize = 32
+
font = ImageFont.truetype(textfont, fontsize)
padding = 10
diff --git a/modules/textual_inversion/preprocess.py b/modules/textual_inversion/preprocess.py index 886cf0c3..33eaddb6 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
@@ -11,7 +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, preprocess_txt_action, 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()
@@ -21,7 +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, preprocess_txt_action, process_flip, process_split, process_caption, process_caption_deepbooru, split_threshold, overlap_ratio)
finally:
@@ -33,11 +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, preprocess_txt_action, 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 = 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'
@@ -48,7 +51,7 @@ 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 +69,49 @@ 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):
- save_pic_with_caption(image, index)
+ 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)
+ 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
for index, imagefile in enumerate(tqdm.tqdm(files)):
subindex = [0]
@@ -86,31 +121,27 @@ def preprocess_work(process_src, process_dst, process_width, process_height, pro except Exception:
continue
+ existing_caption = None
+ 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
- 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, 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()
diff --git a/modules/textual_inversion/textual_inversion.py b/modules/textual_inversion/textual_inversion.py index 3be69562..529ed3e2 100644 --- a/modules/textual_inversion/textual_inversion.py +++ b/modules/textual_inversion/textual_inversion.py @@ -153,7 +153,7 @@ class EmbeddingDatabase: return None, None
-def create_embedding(name, num_vectors_per_token, init_text='*'):
+def create_embedding(name, num_vectors_per_token, overwrite_old, init_text='*'):
cond_model = shared.sd_model.cond_stage_model
embedding_layer = cond_model.wrapped.transformer.text_model.embeddings
@@ -165,7 +165,8 @@ def create_embedding(name, num_vectors_per_token, init_text='*'): vec[i] = embedded[i * int(embedded.shape[0]) // num_vectors_per_token]
fn = os.path.join(shared.cmd_opts.embeddings_dir, f"{name}.pt")
- assert not os.path.exists(fn), f"file {fn} already exists"
+ if not overwrite_old:
+ assert not os.path.exists(fn), f"file {fn} already exists"
embedding = Embedding(vec, name)
embedding.step = 0
@@ -275,6 +276,7 @@ def train_embedding(embedding_name, learn_rate, batch_size, data_root, log_direc loss.backward()
optimizer.step()
+
epoch_num = embedding.step // len(ds)
epoch_step = embedding.step - (epoch_num * len(ds)) + 1
diff --git a/modules/textual_inversion/ui.py b/modules/textual_inversion/ui.py index 36881e7a..e712284d 100644 --- a/modules/textual_inversion/ui.py +++ b/modules/textual_inversion/ui.py @@ -7,8 +7,8 @@ import modules.textual_inversion.preprocess from modules import sd_hijack, shared
-def create_embedding(name, initialization_text, nvpt):
- filename = modules.textual_inversion.textual_inversion.create_embedding(name, nvpt, init_text=initialization_text)
+def create_embedding(name, initialization_text, nvpt, overwrite_old):
+ filename = modules.textual_inversion.textual_inversion.create_embedding(name, nvpt, overwrite_old, init_text=initialization_text)
sd_hijack.model_hijack.embedding_db.load_textual_inversion_embeddings()
|