Spaces:
Running
Running
Upload 2 files
Browse filesAdded "Rule34" and "Xbooru". Removed "OR_tags" as it's not needed anymore and renamed "AND_tags" to "Tags".
- app.py +632 -636
- modules/booru.py +110 -131
app.py
CHANGED
@@ -1,637 +1,633 @@
|
|
1 |
-
import os,io,copy,json,requests,spaces,gradio as gr,numpy as np
|
2 |
-
import argparse,huggingface_hub,onnxruntime as rt,pandas as pd,traceback,tempfile,zipfile,re,ast,time
|
3 |
-
from datetime import datetime,timezone
|
4 |
-
from collections import defaultdict
|
5 |
-
from PIL import Image,ImageOps
|
6 |
-
from modules.booru import
|
7 |
-
from apscheduler.schedulers.background import BackgroundScheduler
|
8 |
-
from modules.classifyTags import classify_tags,process_tags
|
9 |
-
from modules.reorganizer_model import reorganizer_list,reorganizer_class
|
10 |
-
from modules.tag_enhancer import prompt_enhancer
|
11 |
-
from modules.florence2 import process_image,single_task_list,update_task_dropdown
|
12 |
-
|
13 |
-
os.environ['PYTORCH_ENABLE_MPS_FALLBACK']='1'
|
14 |
-
|
15 |
-
TITLE = "Multi-Tagger v1.2"
|
16 |
-
DESCRIPTION = """
|
17 |
-
Multi-Tagger is a versatile application for advanced image analysis and captioning. Perfect for AI artists or enthusiasts, it offers a range of features:
|
18 |
-
|
19 |
-
- **Automatic Tag Categorization**: Tags are grouped into categories.
|
20 |
-
- **Tag Enhancement**: Boost your prompts with enhanced descriptions using a built-in prompt enhancer.
|
21 |
-
- **Reorganizer**: Use a reorganizer model to format tags into a natural-language description.
|
22 |
-
- **Batch Support**: Upload and process multiple images simultaneously.
|
23 |
-
- **Downloadable Output**: Get almost all results as downloadable `.txt`, `.json`, and `.png` files in a `.zip` archive.
|
24 |
-
- **Image Fetcher**: Search for images from **Gelbooru** using flexible tag filters.
|
25 |
-
- **CUDA** and **CPU** support.
|
26 |
-
"""
|
27 |
-
|
28 |
-
# Dataset v3 series of models:
|
29 |
-
SWINV2_MODEL_DSV3_REPO = "SmilingWolf/wd-swinv2-tagger-v3"
|
30 |
-
CONV_MODEL_DSV3_REPO = "SmilingWolf/wd-convnext-tagger-v3"
|
31 |
-
VIT_MODEL_DSV3_REPO = "SmilingWolf/wd-vit-tagger-v3"
|
32 |
-
VIT_LARGE_MODEL_DSV3_REPO = "SmilingWolf/wd-vit-large-tagger-v3"
|
33 |
-
EVA02_LARGE_MODEL_DSV3_REPO = "SmilingWolf/wd-eva02-large-tagger-v3"
|
34 |
-
# Dataset v2 series of models:
|
35 |
-
MOAT_MODEL_DSV2_REPO = "SmilingWolf/wd-v1-4-moat-tagger-v2"
|
36 |
-
SWIN_MODEL_DSV2_REPO = "SmilingWolf/wd-v1-4-swinv2-tagger-v2"
|
37 |
-
CONV_MODEL_DSV2_REPO = "SmilingWolf/wd-v1-4-convnext-tagger-v2"
|
38 |
-
CONV2_MODEL_DSV2_REPO = "SmilingWolf/wd-v1-4-convnextv2-tagger-v2"
|
39 |
-
VIT_MODEL_DSV2_REPO = "SmilingWolf/wd-v1-4-vit-tagger-v2"
|
40 |
-
# IdolSankaku series of models:
|
41 |
-
EVA02_LARGE_MODEL_IS_DSV1_REPO = "deepghs/idolsankaku-eva02-large-tagger-v1"
|
42 |
-
SWINV2_MODEL_IS_DSV1_REPO = "deepghs/idolsankaku-swinv2-tagger-v1"
|
43 |
-
# Files to download from the repos
|
44 |
-
MODEL_FILENAME = "model.onnx"
|
45 |
-
LABEL_FILENAME = "selected_tags.csv"
|
46 |
-
|
47 |
-
kaomojis=['0_0','(o)_(o)','+_+','+_-','._.','<o>_<o>','<|>_<|>','=_=','>_<','3_3','6_9','>_o','@_@','^_^','o_o','u_u','x_x','|_|','||_||']
|
48 |
-
def parse_args()->argparse.Namespace:parser=argparse.ArgumentParser();parser.add_argument('--score-slider-step',type=float,default=.05);parser.add_argument('--score-general-threshold',type=float,default=.35);parser.add_argument('--score-character-threshold',type=float,default=.85);parser.add_argument('--share',action='store_true');return parser.parse_args()
|
49 |
-
def load_labels(dataframe)->list[str]:name_series=dataframe['name'];name_series=name_series.map(lambda x:x.replace('_',' ')if x not in kaomojis else x);tag_names=name_series.tolist();rating_indexes=list(np.where(dataframe['category']==9)[0]);general_indexes=list(np.where(dataframe['category']==0)[0]);character_indexes=list(np.where(dataframe['category']==4)[0]);return tag_names,rating_indexes,general_indexes,character_indexes
|
50 |
-
def mcut_threshold(probs):sorted_probs=probs[probs.argsort()[::-1]];difs=sorted_probs[:-1]-sorted_probs[1:];t=difs.argmax();thresh=(sorted_probs[t]+sorted_probs[t+1])/2;return thresh
|
51 |
-
|
52 |
-
class Timer:
|
53 |
-
def __init__(self):self.start_time=time.perf_counter();self.checkpoints=[('Start',self.start_time)]
|
54 |
-
def checkpoint(self,label='Checkpoint'):now=time.perf_counter();self.checkpoints.append((label,now))
|
55 |
-
def report(self,is_clear_checkpoints=True):
|
56 |
-
max_label_length=max(len(label)for(label,_)in self.checkpoints);prev_time=self.checkpoints[0][1]
|
57 |
-
for(label,curr_time)in self.checkpoints[1:]:elapsed=curr_time-prev_time;print(f"{label.ljust(max_label_length)}: {elapsed:.3f} seconds");prev_time=curr_time
|
58 |
-
if is_clear_checkpoints:self.checkpoints.clear();self.checkpoint()
|
59 |
-
def report_all(self):
|
60 |
-
print('\n> Execution Time Report:');max_label_length=max(len(label)for(label,_)in self.checkpoints)if len(self.checkpoints)>0 else 0;prev_time=self.start_time
|
61 |
-
for(label,curr_time)in self.checkpoints[1:]:elapsed=curr_time-prev_time;print(f"{label.ljust(max_label_length)}: {elapsed:.3f} seconds");prev_time=curr_time
|
62 |
-
total_time=self.checkpoints[-1][1]-self.start_time;print(f"{'Total Execution Time'.ljust(max_label_length)}: {total_time:.3f} seconds\n");self.checkpoints.clear()
|
63 |
-
def restart(self):self.start_time=time.perf_counter();self.checkpoints=[('Start',self.start_time)]
|
64 |
-
class Predictor:
|
65 |
-
def __init__(self):
|
66 |
-
self.model_target_size = None
|
67 |
-
self.last_loaded_repo = None
|
68 |
-
def download_model(self, model_repo):
|
69 |
-
csv_path = huggingface_hub.hf_hub_download(
|
70 |
-
model_repo,
|
71 |
-
LABEL_FILENAME,
|
72 |
-
)
|
73 |
-
model_path = huggingface_hub.hf_hub_download(
|
74 |
-
model_repo,
|
75 |
-
MODEL_FILENAME,
|
76 |
-
)
|
77 |
-
return csv_path, model_path
|
78 |
-
def load_model(self, model_repo):
|
79 |
-
if model_repo == self.last_loaded_repo:
|
80 |
-
return
|
81 |
-
|
82 |
-
csv_path, model_path = self.download_model(model_repo)
|
83 |
-
|
84 |
-
tags_df = pd.read_csv(csv_path)
|
85 |
-
sep_tags = load_labels(tags_df)
|
86 |
-
|
87 |
-
self.tag_names = sep_tags[0]
|
88 |
-
self.rating_indexes = sep_tags[1]
|
89 |
-
self.general_indexes = sep_tags[2]
|
90 |
-
self.character_indexes = sep_tags[3]
|
91 |
-
|
92 |
-
model = rt.InferenceSession(model_path)
|
93 |
-
_, height, width, _ = model.get_inputs()[0].shape
|
94 |
-
self.model_target_size = height
|
95 |
-
|
96 |
-
self.last_loaded_repo = model_repo
|
97 |
-
self.model = model
|
98 |
-
def prepare_image(self, path):
|
99 |
-
image = Image.open(path)
|
100 |
-
image = image.convert("RGBA")
|
101 |
-
target_size = self.model_target_size
|
102 |
-
|
103 |
-
canvas = Image.new("RGBA", image.size, (255, 255, 255))
|
104 |
-
canvas.alpha_composite(image)
|
105 |
-
image = canvas.convert("RGB")
|
106 |
-
|
107 |
-
# Pad image to square
|
108 |
-
image_shape = image.size
|
109 |
-
max_dim = max(image_shape)
|
110 |
-
pad_left = (max_dim - image_shape[0]) // 2
|
111 |
-
pad_top = (max_dim - image_shape[1]) // 2
|
112 |
-
|
113 |
-
padded_image = Image.new("RGB", (max_dim, max_dim), (255, 255, 255))
|
114 |
-
padded_image.paste(image, (pad_left, pad_top))
|
115 |
-
|
116 |
-
# Resize
|
117 |
-
if max_dim != target_size:
|
118 |
-
padded_image = padded_image.resize(
|
119 |
-
(target_size, target_size),
|
120 |
-
Image.BICUBIC,
|
121 |
-
)
|
122 |
-
# Convert to numpy array
|
123 |
-
image_array = np.asarray(padded_image, dtype=np.float32)
|
124 |
-
# Convert PIL-native RGB to BGR
|
125 |
-
image_array = image_array[:, :, ::-1]
|
126 |
-
return np.expand_dims(image_array, axis=0)
|
127 |
-
|
128 |
-
def create_file(self, content: str, directory: str, fileName: str) -> str:
|
129 |
-
# Write the content to a file
|
130 |
-
file_path = os.path.join(directory, fileName)
|
131 |
-
if fileName.endswith('.json'):
|
132 |
-
with open(file_path, 'w', encoding="utf-8") as file:
|
133 |
-
file.write(content)
|
134 |
-
else:
|
135 |
-
with open(file_path, 'w+', encoding="utf-8") as file:
|
136 |
-
file.write(content)
|
137 |
-
|
138 |
-
return file_path
|
139 |
-
|
140 |
-
def predict(
|
141 |
-
self,
|
142 |
-
gallery,
|
143 |
-
model_repo,
|
144 |
-
general_thresh,
|
145 |
-
general_mcut_enabled,
|
146 |
-
character_thresh,
|
147 |
-
character_mcut_enabled,
|
148 |
-
characters_merge_enabled,
|
149 |
-
reorganizer_model_repo,
|
150 |
-
additional_tags_prepend,
|
151 |
-
additional_tags_append,
|
152 |
-
tag_results,
|
153 |
-
progress=gr.Progress()
|
154 |
-
):
|
155 |
-
# Clear tag_results before starting a new prediction
|
156 |
-
tag_results.clear()
|
157 |
-
|
158 |
-
gallery_len = len(gallery)
|
159 |
-
print(f"Predict load model: {model_repo}, gallery length: {gallery_len}")
|
160 |
-
|
161 |
-
timer = Timer() # Create a timer
|
162 |
-
progressRatio = 0.5 if reorganizer_model_repo else 1
|
163 |
-
progressTotal = gallery_len + 1
|
164 |
-
current_progress = 0
|
165 |
-
|
166 |
-
self.load_model(model_repo)
|
167 |
-
current_progress += progressRatio/progressTotal;
|
168 |
-
progress(current_progress, desc="Initialize wd model finished")
|
169 |
-
timer.checkpoint(f"Initialize wd model")
|
170 |
-
|
171 |
-
txt_infos = []
|
172 |
-
output_dir = tempfile.mkdtemp()
|
173 |
-
if not os.path.exists(output_dir):
|
174 |
-
os.makedirs(output_dir)
|
175 |
-
|
176 |
-
sorted_general_strings = ""
|
177 |
-
# Create categorized output string
|
178 |
-
categorized_output_strings = []
|
179 |
-
rating = None
|
180 |
-
character_res = None
|
181 |
-
general_res = None
|
182 |
-
|
183 |
-
if reorganizer_model_repo:
|
184 |
-
print(f"Reorganizer load model {reorganizer_model_repo}")
|
185 |
-
reorganizer = reorganizer_class(reorganizer_model_repo, loadModel=True)
|
186 |
-
current_progress += progressRatio/progressTotal;
|
187 |
-
progress(current_progress, desc="Initialize reoganizer model finished")
|
188 |
-
timer.checkpoint(f"Initialize reoganizer model")
|
189 |
-
|
190 |
-
timer.report()
|
191 |
-
|
192 |
-
prepend_list = [tag.strip() for tag in additional_tags_prepend.split(",") if tag.strip()]
|
193 |
-
append_list = [tag.strip() for tag in additional_tags_append.split(",") if tag.strip()]
|
194 |
-
if prepend_list and append_list:
|
195 |
-
append_list = [item for item in append_list if item not in prepend_list]
|
196 |
-
|
197 |
-
# Dictionary to track counters for each filename
|
198 |
-
name_counters = defaultdict(int)
|
199 |
-
|
200 |
-
for idx, value in enumerate(gallery):
|
201 |
-
try:
|
202 |
-
image_path = value[0]
|
203 |
-
image_name = os.path.splitext(os.path.basename(image_path))[0]
|
204 |
-
|
205 |
-
# Increment the counter for the current name
|
206 |
-
name_counters[image_name] += 1
|
207 |
-
|
208 |
-
if name_counters[image_name] > 1:
|
209 |
-
image_name = f"{image_name}_{name_counters[image_name]:02d}"
|
210 |
-
|
211 |
-
image = self.prepare_image(image_path)
|
212 |
-
|
213 |
-
input_name = self.model.get_inputs()[0].name
|
214 |
-
label_name = self.model.get_outputs()[0].name
|
215 |
-
print(f"Gallery {idx:02d}: Starting run wd model...")
|
216 |
-
preds = self.model.run([label_name], {input_name: image})[0]
|
217 |
-
|
218 |
-
labels = list(zip(self.tag_names, preds[0].astype(float)))
|
219 |
-
|
220 |
-
# First 4 labels are actually ratings: pick one with argmax
|
221 |
-
ratings_names = [labels[i] for i in self.rating_indexes]
|
222 |
-
rating = dict(ratings_names)
|
223 |
-
|
224 |
-
# Then we have general tags: pick any where prediction confidence > threshold
|
225 |
-
general_names = [labels[i] for i in self.general_indexes]
|
226 |
-
|
227 |
-
if general_mcut_enabled:
|
228 |
-
general_probs = np.array([x[1] for x in general_names])
|
229 |
-
general_thresh = mcut_threshold(general_probs)
|
230 |
-
|
231 |
-
general_res = [x for x in general_names if x[1] > general_thresh]
|
232 |
-
general_res = dict(general_res)
|
233 |
-
|
234 |
-
# Everything else is characters: pick any where prediction confidence > threshold
|
235 |
-
character_names = [labels[i] for i in self.character_indexes]
|
236 |
-
|
237 |
-
if character_mcut_enabled:
|
238 |
-
character_probs = np.array([x[1] for x in character_names])
|
239 |
-
character_thresh = mcut_threshold(character_probs)
|
240 |
-
character_thresh = max(0.15, character_thresh)
|
241 |
-
|
242 |
-
character_res = [x for x in character_names if x[1] > character_thresh]
|
243 |
-
character_res = dict(character_res)
|
244 |
-
character_list = list(character_res.keys())
|
245 |
-
|
246 |
-
sorted_general_list = sorted(
|
247 |
-
general_res.items(),
|
248 |
-
key=lambda x: x[1],
|
249 |
-
reverse=True,
|
250 |
-
)
|
251 |
-
sorted_general_list = [x[0] for x in sorted_general_list]
|
252 |
-
# Remove values from character_list that already exist in sorted_general_list
|
253 |
-
character_list = [item for item in character_list if item not in sorted_general_list]
|
254 |
-
# Remove values from sorted_general_list that already exist in prepend_list or append_list
|
255 |
-
if prepend_list:
|
256 |
-
sorted_general_list = [item for item in sorted_general_list if item not in prepend_list]
|
257 |
-
if append_list:
|
258 |
-
sorted_general_list = [item for item in sorted_general_list if item not in append_list]
|
259 |
-
|
260 |
-
sorted_general_list = prepend_list + sorted_general_list + append_list
|
261 |
-
|
262 |
-
sorted_general_strings = ", ".join((character_list if characters_merge_enabled else []) + sorted_general_list).replace("(", "\(").replace(")", "\)")
|
263 |
-
|
264 |
-
classified_tags, unclassified_tags = classify_tags(sorted_general_list)
|
265 |
-
|
266 |
-
# Create a single string of ALL categorized tags for the current image
|
267 |
-
categorized_output_string = ', '.join([', '.join(tags) for tags in classified_tags.values()])
|
268 |
-
categorized_output_strings.append(categorized_output_string)
|
269 |
-
# Collect all categorized output strings into a single string
|
270 |
-
final_categorized_output = ', '.join(categorized_output_strings)
|
271 |
-
|
272 |
-
# Create a .txt file for "Output (string)" and "Categorized Output (string)"
|
273 |
-
txt_content = f"Output (string): {sorted_general_strings}\nCategorized Output (string): {final_categorized_output}"
|
274 |
-
txt_file = self.create_file(txt_content, output_dir, f"{image_name}_output.txt")
|
275 |
-
txt_infos.append({"path": txt_file, "name": f"{image_name}_output.txt"})
|
276 |
-
|
277 |
-
# Create a .json file for "Categorized (tags)"
|
278 |
-
json_content = json.dumps(classified_tags, indent=4)
|
279 |
-
json_file = self.create_file(json_content, output_dir, f"{image_name}_categorized_tags.json")
|
280 |
-
txt_infos.append({"path": json_file, "name": f"{image_name}_categorized_tags.json"})
|
281 |
-
|
282 |
-
# Save a copy of the uploaded image in PNG format
|
283 |
-
image_path = value[0]
|
284 |
-
image = Image.open(image_path)
|
285 |
-
image.save(os.path.join(output_dir, f"{image_name}.png"), format="PNG")
|
286 |
-
txt_infos.append({"path": os.path.join(output_dir, f"{image_name}.png"), "name": f"{image_name}.png"})
|
287 |
-
|
288 |
-
current_progress += progressRatio/progressTotal;
|
289 |
-
progress(current_progress, desc=f"image{idx:02d}, predict finished")
|
290 |
-
timer.checkpoint(f"image{idx:02d}, predict finished")
|
291 |
-
|
292 |
-
if reorganizer_model_repo:
|
293 |
-
print(f"Starting reorganizer...")
|
294 |
-
reorganize_strings = reorganizer.reorganize(sorted_general_strings)
|
295 |
-
reorganize_strings = re.sub(r" *Title: *", "", reorganize_strings)
|
296 |
-
reorganize_strings = re.sub(r"\n+", ",", reorganize_strings)
|
297 |
-
reorganize_strings = re.sub(r",,+", ",", reorganize_strings)
|
298 |
-
sorted_general_strings += ",\n\n" + reorganize_strings
|
299 |
-
|
300 |
-
current_progress += progressRatio/progressTotal;
|
301 |
-
progress(current_progress, desc=f"image{idx:02d}, reorganizer finished")
|
302 |
-
timer.checkpoint(f"image{idx:02d}, reorganizer finished")
|
303 |
-
|
304 |
-
txt_file = self.create_file(sorted_general_strings, output_dir, image_name + ".txt")
|
305 |
-
txt_infos.append({"path":txt_file, "name": image_name + ".txt"})
|
306 |
-
|
307 |
-
# Store the result in tag_results using image_path as the key
|
308 |
-
tag_results[image_path] = {
|
309 |
-
"strings": sorted_general_strings,
|
310 |
-
"strings2": categorized_output_string, # Store the categorized output string here
|
311 |
-
"classified_tags": classified_tags,
|
312 |
-
"rating": rating,
|
313 |
-
"character_res": character_res,
|
314 |
-
"general_res": general_res,
|
315 |
-
"unclassified_tags": unclassified_tags,
|
316 |
-
"enhanced_tags": "" # Initialize as empty string
|
317 |
-
}
|
318 |
-
|
319 |
-
timer.report()
|
320 |
-
except Exception as e:
|
321 |
-
print(traceback.format_exc())
|
322 |
-
print("Error predict: " + str(e))
|
323 |
-
# Zip creation logic:
|
324 |
-
download = []
|
325 |
-
if txt_infos is not None and len(txt_infos) > 0:
|
326 |
-
downloadZipPath = os.path.join(output_dir, "Multi-tagger-" + datetime.now().strftime("%Y%m%d-%H%M%S") + ".zip")
|
327 |
-
with zipfile.ZipFile(downloadZipPath, 'w', zipfile.ZIP_DEFLATED) as taggers_zip:
|
328 |
-
for info in txt_infos:
|
329 |
-
# Get file name from lookup
|
330 |
-
taggers_zip.write(info["path"], arcname=info["name"])
|
331 |
-
download.append(downloadZipPath)
|
332 |
-
# End zip creation logic
|
333 |
-
if reorganizer_model_repo:
|
334 |
-
reorganizer.release_vram()
|
335 |
-
del reorganizer
|
336 |
-
|
337 |
-
progress(1, desc=f"Predict completed")
|
338 |
-
timer.report_all() # Print all recorded times
|
339 |
-
print("Predict is complete.")
|
340 |
-
|
341 |
-
return download, sorted_general_strings, final_categorized_output, classified_tags, rating, character_res, general_res, unclassified_tags, tag_results
|
342 |
-
def get_selection_from_gallery(gallery: list, tag_results: dict, selected_state: gr.SelectData):
|
343 |
-
if not selected_state:
|
344 |
-
return selected_state
|
345 |
-
tag_result = {
|
346 |
-
"strings": "",
|
347 |
-
"strings2": "",
|
348 |
-
"classified_tags": "{}",
|
349 |
-
"rating": "",
|
350 |
-
"character_res": "",
|
351 |
-
"general_res": "",
|
352 |
-
"unclassified_tags": "{}",
|
353 |
-
"enhanced_tags": ""
|
354 |
-
}
|
355 |
-
if selected_state.value["image"]["path"] in tag_results:
|
356 |
-
tag_result = tag_results[selected_state.value["image"]["path"]]
|
357 |
-
return (selected_state.value["image"]["path"], selected_state.value["caption"]), tag_result["strings"], tag_result["strings2"], tag_result["classified_tags"], tag_result["rating"], tag_result["character_res"], tag_result["general_res"], tag_result["unclassified_tags"], tag_result["enhanced_tags"]
|
358 |
-
def append_gallery(gallery:list,image:str):
|
359 |
-
if gallery is None:gallery=[]
|
360 |
-
if not image:return gallery,None
|
361 |
-
gallery.append(image);return gallery,None
|
362 |
-
def extend_gallery(gallery:list,images):
|
363 |
-
if gallery is None:gallery=[]
|
364 |
-
if not images:return gallery
|
365 |
-
gallery.extend(images);return gallery
|
366 |
-
def remove_image_from_gallery(gallery:list,selected_image:str):
|
367 |
-
if not gallery or not selected_image:return gallery
|
368 |
-
selected_image=ast.literal_eval(selected_image)
|
369 |
-
if selected_image in gallery:gallery.remove(selected_image)
|
370 |
-
return gallery
|
371 |
-
args = parse_args()
|
372 |
-
predictor = Predictor()
|
373 |
-
dropdown_list = [
|
374 |
-
EVA02_LARGE_MODEL_DSV3_REPO,
|
375 |
-
SWINV2_MODEL_DSV3_REPO,
|
376 |
-
CONV_MODEL_DSV3_REPO,
|
377 |
-
VIT_MODEL_DSV3_REPO,
|
378 |
-
VIT_LARGE_MODEL_DSV3_REPO,
|
379 |
-
# ---
|
380 |
-
MOAT_MODEL_DSV2_REPO,
|
381 |
-
SWIN_MODEL_DSV2_REPO,
|
382 |
-
CONV_MODEL_DSV2_REPO,
|
383 |
-
CONV2_MODEL_DSV2_REPO,
|
384 |
-
VIT_MODEL_DSV2_REPO,
|
385 |
-
# ---
|
386 |
-
SWINV2_MODEL_IS_DSV1_REPO,
|
387 |
-
EVA02_LARGE_MODEL_IS_DSV1_REPO,
|
388 |
-
]
|
389 |
-
|
390 |
-
def _restart_space():
|
391 |
-
HF_TOKEN=os.getenv('HF_TOKEN')
|
392 |
-
if not HF_TOKEN:raise ValueError('HF_TOKEN environment variable is not set.')
|
393 |
-
huggingface_hub.HfApi().restart_space(repo_id='Werli/Multi-Tagger',token=HF_TOKEN,factory_reboot=False)
|
394 |
-
scheduler=BackgroundScheduler()
|
395 |
-
# Add a job to restart the space every 2 days (172800 seconds)
|
396 |
-
restart_space_job = scheduler.add_job(_restart_space, "interval", seconds=172800)
|
397 |
-
scheduler.start()
|
398 |
-
next_run_time_utc=restart_space_job.next_run_time.astimezone(timezone.utc)
|
399 |
-
NEXT_RESTART=f"Next Restart: {next_run_time_utc.strftime('%Y-%m-%d %H:%M:%S')} (UTC) - The space will restart every 2 days to ensure stability and performance. It uses a background scheduler to handle the restart process."
|
400 |
-
|
401 |
-
css = """
|
402 |
-
#output {height: 500px; overflow: auto; border: 1px solid #ccc;}
|
403 |
-
label.float.svelte-i3tvor {position: relative !important;}
|
404 |
-
.reduced-height.svelte-11chud3 {height: calc(80% - var(--size-10));}
|
405 |
-
"""
|
406 |
-
|
407 |
-
with gr.Blocks(title=TITLE, css=css, theme=gr.themes.Soft(), fill_width=True) as demo:
|
408 |
-
gr.Markdown(value=f"<h1 style='text-align: center; margin-bottom: 1rem'>{TITLE}</h1>")
|
409 |
-
gr.Markdown(value=DESCRIPTION)
|
410 |
-
gr.Markdown(NEXT_RESTART)
|
411 |
-
with gr.Tab(label="Waifu Diffusion"):
|
412 |
-
with gr.Row():
|
413 |
-
with gr.Column():
|
414 |
-
submit = gr.Button(value="Submit", variant="primary", size="lg")
|
415 |
-
with gr.Column(variant="panel"):
|
416 |
-
# Create an Image component for uploading images
|
417 |
-
image_input = gr.Image(label="Upload an Image or clicking paste from clipboard button", type="filepath", sources=["upload", "clipboard"], height=150)
|
418 |
-
with gr.Row():
|
419 |
-
upload_button = gr.UploadButton("Upload multiple images", file_types=["image"], file_count="multiple", size="sm")
|
420 |
-
remove_button = gr.Button("Remove Selected Image", size="sm")
|
421 |
-
gallery = gr.Gallery(columns=5, rows=5, show_share_button=False, interactive=True, height="500px", label="Grid of images")
|
422 |
-
model_repo = gr.Dropdown(
|
423 |
-
dropdown_list,
|
424 |
-
value=EVA02_LARGE_MODEL_DSV3_REPO,
|
425 |
-
label="Model",
|
426 |
-
)
|
427 |
-
with gr.Row():
|
428 |
-
general_thresh = gr.Slider(
|
429 |
-
0,
|
430 |
-
1,
|
431 |
-
step=args.score_slider_step,
|
432 |
-
value=args.score_general_threshold,
|
433 |
-
label="General Tags Threshold",
|
434 |
-
scale=3,
|
435 |
-
)
|
436 |
-
general_mcut_enabled = gr.Checkbox(
|
437 |
-
value=False,
|
438 |
-
label="Use MCut threshold",
|
439 |
-
scale=1,
|
440 |
-
)
|
441 |
-
with gr.Row():
|
442 |
-
character_thresh = gr.Slider(
|
443 |
-
0,
|
444 |
-
1,
|
445 |
-
step=args.score_slider_step,
|
446 |
-
value=args.score_character_threshold,
|
447 |
-
label="Character Tags Threshold",
|
448 |
-
scale=3,
|
449 |
-
)
|
450 |
-
character_mcut_enabled = gr.Checkbox(
|
451 |
-
value=False,
|
452 |
-
label="Use MCut threshold",
|
453 |
-
scale=1,
|
454 |
-
)
|
455 |
-
with gr.Row():
|
456 |
-
characters_merge_enabled = gr.Checkbox(
|
457 |
-
value=True,
|
458 |
-
label="Merge characters into the string output",
|
459 |
-
scale=1,
|
460 |
-
)
|
461 |
-
with gr.Row():
|
462 |
-
reorganizer_model_repo = gr.Dropdown(
|
463 |
-
[None] + reorganizer_list,
|
464 |
-
value=None,
|
465 |
-
label="Reorganizer Model",
|
466 |
-
info="Use a model to create a description for you",
|
467 |
-
)
|
468 |
-
with gr.Row():
|
469 |
-
additional_tags_prepend = gr.Text(label="Prepend Additional tags (comma split)")
|
470 |
-
additional_tags_append = gr.Text(label="Append Additional tags (comma split)")
|
471 |
-
with gr.Row():
|
472 |
-
clear = gr.ClearButton(
|
473 |
-
components=[
|
474 |
-
gallery,
|
475 |
-
model_repo,
|
476 |
-
general_thresh,
|
477 |
-
general_mcut_enabled,
|
478 |
-
character_thresh,
|
479 |
-
character_mcut_enabled,
|
480 |
-
characters_merge_enabled,
|
481 |
-
reorganizer_model_repo,
|
482 |
-
additional_tags_prepend,
|
483 |
-
additional_tags_append,
|
484 |
-
],
|
485 |
-
variant="secondary",
|
486 |
-
size="lg",
|
487 |
-
)
|
488 |
-
with gr.Column(variant="panel"):
|
489 |
-
download_file = gr.File(label="Download includes: All outputs* and image(s)") # 0
|
490 |
-
character_res = gr.Label(label="Output (characters)") # 1
|
491 |
-
sorted_general_strings = gr.Textbox(label="Output (string)*", show_label=True, show_copy_button=True) # 2
|
492 |
-
final_categorized_output = gr.Textbox(label="Categorized (string)* - If it's too long, select an image to display tags correctly.", show_label=True, show_copy_button=True) # 3
|
493 |
-
pe_generate_btn = gr.Button(value="ENHANCE TAGS", size="lg", variant="primary") # 4
|
494 |
-
enhanced_tags = gr.Textbox(label="Enhanced Tags", show_label=True, show_copy_button=True) # 5
|
495 |
-
prompt_enhancer_model = gr.Radio(["Medium", "Long", "Flux"], label="Model Choice", value="Medium", info="Enhance your prompts with Medium or Long answers") # 6
|
496 |
-
categorized = gr.JSON(label="Categorized (tags)* - JSON") # 7
|
497 |
-
rating = gr.Label(label="Rating") # 8
|
498 |
-
general_res = gr.Label(label="Output (tags)") # 9
|
499 |
-
unclassified = gr.JSON(label="Unclassified (tags)") # 10
|
500 |
-
clear.add(
|
501 |
-
[
|
502 |
-
download_file,
|
503 |
-
sorted_general_strings,
|
504 |
-
final_categorized_output,
|
505 |
-
categorized,
|
506 |
-
rating,
|
507 |
-
character_res,
|
508 |
-
general_res,
|
509 |
-
unclassified,
|
510 |
-
prompt_enhancer_model,
|
511 |
-
enhanced_tags,
|
512 |
-
]
|
513 |
-
)
|
514 |
-
tag_results = gr.State({})
|
515 |
-
# Define the event listener to add the uploaded image to the gallery
|
516 |
-
image_input.change(append_gallery, inputs=[gallery, image_input], outputs=[gallery, image_input])
|
517 |
-
# When the upload button is clicked, add the new images to the gallery
|
518 |
-
upload_button.upload(extend_gallery, inputs=[gallery, upload_button], outputs=gallery)
|
519 |
-
# Event to update the selected image when an image is clicked in the gallery
|
520 |
-
selected_image = gr.Textbox(label="Selected Image", visible=False)
|
521 |
-
gallery.select(get_selection_from_gallery,inputs=[gallery, tag_results],outputs=[selected_image, sorted_general_strings, final_categorized_output, categorized, rating, character_res, general_res, unclassified, enhanced_tags])
|
522 |
-
# Event to remove a selected image from the gallery
|
523 |
-
remove_button.click(remove_image_from_gallery, inputs=[gallery, selected_image], outputs=gallery)
|
524 |
-
# Event to for the Prompt Enhancer Button
|
525 |
-
pe_generate_btn.click(lambda tags,model:prompt_enhancer('','',tags,model)[0],inputs=[final_categorized_output,prompt_enhancer_model],outputs=[enhanced_tags])
|
526 |
-
submit.click(
|
527 |
-
predictor.predict,
|
528 |
-
inputs=[
|
529 |
-
gallery,
|
530 |
-
model_repo,
|
531 |
-
general_thresh,
|
532 |
-
general_mcut_enabled,
|
533 |
-
character_thresh,
|
534 |
-
character_mcut_enabled,
|
535 |
-
characters_merge_enabled,
|
536 |
-
reorganizer_model_repo,
|
537 |
-
additional_tags_prepend,
|
538 |
-
additional_tags_append,
|
539 |
-
tag_results,
|
540 |
-
],
|
541 |
-
outputs=[download_file, sorted_general_strings, final_categorized_output, categorized, rating, character_res, general_res, unclassified, tag_results,],
|
542 |
-
)
|
543 |
-
gr.Examples(
|
544 |
-
[["images/1girl.png", VIT_LARGE_MODEL_DSV3_REPO, 0.35, False, 0.85, False]],
|
545 |
-
inputs=[
|
546 |
-
image_input,
|
547 |
-
model_repo,
|
548 |
-
general_thresh,
|
549 |
-
general_mcut_enabled,
|
550 |
-
character_thresh,
|
551 |
-
character_mcut_enabled,
|
552 |
-
],
|
553 |
-
)
|
554 |
-
with gr.Tab(label="Florence 2 Image Captioning"):
|
555 |
-
with gr.Row():
|
556 |
-
with gr.Column(variant="panel"):
|
557 |
-
input_img = gr.Image(label="Input Picture")
|
558 |
-
task_type = gr.Radio(choices=['Single task', 'Cascaded task'], label='Task type selector', value='Single task')
|
559 |
-
task_prompt = gr.Dropdown(choices=single_task_list, label="Task Prompt", value="Caption")
|
560 |
-
task_type.change(fn=update_task_dropdown, inputs=task_type, outputs=task_prompt)
|
561 |
-
text_input = gr.Textbox(label="Text Input (optional)")
|
562 |
-
submit_btn = gr.Button(value="Submit")
|
563 |
-
with gr.Column(variant="panel"):
|
564 |
-
output_text = gr.Textbox(label="Output Text", show_label=True, show_copy_button=True, lines=8)
|
565 |
-
output_img = gr.Image(label="Output Image")
|
566 |
-
gr.Examples(
|
567 |
-
examples=[
|
568 |
-
["images/image1.png", 'Object Detection'],
|
569 |
-
["images/image2.png", 'OCR with Region']
|
570 |
-
],
|
571 |
-
inputs=[input_img, task_prompt],
|
572 |
-
outputs=[output_text, output_img],
|
573 |
-
fn=process_image,
|
574 |
-
cache_examples=False,
|
575 |
-
label='Try examples'
|
576 |
-
)
|
577 |
-
submit_btn.click(process_image, [input_img, task_prompt, text_input], [output_text, output_img])
|
578 |
-
with gr.Tab(
|
579 |
-
with gr.Row():
|
580 |
-
with gr.Column():
|
581 |
-
gr.Markdown("### ⚙️ Search Parameters")
|
582 |
-
site = gr.Dropdown(label="Select Source", choices=["Gelbooru", "
|
583 |
-
|
584 |
-
|
585 |
-
|
586 |
-
|
587 |
-
|
588 |
-
|
589 |
-
|
590 |
-
|
591 |
-
|
592 |
-
|
593 |
-
|
594 |
-
|
595 |
-
|
596 |
-
|
597 |
-
gr.
|
598 |
-
|
599 |
-
|
600 |
-
|
601 |
-
|
602 |
-
|
603 |
-
|
604 |
-
|
605 |
-
|
606 |
-
|
607 |
-
|
608 |
-
|
609 |
-
|
610 |
-
|
611 |
-
|
612 |
-
|
613 |
-
|
614 |
-
|
615 |
-
|
616 |
-
|
617 |
-
|
618 |
-
)
|
619 |
-
|
620 |
-
|
621 |
-
|
622 |
-
|
623 |
-
|
624 |
-
|
625 |
-
|
626 |
-
|
627 |
-
submit_button
|
628 |
-
with gr.Column(variant="panel"):
|
629 |
-
|
630 |
-
|
631 |
-
|
632 |
-
|
633 |
-
pe_generate_btn = gr.Button(value="ENHANCE TAGS", size="lg", variant="primary")
|
634 |
-
enhanced_tags = gr.Textbox(label="Enhanced Tags", show_label=True, show_copy_button=True)
|
635 |
-
prompt_enhancer_model = gr.Radio(["Medium", "Long", "Flux"], label="Model Choice", value="Medium", info="Enhance your prompts with Medium or Long answers")
|
636 |
-
pe_generate_btn.click(lambda tags,model:prompt_enhancer('','',tags,model)[0],inputs=[categorized_string,prompt_enhancer_model],outputs=[enhanced_tags])
|
637 |
demo.queue(max_size=2).launch()
|
|
|
1 |
+
import os,io,copy,json,requests,spaces,gradio as gr,numpy as np
|
2 |
+
import argparse,huggingface_hub,onnxruntime as rt,pandas as pd,traceback,tempfile,zipfile,re,ast,time
|
3 |
+
from datetime import datetime,timezone
|
4 |
+
from collections import defaultdict
|
5 |
+
from PIL import Image,ImageOps
|
6 |
+
from modules.booru import booru_gradio,on_select
|
7 |
+
from apscheduler.schedulers.background import BackgroundScheduler
|
8 |
+
from modules.classifyTags import classify_tags,process_tags
|
9 |
+
from modules.reorganizer_model import reorganizer_list,reorganizer_class
|
10 |
+
from modules.tag_enhancer import prompt_enhancer
|
11 |
+
from modules.florence2 import process_image,single_task_list,update_task_dropdown
|
12 |
+
|
13 |
+
os.environ['PYTORCH_ENABLE_MPS_FALLBACK']='1'
|
14 |
+
|
15 |
+
TITLE = "Multi-Tagger v1.2"
|
16 |
+
DESCRIPTION = """
|
17 |
+
Multi-Tagger is a versatile application for advanced image analysis and captioning. Perfect for AI artists or enthusiasts, it offers a range of features:
|
18 |
+
|
19 |
+
- **Automatic Tag Categorization**: Tags are grouped into categories.
|
20 |
+
- **Tag Enhancement**: Boost your prompts with enhanced descriptions using a built-in prompt enhancer.
|
21 |
+
- **Reorganizer**: Use a reorganizer model to format tags into a natural-language description.
|
22 |
+
- **Batch Support**: Upload and process multiple images simultaneously.
|
23 |
+
- **Downloadable Output**: Get almost all results as downloadable `.txt`, `.json`, and `.png` files in a `.zip` archive.
|
24 |
+
- **Image Fetcher**: Search for images from **Gelbooru** using flexible tag filters.
|
25 |
+
- **CUDA** and **CPU** support.
|
26 |
+
"""
|
27 |
+
|
28 |
+
# Dataset v3 series of models:
|
29 |
+
SWINV2_MODEL_DSV3_REPO = "SmilingWolf/wd-swinv2-tagger-v3"
|
30 |
+
CONV_MODEL_DSV3_REPO = "SmilingWolf/wd-convnext-tagger-v3"
|
31 |
+
VIT_MODEL_DSV3_REPO = "SmilingWolf/wd-vit-tagger-v3"
|
32 |
+
VIT_LARGE_MODEL_DSV3_REPO = "SmilingWolf/wd-vit-large-tagger-v3"
|
33 |
+
EVA02_LARGE_MODEL_DSV3_REPO = "SmilingWolf/wd-eva02-large-tagger-v3"
|
34 |
+
# Dataset v2 series of models:
|
35 |
+
MOAT_MODEL_DSV2_REPO = "SmilingWolf/wd-v1-4-moat-tagger-v2"
|
36 |
+
SWIN_MODEL_DSV2_REPO = "SmilingWolf/wd-v1-4-swinv2-tagger-v2"
|
37 |
+
CONV_MODEL_DSV2_REPO = "SmilingWolf/wd-v1-4-convnext-tagger-v2"
|
38 |
+
CONV2_MODEL_DSV2_REPO = "SmilingWolf/wd-v1-4-convnextv2-tagger-v2"
|
39 |
+
VIT_MODEL_DSV2_REPO = "SmilingWolf/wd-v1-4-vit-tagger-v2"
|
40 |
+
# IdolSankaku series of models:
|
41 |
+
EVA02_LARGE_MODEL_IS_DSV1_REPO = "deepghs/idolsankaku-eva02-large-tagger-v1"
|
42 |
+
SWINV2_MODEL_IS_DSV1_REPO = "deepghs/idolsankaku-swinv2-tagger-v1"
|
43 |
+
# Files to download from the repos
|
44 |
+
MODEL_FILENAME = "model.onnx"
|
45 |
+
LABEL_FILENAME = "selected_tags.csv"
|
46 |
+
|
47 |
+
kaomojis=['0_0','(o)_(o)','+_+','+_-','._.','<o>_<o>','<|>_<|>','=_=','>_<','3_3','6_9','>_o','@_@','^_^','o_o','u_u','x_x','|_|','||_||']
|
48 |
+
def parse_args()->argparse.Namespace:parser=argparse.ArgumentParser();parser.add_argument('--score-slider-step',type=float,default=.05);parser.add_argument('--score-general-threshold',type=float,default=.35);parser.add_argument('--score-character-threshold',type=float,default=.85);parser.add_argument('--share',action='store_true');return parser.parse_args()
|
49 |
+
def load_labels(dataframe)->list[str]:name_series=dataframe['name'];name_series=name_series.map(lambda x:x.replace('_',' ')if x not in kaomojis else x);tag_names=name_series.tolist();rating_indexes=list(np.where(dataframe['category']==9)[0]);general_indexes=list(np.where(dataframe['category']==0)[0]);character_indexes=list(np.where(dataframe['category']==4)[0]);return tag_names,rating_indexes,general_indexes,character_indexes
|
50 |
+
def mcut_threshold(probs):sorted_probs=probs[probs.argsort()[::-1]];difs=sorted_probs[:-1]-sorted_probs[1:];t=difs.argmax();thresh=(sorted_probs[t]+sorted_probs[t+1])/2;return thresh
|
51 |
+
|
52 |
+
class Timer:
|
53 |
+
def __init__(self):self.start_time=time.perf_counter();self.checkpoints=[('Start',self.start_time)]
|
54 |
+
def checkpoint(self,label='Checkpoint'):now=time.perf_counter();self.checkpoints.append((label,now))
|
55 |
+
def report(self,is_clear_checkpoints=True):
|
56 |
+
max_label_length=max(len(label)for(label,_)in self.checkpoints);prev_time=self.checkpoints[0][1]
|
57 |
+
for(label,curr_time)in self.checkpoints[1:]:elapsed=curr_time-prev_time;print(f"{label.ljust(max_label_length)}: {elapsed:.3f} seconds");prev_time=curr_time
|
58 |
+
if is_clear_checkpoints:self.checkpoints.clear();self.checkpoint()
|
59 |
+
def report_all(self):
|
60 |
+
print('\n> Execution Time Report:');max_label_length=max(len(label)for(label,_)in self.checkpoints)if len(self.checkpoints)>0 else 0;prev_time=self.start_time
|
61 |
+
for(label,curr_time)in self.checkpoints[1:]:elapsed=curr_time-prev_time;print(f"{label.ljust(max_label_length)}: {elapsed:.3f} seconds");prev_time=curr_time
|
62 |
+
total_time=self.checkpoints[-1][1]-self.start_time;print(f"{'Total Execution Time'.ljust(max_label_length)}: {total_time:.3f} seconds\n");self.checkpoints.clear()
|
63 |
+
def restart(self):self.start_time=time.perf_counter();self.checkpoints=[('Start',self.start_time)]
|
64 |
+
class Predictor:
|
65 |
+
def __init__(self):
|
66 |
+
self.model_target_size = None
|
67 |
+
self.last_loaded_repo = None
|
68 |
+
def download_model(self, model_repo):
|
69 |
+
csv_path = huggingface_hub.hf_hub_download(
|
70 |
+
model_repo,
|
71 |
+
LABEL_FILENAME,
|
72 |
+
)
|
73 |
+
model_path = huggingface_hub.hf_hub_download(
|
74 |
+
model_repo,
|
75 |
+
MODEL_FILENAME,
|
76 |
+
)
|
77 |
+
return csv_path, model_path
|
78 |
+
def load_model(self, model_repo):
|
79 |
+
if model_repo == self.last_loaded_repo:
|
80 |
+
return
|
81 |
+
|
82 |
+
csv_path, model_path = self.download_model(model_repo)
|
83 |
+
|
84 |
+
tags_df = pd.read_csv(csv_path)
|
85 |
+
sep_tags = load_labels(tags_df)
|
86 |
+
|
87 |
+
self.tag_names = sep_tags[0]
|
88 |
+
self.rating_indexes = sep_tags[1]
|
89 |
+
self.general_indexes = sep_tags[2]
|
90 |
+
self.character_indexes = sep_tags[3]
|
91 |
+
|
92 |
+
model = rt.InferenceSession(model_path)
|
93 |
+
_, height, width, _ = model.get_inputs()[0].shape
|
94 |
+
self.model_target_size = height
|
95 |
+
|
96 |
+
self.last_loaded_repo = model_repo
|
97 |
+
self.model = model
|
98 |
+
def prepare_image(self, path):
|
99 |
+
image = Image.open(path)
|
100 |
+
image = image.convert("RGBA")
|
101 |
+
target_size = self.model_target_size
|
102 |
+
|
103 |
+
canvas = Image.new("RGBA", image.size, (255, 255, 255))
|
104 |
+
canvas.alpha_composite(image)
|
105 |
+
image = canvas.convert("RGB")
|
106 |
+
|
107 |
+
# Pad image to square
|
108 |
+
image_shape = image.size
|
109 |
+
max_dim = max(image_shape)
|
110 |
+
pad_left = (max_dim - image_shape[0]) // 2
|
111 |
+
pad_top = (max_dim - image_shape[1]) // 2
|
112 |
+
|
113 |
+
padded_image = Image.new("RGB", (max_dim, max_dim), (255, 255, 255))
|
114 |
+
padded_image.paste(image, (pad_left, pad_top))
|
115 |
+
|
116 |
+
# Resize
|
117 |
+
if max_dim != target_size:
|
118 |
+
padded_image = padded_image.resize(
|
119 |
+
(target_size, target_size),
|
120 |
+
Image.BICUBIC,
|
121 |
+
)
|
122 |
+
# Convert to numpy array
|
123 |
+
image_array = np.asarray(padded_image, dtype=np.float32)
|
124 |
+
# Convert PIL-native RGB to BGR
|
125 |
+
image_array = image_array[:, :, ::-1]
|
126 |
+
return np.expand_dims(image_array, axis=0)
|
127 |
+
|
128 |
+
def create_file(self, content: str, directory: str, fileName: str) -> str:
|
129 |
+
# Write the content to a file
|
130 |
+
file_path = os.path.join(directory, fileName)
|
131 |
+
if fileName.endswith('.json'):
|
132 |
+
with open(file_path, 'w', encoding="utf-8") as file:
|
133 |
+
file.write(content)
|
134 |
+
else:
|
135 |
+
with open(file_path, 'w+', encoding="utf-8") as file:
|
136 |
+
file.write(content)
|
137 |
+
|
138 |
+
return file_path
|
139 |
+
|
140 |
+
def predict(
|
141 |
+
self,
|
142 |
+
gallery,
|
143 |
+
model_repo,
|
144 |
+
general_thresh,
|
145 |
+
general_mcut_enabled,
|
146 |
+
character_thresh,
|
147 |
+
character_mcut_enabled,
|
148 |
+
characters_merge_enabled,
|
149 |
+
reorganizer_model_repo,
|
150 |
+
additional_tags_prepend,
|
151 |
+
additional_tags_append,
|
152 |
+
tag_results,
|
153 |
+
progress=gr.Progress()
|
154 |
+
):
|
155 |
+
# Clear tag_results before starting a new prediction
|
156 |
+
tag_results.clear()
|
157 |
+
|
158 |
+
gallery_len = len(gallery)
|
159 |
+
print(f"Predict load model: {model_repo}, gallery length: {gallery_len}")
|
160 |
+
|
161 |
+
timer = Timer() # Create a timer
|
162 |
+
progressRatio = 0.5 if reorganizer_model_repo else 1
|
163 |
+
progressTotal = gallery_len + 1
|
164 |
+
current_progress = 0
|
165 |
+
|
166 |
+
self.load_model(model_repo)
|
167 |
+
current_progress += progressRatio/progressTotal;
|
168 |
+
progress(current_progress, desc="Initialize wd model finished")
|
169 |
+
timer.checkpoint(f"Initialize wd model")
|
170 |
+
|
171 |
+
txt_infos = []
|
172 |
+
output_dir = tempfile.mkdtemp()
|
173 |
+
if not os.path.exists(output_dir):
|
174 |
+
os.makedirs(output_dir)
|
175 |
+
|
176 |
+
sorted_general_strings = ""
|
177 |
+
# Create categorized output string
|
178 |
+
categorized_output_strings = []
|
179 |
+
rating = None
|
180 |
+
character_res = None
|
181 |
+
general_res = None
|
182 |
+
|
183 |
+
if reorganizer_model_repo:
|
184 |
+
print(f"Reorganizer load model {reorganizer_model_repo}")
|
185 |
+
reorganizer = reorganizer_class(reorganizer_model_repo, loadModel=True)
|
186 |
+
current_progress += progressRatio/progressTotal;
|
187 |
+
progress(current_progress, desc="Initialize reoganizer model finished")
|
188 |
+
timer.checkpoint(f"Initialize reoganizer model")
|
189 |
+
|
190 |
+
timer.report()
|
191 |
+
|
192 |
+
prepend_list = [tag.strip() for tag in additional_tags_prepend.split(",") if tag.strip()]
|
193 |
+
append_list = [tag.strip() for tag in additional_tags_append.split(",") if tag.strip()]
|
194 |
+
if prepend_list and append_list:
|
195 |
+
append_list = [item for item in append_list if item not in prepend_list]
|
196 |
+
|
197 |
+
# Dictionary to track counters for each filename
|
198 |
+
name_counters = defaultdict(int)
|
199 |
+
|
200 |
+
for idx, value in enumerate(gallery):
|
201 |
+
try:
|
202 |
+
image_path = value[0]
|
203 |
+
image_name = os.path.splitext(os.path.basename(image_path))[0]
|
204 |
+
|
205 |
+
# Increment the counter for the current name
|
206 |
+
name_counters[image_name] += 1
|
207 |
+
|
208 |
+
if name_counters[image_name] > 1:
|
209 |
+
image_name = f"{image_name}_{name_counters[image_name]:02d}"
|
210 |
+
|
211 |
+
image = self.prepare_image(image_path)
|
212 |
+
|
213 |
+
input_name = self.model.get_inputs()[0].name
|
214 |
+
label_name = self.model.get_outputs()[0].name
|
215 |
+
print(f"Gallery {idx:02d}: Starting run wd model...")
|
216 |
+
preds = self.model.run([label_name], {input_name: image})[0]
|
217 |
+
|
218 |
+
labels = list(zip(self.tag_names, preds[0].astype(float)))
|
219 |
+
|
220 |
+
# First 4 labels are actually ratings: pick one with argmax
|
221 |
+
ratings_names = [labels[i] for i in self.rating_indexes]
|
222 |
+
rating = dict(ratings_names)
|
223 |
+
|
224 |
+
# Then we have general tags: pick any where prediction confidence > threshold
|
225 |
+
general_names = [labels[i] for i in self.general_indexes]
|
226 |
+
|
227 |
+
if general_mcut_enabled:
|
228 |
+
general_probs = np.array([x[1] for x in general_names])
|
229 |
+
general_thresh = mcut_threshold(general_probs)
|
230 |
+
|
231 |
+
general_res = [x for x in general_names if x[1] > general_thresh]
|
232 |
+
general_res = dict(general_res)
|
233 |
+
|
234 |
+
# Everything else is characters: pick any where prediction confidence > threshold
|
235 |
+
character_names = [labels[i] for i in self.character_indexes]
|
236 |
+
|
237 |
+
if character_mcut_enabled:
|
238 |
+
character_probs = np.array([x[1] for x in character_names])
|
239 |
+
character_thresh = mcut_threshold(character_probs)
|
240 |
+
character_thresh = max(0.15, character_thresh)
|
241 |
+
|
242 |
+
character_res = [x for x in character_names if x[1] > character_thresh]
|
243 |
+
character_res = dict(character_res)
|
244 |
+
character_list = list(character_res.keys())
|
245 |
+
|
246 |
+
sorted_general_list = sorted(
|
247 |
+
general_res.items(),
|
248 |
+
key=lambda x: x[1],
|
249 |
+
reverse=True,
|
250 |
+
)
|
251 |
+
sorted_general_list = [x[0] for x in sorted_general_list]
|
252 |
+
# Remove values from character_list that already exist in sorted_general_list
|
253 |
+
character_list = [item for item in character_list if item not in sorted_general_list]
|
254 |
+
# Remove values from sorted_general_list that already exist in prepend_list or append_list
|
255 |
+
if prepend_list:
|
256 |
+
sorted_general_list = [item for item in sorted_general_list if item not in prepend_list]
|
257 |
+
if append_list:
|
258 |
+
sorted_general_list = [item for item in sorted_general_list if item not in append_list]
|
259 |
+
|
260 |
+
sorted_general_list = prepend_list + sorted_general_list + append_list
|
261 |
+
|
262 |
+
sorted_general_strings = ", ".join((character_list if characters_merge_enabled else []) + sorted_general_list).replace("(", "\(").replace(")", "\)")
|
263 |
+
|
264 |
+
classified_tags, unclassified_tags = classify_tags(sorted_general_list)
|
265 |
+
|
266 |
+
# Create a single string of ALL categorized tags for the current image
|
267 |
+
categorized_output_string = ', '.join([', '.join(tags) for tags in classified_tags.values()])
|
268 |
+
categorized_output_strings.append(categorized_output_string)
|
269 |
+
# Collect all categorized output strings into a single string
|
270 |
+
final_categorized_output = ', '.join(categorized_output_strings)
|
271 |
+
|
272 |
+
# Create a .txt file for "Output (string)" and "Categorized Output (string)"
|
273 |
+
txt_content = f"Output (string): {sorted_general_strings}\nCategorized Output (string): {final_categorized_output}"
|
274 |
+
txt_file = self.create_file(txt_content, output_dir, f"{image_name}_output.txt")
|
275 |
+
txt_infos.append({"path": txt_file, "name": f"{image_name}_output.txt"})
|
276 |
+
|
277 |
+
# Create a .json file for "Categorized (tags)"
|
278 |
+
json_content = json.dumps(classified_tags, indent=4)
|
279 |
+
json_file = self.create_file(json_content, output_dir, f"{image_name}_categorized_tags.json")
|
280 |
+
txt_infos.append({"path": json_file, "name": f"{image_name}_categorized_tags.json"})
|
281 |
+
|
282 |
+
# Save a copy of the uploaded image in PNG format
|
283 |
+
image_path = value[0]
|
284 |
+
image = Image.open(image_path)
|
285 |
+
image.save(os.path.join(output_dir, f"{image_name}.png"), format="PNG")
|
286 |
+
txt_infos.append({"path": os.path.join(output_dir, f"{image_name}.png"), "name": f"{image_name}.png"})
|
287 |
+
|
288 |
+
current_progress += progressRatio/progressTotal;
|
289 |
+
progress(current_progress, desc=f"image{idx:02d}, predict finished")
|
290 |
+
timer.checkpoint(f"image{idx:02d}, predict finished")
|
291 |
+
|
292 |
+
if reorganizer_model_repo:
|
293 |
+
print(f"Starting reorganizer...")
|
294 |
+
reorganize_strings = reorganizer.reorganize(sorted_general_strings)
|
295 |
+
reorganize_strings = re.sub(r" *Title: *", "", reorganize_strings)
|
296 |
+
reorganize_strings = re.sub(r"\n+", ",", reorganize_strings)
|
297 |
+
reorganize_strings = re.sub(r",,+", ",", reorganize_strings)
|
298 |
+
sorted_general_strings += ",\n\n" + reorganize_strings
|
299 |
+
|
300 |
+
current_progress += progressRatio/progressTotal;
|
301 |
+
progress(current_progress, desc=f"image{idx:02d}, reorganizer finished")
|
302 |
+
timer.checkpoint(f"image{idx:02d}, reorganizer finished")
|
303 |
+
|
304 |
+
txt_file = self.create_file(sorted_general_strings, output_dir, image_name + ".txt")
|
305 |
+
txt_infos.append({"path":txt_file, "name": image_name + ".txt"})
|
306 |
+
|
307 |
+
# Store the result in tag_results using image_path as the key
|
308 |
+
tag_results[image_path] = {
|
309 |
+
"strings": sorted_general_strings,
|
310 |
+
"strings2": categorized_output_string, # Store the categorized output string here
|
311 |
+
"classified_tags": classified_tags,
|
312 |
+
"rating": rating,
|
313 |
+
"character_res": character_res,
|
314 |
+
"general_res": general_res,
|
315 |
+
"unclassified_tags": unclassified_tags,
|
316 |
+
"enhanced_tags": "" # Initialize as empty string
|
317 |
+
}
|
318 |
+
|
319 |
+
timer.report()
|
320 |
+
except Exception as e:
|
321 |
+
print(traceback.format_exc())
|
322 |
+
print("Error predict: " + str(e))
|
323 |
+
# Zip creation logic:
|
324 |
+
download = []
|
325 |
+
if txt_infos is not None and len(txt_infos) > 0:
|
326 |
+
downloadZipPath = os.path.join(output_dir, "Multi-tagger-" + datetime.now().strftime("%Y%m%d-%H%M%S") + ".zip")
|
327 |
+
with zipfile.ZipFile(downloadZipPath, 'w', zipfile.ZIP_DEFLATED) as taggers_zip:
|
328 |
+
for info in txt_infos:
|
329 |
+
# Get file name from lookup
|
330 |
+
taggers_zip.write(info["path"], arcname=info["name"])
|
331 |
+
download.append(downloadZipPath)
|
332 |
+
# End zip creation logic
|
333 |
+
if reorganizer_model_repo:
|
334 |
+
reorganizer.release_vram()
|
335 |
+
del reorganizer
|
336 |
+
|
337 |
+
progress(1, desc=f"Predict completed")
|
338 |
+
timer.report_all() # Print all recorded times
|
339 |
+
print("Predict is complete.")
|
340 |
+
|
341 |
+
return download, sorted_general_strings, final_categorized_output, classified_tags, rating, character_res, general_res, unclassified_tags, tag_results
|
342 |
+
def get_selection_from_gallery(gallery: list, tag_results: dict, selected_state: gr.SelectData):
|
343 |
+
if not selected_state:
|
344 |
+
return selected_state
|
345 |
+
tag_result = {
|
346 |
+
"strings": "",
|
347 |
+
"strings2": "",
|
348 |
+
"classified_tags": "{}",
|
349 |
+
"rating": "",
|
350 |
+
"character_res": "",
|
351 |
+
"general_res": "",
|
352 |
+
"unclassified_tags": "{}",
|
353 |
+
"enhanced_tags": ""
|
354 |
+
}
|
355 |
+
if selected_state.value["image"]["path"] in tag_results:
|
356 |
+
tag_result = tag_results[selected_state.value["image"]["path"]]
|
357 |
+
return (selected_state.value["image"]["path"], selected_state.value["caption"]), tag_result["strings"], tag_result["strings2"], tag_result["classified_tags"], tag_result["rating"], tag_result["character_res"], tag_result["general_res"], tag_result["unclassified_tags"], tag_result["enhanced_tags"]
|
358 |
+
def append_gallery(gallery:list,image:str):
|
359 |
+
if gallery is None:gallery=[]
|
360 |
+
if not image:return gallery,None
|
361 |
+
gallery.append(image);return gallery,None
|
362 |
+
def extend_gallery(gallery:list,images):
|
363 |
+
if gallery is None:gallery=[]
|
364 |
+
if not images:return gallery
|
365 |
+
gallery.extend(images);return gallery
|
366 |
+
def remove_image_from_gallery(gallery:list,selected_image:str):
|
367 |
+
if not gallery or not selected_image:return gallery
|
368 |
+
selected_image=ast.literal_eval(selected_image)
|
369 |
+
if selected_image in gallery:gallery.remove(selected_image)
|
370 |
+
return gallery
|
371 |
+
args = parse_args()
|
372 |
+
predictor = Predictor()
|
373 |
+
dropdown_list = [
|
374 |
+
EVA02_LARGE_MODEL_DSV3_REPO,
|
375 |
+
SWINV2_MODEL_DSV3_REPO,
|
376 |
+
CONV_MODEL_DSV3_REPO,
|
377 |
+
VIT_MODEL_DSV3_REPO,
|
378 |
+
VIT_LARGE_MODEL_DSV3_REPO,
|
379 |
+
# ---
|
380 |
+
MOAT_MODEL_DSV2_REPO,
|
381 |
+
SWIN_MODEL_DSV2_REPO,
|
382 |
+
CONV_MODEL_DSV2_REPO,
|
383 |
+
CONV2_MODEL_DSV2_REPO,
|
384 |
+
VIT_MODEL_DSV2_REPO,
|
385 |
+
# ---
|
386 |
+
SWINV2_MODEL_IS_DSV1_REPO,
|
387 |
+
EVA02_LARGE_MODEL_IS_DSV1_REPO,
|
388 |
+
]
|
389 |
+
|
390 |
+
def _restart_space():
|
391 |
+
HF_TOKEN=os.getenv('HF_TOKEN')
|
392 |
+
if not HF_TOKEN:raise ValueError('HF_TOKEN environment variable is not set.')
|
393 |
+
huggingface_hub.HfApi().restart_space(repo_id='Werli/Multi-Tagger',token=HF_TOKEN,factory_reboot=False)
|
394 |
+
scheduler=BackgroundScheduler()
|
395 |
+
# Add a job to restart the space every 2 days (172800 seconds)
|
396 |
+
restart_space_job = scheduler.add_job(_restart_space, "interval", seconds=172800)
|
397 |
+
scheduler.start()
|
398 |
+
next_run_time_utc=restart_space_job.next_run_time.astimezone(timezone.utc)
|
399 |
+
NEXT_RESTART=f"Next Restart: {next_run_time_utc.strftime('%Y-%m-%d %H:%M:%S')} (UTC) - The space will restart every 2 days to ensure stability and performance. It uses a background scheduler to handle the restart process."
|
400 |
+
|
401 |
+
css = """
|
402 |
+
#output {height: 500px; overflow: auto; border: 1px solid #ccc;}
|
403 |
+
label.float.svelte-i3tvor {position: relative !important;}
|
404 |
+
.reduced-height.svelte-11chud3 {height: calc(80% - var(--size-10));}
|
405 |
+
"""
|
406 |
+
|
407 |
+
with gr.Blocks(title=TITLE, css=css, theme=gr.themes.Soft(), fill_width=True) as demo:
|
408 |
+
gr.Markdown(value=f"<h1 style='text-align: center; margin-bottom: 1rem'>{TITLE}</h1>")
|
409 |
+
gr.Markdown(value=DESCRIPTION)
|
410 |
+
gr.Markdown(NEXT_RESTART)
|
411 |
+
with gr.Tab(label="Waifu Diffusion"):
|
412 |
+
with gr.Row():
|
413 |
+
with gr.Column():
|
414 |
+
submit = gr.Button(value="Submit", variant="primary", size="lg")
|
415 |
+
with gr.Column(variant="panel"):
|
416 |
+
# Create an Image component for uploading images
|
417 |
+
image_input = gr.Image(label="Upload an Image or clicking paste from clipboard button", type="filepath", sources=["upload", "clipboard"], height=150)
|
418 |
+
with gr.Row():
|
419 |
+
upload_button = gr.UploadButton("Upload multiple images", file_types=["image"], file_count="multiple", size="sm")
|
420 |
+
remove_button = gr.Button("Remove Selected Image", size="sm")
|
421 |
+
gallery = gr.Gallery(columns=5, rows=5, show_share_button=False, interactive=True, height="500px", label="Grid of images")
|
422 |
+
model_repo = gr.Dropdown(
|
423 |
+
dropdown_list,
|
424 |
+
value=EVA02_LARGE_MODEL_DSV3_REPO,
|
425 |
+
label="Model",
|
426 |
+
)
|
427 |
+
with gr.Row():
|
428 |
+
general_thresh = gr.Slider(
|
429 |
+
0,
|
430 |
+
1,
|
431 |
+
step=args.score_slider_step,
|
432 |
+
value=args.score_general_threshold,
|
433 |
+
label="General Tags Threshold",
|
434 |
+
scale=3,
|
435 |
+
)
|
436 |
+
general_mcut_enabled = gr.Checkbox(
|
437 |
+
value=False,
|
438 |
+
label="Use MCut threshold",
|
439 |
+
scale=1,
|
440 |
+
)
|
441 |
+
with gr.Row():
|
442 |
+
character_thresh = gr.Slider(
|
443 |
+
0,
|
444 |
+
1,
|
445 |
+
step=args.score_slider_step,
|
446 |
+
value=args.score_character_threshold,
|
447 |
+
label="Character Tags Threshold",
|
448 |
+
scale=3,
|
449 |
+
)
|
450 |
+
character_mcut_enabled = gr.Checkbox(
|
451 |
+
value=False,
|
452 |
+
label="Use MCut threshold",
|
453 |
+
scale=1,
|
454 |
+
)
|
455 |
+
with gr.Row():
|
456 |
+
characters_merge_enabled = gr.Checkbox(
|
457 |
+
value=True,
|
458 |
+
label="Merge characters into the string output",
|
459 |
+
scale=1,
|
460 |
+
)
|
461 |
+
with gr.Row():
|
462 |
+
reorganizer_model_repo = gr.Dropdown(
|
463 |
+
[None] + reorganizer_list,
|
464 |
+
value=None,
|
465 |
+
label="Reorganizer Model",
|
466 |
+
info="Use a model to create a description for you",
|
467 |
+
)
|
468 |
+
with gr.Row():
|
469 |
+
additional_tags_prepend = gr.Text(label="Prepend Additional tags (comma split)")
|
470 |
+
additional_tags_append = gr.Text(label="Append Additional tags (comma split)")
|
471 |
+
with gr.Row():
|
472 |
+
clear = gr.ClearButton(
|
473 |
+
components=[
|
474 |
+
gallery,
|
475 |
+
model_repo,
|
476 |
+
general_thresh,
|
477 |
+
general_mcut_enabled,
|
478 |
+
character_thresh,
|
479 |
+
character_mcut_enabled,
|
480 |
+
characters_merge_enabled,
|
481 |
+
reorganizer_model_repo,
|
482 |
+
additional_tags_prepend,
|
483 |
+
additional_tags_append,
|
484 |
+
],
|
485 |
+
variant="secondary",
|
486 |
+
size="lg",
|
487 |
+
)
|
488 |
+
with gr.Column(variant="panel"):
|
489 |
+
download_file = gr.File(label="Download includes: All outputs* and image(s)") # 0
|
490 |
+
character_res = gr.Label(label="Output (characters)") # 1
|
491 |
+
sorted_general_strings = gr.Textbox(label="Output (string)*", show_label=True, show_copy_button=True) # 2
|
492 |
+
final_categorized_output = gr.Textbox(label="Categorized (string)* - If it's too long, select an image to display tags correctly.", show_label=True, show_copy_button=True) # 3
|
493 |
+
pe_generate_btn = gr.Button(value="ENHANCE TAGS", size="lg", variant="primary") # 4
|
494 |
+
enhanced_tags = gr.Textbox(label="Enhanced Tags", show_label=True, show_copy_button=True) # 5
|
495 |
+
prompt_enhancer_model = gr.Radio(["Medium", "Long", "Flux"], label="Model Choice", value="Medium", info="Enhance your prompts with Medium or Long answers") # 6
|
496 |
+
categorized = gr.JSON(label="Categorized (tags)* - JSON") # 7
|
497 |
+
rating = gr.Label(label="Rating") # 8
|
498 |
+
general_res = gr.Label(label="Output (tags)") # 9
|
499 |
+
unclassified = gr.JSON(label="Unclassified (tags)") # 10
|
500 |
+
clear.add(
|
501 |
+
[
|
502 |
+
download_file,
|
503 |
+
sorted_general_strings,
|
504 |
+
final_categorized_output,
|
505 |
+
categorized,
|
506 |
+
rating,
|
507 |
+
character_res,
|
508 |
+
general_res,
|
509 |
+
unclassified,
|
510 |
+
prompt_enhancer_model,
|
511 |
+
enhanced_tags,
|
512 |
+
]
|
513 |
+
)
|
514 |
+
tag_results = gr.State({})
|
515 |
+
# Define the event listener to add the uploaded image to the gallery
|
516 |
+
image_input.change(append_gallery, inputs=[gallery, image_input], outputs=[gallery, image_input])
|
517 |
+
# When the upload button is clicked, add the new images to the gallery
|
518 |
+
upload_button.upload(extend_gallery, inputs=[gallery, upload_button], outputs=gallery)
|
519 |
+
# Event to update the selected image when an image is clicked in the gallery
|
520 |
+
selected_image = gr.Textbox(label="Selected Image", visible=False)
|
521 |
+
gallery.select(get_selection_from_gallery,inputs=[gallery, tag_results],outputs=[selected_image, sorted_general_strings, final_categorized_output, categorized, rating, character_res, general_res, unclassified, enhanced_tags])
|
522 |
+
# Event to remove a selected image from the gallery
|
523 |
+
remove_button.click(remove_image_from_gallery, inputs=[gallery, selected_image], outputs=gallery)
|
524 |
+
# Event to for the Prompt Enhancer Button
|
525 |
+
pe_generate_btn.click(lambda tags,model:prompt_enhancer('','',tags,model)[0],inputs=[final_categorized_output,prompt_enhancer_model],outputs=[enhanced_tags])
|
526 |
+
submit.click(
|
527 |
+
predictor.predict,
|
528 |
+
inputs=[
|
529 |
+
gallery,
|
530 |
+
model_repo,
|
531 |
+
general_thresh,
|
532 |
+
general_mcut_enabled,
|
533 |
+
character_thresh,
|
534 |
+
character_mcut_enabled,
|
535 |
+
characters_merge_enabled,
|
536 |
+
reorganizer_model_repo,
|
537 |
+
additional_tags_prepend,
|
538 |
+
additional_tags_append,
|
539 |
+
tag_results,
|
540 |
+
],
|
541 |
+
outputs=[download_file, sorted_general_strings, final_categorized_output, categorized, rating, character_res, general_res, unclassified, tag_results,],
|
542 |
+
)
|
543 |
+
gr.Examples(
|
544 |
+
[["images/1girl.png", VIT_LARGE_MODEL_DSV3_REPO, 0.35, False, 0.85, False]],
|
545 |
+
inputs=[
|
546 |
+
image_input,
|
547 |
+
model_repo,
|
548 |
+
general_thresh,
|
549 |
+
general_mcut_enabled,
|
550 |
+
character_thresh,
|
551 |
+
character_mcut_enabled,
|
552 |
+
],
|
553 |
+
)
|
554 |
+
with gr.Tab(label="Florence 2 Image Captioning"):
|
555 |
+
with gr.Row():
|
556 |
+
with gr.Column(variant="panel"):
|
557 |
+
input_img = gr.Image(label="Input Picture")
|
558 |
+
task_type = gr.Radio(choices=['Single task', 'Cascaded task'], label='Task type selector', value='Single task')
|
559 |
+
task_prompt = gr.Dropdown(choices=single_task_list, label="Task Prompt", value="Caption")
|
560 |
+
task_type.change(fn=update_task_dropdown, inputs=task_type, outputs=task_prompt)
|
561 |
+
text_input = gr.Textbox(label="Text Input (optional)")
|
562 |
+
submit_btn = gr.Button(value="Submit")
|
563 |
+
with gr.Column(variant="panel"):
|
564 |
+
output_text = gr.Textbox(label="Output Text", show_label=True, show_copy_button=True, lines=8)
|
565 |
+
output_img = gr.Image(label="Output Image")
|
566 |
+
gr.Examples(
|
567 |
+
examples=[
|
568 |
+
["images/image1.png", 'Object Detection'],
|
569 |
+
["images/image2.png", 'OCR with Region']
|
570 |
+
],
|
571 |
+
inputs=[input_img, task_prompt],
|
572 |
+
outputs=[output_text, output_img],
|
573 |
+
fn=process_image,
|
574 |
+
cache_examples=False,
|
575 |
+
label='Try examples'
|
576 |
+
)
|
577 |
+
submit_btn.click(process_image, [input_img, task_prompt, text_input], [output_text, output_img])
|
578 |
+
with gr.Tab("Booru Image Fetcher"):
|
579 |
+
with gr.Row():
|
580 |
+
with gr.Column():
|
581 |
+
gr.Markdown("### ⚙️ Search Parameters")
|
582 |
+
site = gr.Dropdown(label="Select Source", choices=["Gelbooru", "Rule34", "Xbooru"], value="Gelbooru")
|
583 |
+
Tags = gr.Textbox(label="Tags (comma-separated)", placeholder="e.g. solo, 1girl, 1boy, artist name, character, black hair, cat ears, holding, granblue fantasy, ...")
|
584 |
+
exclude_tags = gr.Textbox(label="Exclude Tags (comma-separated)", placeholder="e.g. animated, watermark, username, ...")
|
585 |
+
score = gr.Number(label="Minimum Score", value=0)
|
586 |
+
count = gr.Slider(label="Number of Images", minimum=1, maximum=4, step=1, value=1)
|
587 |
+
Safe = gr.Checkbox(label="Include Safe", value=True)
|
588 |
+
Questionable = gr.Checkbox(label="Include Questionable", value=True)
|
589 |
+
Explicit = gr.Checkbox(label="Include Explicit", value=False)
|
590 |
+
submit_btn = gr.Button("Fetch Images", variant="primary")
|
591 |
+
|
592 |
+
with gr.Column():
|
593 |
+
gr.Markdown("### 📄 Results")
|
594 |
+
images_output = gr.Gallery(label="Images", columns=3, rows=2, object_fit="contain", height=500)
|
595 |
+
tags_output = gr.Textbox(label="Tags", placeholder="Select an image to show tags", lines=5, show_copy_button=True)
|
596 |
+
post_url_output = gr.Textbox(label="Post URL", lines=1, show_copy_button=True)
|
597 |
+
image_url_output = gr.Textbox(label="Image URL", lines=1, show_copy_button=True)
|
598 |
+
|
599 |
+
# State to store tags, URLs
|
600 |
+
tags_state = gr.State([])
|
601 |
+
post_url_state = gr.State([])
|
602 |
+
image_url_state = gr.State([])
|
603 |
+
|
604 |
+
submit_btn.click(
|
605 |
+
fn=booru_gradio,
|
606 |
+
inputs=[Tags, exclude_tags, score, count, Safe, Questionable, Explicit, site],
|
607 |
+
outputs=[images_output, tags_state, post_url_state, image_url_state],
|
608 |
+
)
|
609 |
+
|
610 |
+
images_output.select(
|
611 |
+
fn=on_select,
|
612 |
+
inputs=[tags_state, post_url_state, image_url_state],
|
613 |
+
outputs=[tags_output, post_url_output, image_url_output],
|
614 |
+
)
|
615 |
+
gr.Markdown("""
|
616 |
+
---
|
617 |
+
ComfyUI version: [Comfyui-Gelbooru](https://github.com/1mckw/Comfyui-Gelbooru)
|
618 |
+
""")
|
619 |
+
with gr.Tab(label="Categorizer++"):
|
620 |
+
with gr.Row():
|
621 |
+
with gr.Column(variant="panel"):
|
622 |
+
input_tags = gr.Textbox(label="Input Tags", placeholder="1girl, cat, horns, blue hair, ...\nor\n? 1girl 1234567? cat 1234567? horns 1234567? blue hair 1234567? ...", lines=4)
|
623 |
+
submit_button = gr.Button(value="Submit", variant="primary", size="lg")
|
624 |
+
with gr.Column(variant="panel"):
|
625 |
+
categorized_string = gr.Textbox(label="Categorized (string)", show_label=True, show_copy_button=True, lines=8)
|
626 |
+
categorized_json = gr.JSON(label="Categorized (tags) - JSON")
|
627 |
+
submit_button.click(process_tags, inputs=[input_tags], outputs=[categorized_string, categorized_json])
|
628 |
+
with gr.Column(variant="panel"):
|
629 |
+
pe_generate_btn = gr.Button(value="ENHANCE TAGS", size="lg", variant="primary")
|
630 |
+
enhanced_tags = gr.Textbox(label="Enhanced Tags", show_label=True, show_copy_button=True)
|
631 |
+
prompt_enhancer_model = gr.Radio(["Medium", "Long", "Flux"], label="Model Choice", value="Medium", info="Enhance your prompts with Medium or Long answers")
|
632 |
+
pe_generate_btn.click(lambda tags,model:prompt_enhancer('','',tags,model)[0],inputs=[categorized_string,prompt_enhancer_model],outputs=[enhanced_tags])
|
|
|
|
|
|
|
|
|
633 |
demo.queue(max_size=2).launch()
|
modules/booru.py
CHANGED
@@ -1,132 +1,111 @@
|
|
1 |
-
import requests,re,base64,io,numpy as np
|
2 |
-
from PIL import Image,ImageOps
|
3 |
-
import torch,gradio as gr
|
4 |
-
|
5 |
-
# Helper to load image from URL
|
6 |
-
def loadImageFromUrl(url):
|
7 |
-
|
8 |
-
|
9 |
-
|
10 |
-
|
11 |
-
|
12 |
-
|
13 |
-
|
14 |
-
|
15 |
-
|
16 |
-
|
17 |
-
|
18 |
-
|
19 |
-
|
20 |
-
|
21 |
-
|
22 |
-
|
23 |
-
|
24 |
-
|
25 |
-
|
26 |
-
|
27 |
-
|
28 |
-
|
29 |
-
|
30 |
-
|
31 |
-
|
32 |
-
|
33 |
-
|
34 |
-
|
35 |
-
|
36 |
-
|
37 |
-
|
38 |
-
|
39 |
-
|
40 |
-
|
41 |
-
#
|
42 |
-
|
43 |
-
|
44 |
-
|
45 |
-
|
46 |
-
if
|
47 |
-
|
48 |
-
|
49 |
-
|
50 |
-
|
51 |
-
|
52 |
-
|
53 |
-
|
54 |
-
|
55 |
-
|
56 |
-
|
57 |
-
|
58 |
-
|
59 |
-
|
60 |
-
|
61 |
-
|
62 |
-
|
63 |
-
|
64 |
-
|
65 |
-
|
66 |
-
|
67 |
-
|
68 |
-
|
69 |
-
|
70 |
-
|
71 |
-
|
72 |
-
|
73 |
-
|
74 |
-
|
75 |
-
|
76 |
-
|
77 |
-
|
78 |
-
|
79 |
-
|
80 |
-
|
81 |
-
|
82 |
-
|
83 |
-
|
84 |
-
|
85 |
-
|
86 |
-
|
87 |
-
|
88 |
-
|
89 |
-
|
90 |
-
|
91 |
-
|
92 |
-
|
93 |
-
|
94 |
-
|
95 |
-
|
96 |
-
|
97 |
-
|
98 |
-
|
99 |
-
|
100 |
-
|
101 |
-
|
102 |
-
|
103 |
-
|
104 |
-
|
105 |
-
|
106 |
-
|
107 |
-
|
108 |
-
|
109 |
-
)
|
110 |
-
|
111 |
-
if not image_urls:
|
112 |
-
return [], [], [], []
|
113 |
-
|
114 |
-
image_data = []
|
115 |
-
for url in image_urls:
|
116 |
-
try:
|
117 |
-
image = loadImageFromUrl(url)
|
118 |
-
image = (image * 255).clamp(0, 255).cpu().numpy().astype(np.uint8)[0]
|
119 |
-
image = Image.fromarray(image)
|
120 |
-
image_data.append(image)
|
121 |
-
except Exception as e:
|
122 |
-
print(f"Error loading image from {url}: {e}")
|
123 |
-
continue
|
124 |
-
|
125 |
-
return image_data, tags_list, post_urls, image_urls
|
126 |
-
|
127 |
-
# Update UI on image click
|
128 |
-
def on_select(evt: gr.SelectData, tags_list, post_url_list, image_url_list):
|
129 |
-
idx = evt.index
|
130 |
-
if idx < len(tags_list):
|
131 |
-
return tags_list[idx], post_url_list[idx], image_url_list[idx]
|
132 |
return "No tags", "", ""
|
|
|
1 |
+
import requests,re,base64,io,numpy as np
|
2 |
+
from PIL import Image,ImageOps
|
3 |
+
import torch,gradio as gr
|
4 |
+
|
5 |
+
# Helper to load image from URL
|
6 |
+
def loadImageFromUrl(url):
|
7 |
+
response = requests.get(url, timeout=10)
|
8 |
+
if response.status_code != 200:
|
9 |
+
raise Exception(f"Failed to load image from {url}")
|
10 |
+
i = Image.open(io.BytesIO(response.content))
|
11 |
+
i = ImageOps.exif_transpose(i)
|
12 |
+
if i.mode != "RGBA":
|
13 |
+
i = i.convert("RGBA")
|
14 |
+
alpha = i.split()[-1]
|
15 |
+
image = Image.new("RGB", i.size, (0, 0, 0))
|
16 |
+
image.paste(i, mask=alpha)
|
17 |
+
image = np.array(image).astype(np.float32) / 255.0
|
18 |
+
image = torch.from_numpy(image)[None,]
|
19 |
+
return image
|
20 |
+
|
21 |
+
# Fetch data from multiple booru platforms
|
22 |
+
def fetch_booru_images(site, Tags, exclude_tags, score, count, Safe, Questionable, Explicit):
|
23 |
+
# Clean and format tags
|
24 |
+
def clean_tag_list(tags):
|
25 |
+
return [item.strip().replace(' ', '_') for item in tags.split(',') if item.strip()]
|
26 |
+
|
27 |
+
Tags = '+'.join(clean_tag_list(Tags)) if Tags else ''
|
28 |
+
exclude_tags = '+'.join('-' + tag for tag in clean_tag_list(exclude_tags))
|
29 |
+
|
30 |
+
rating_filters = []
|
31 |
+
if not Safe:
|
32 |
+
rating_filters.extend(["rating:safe", "rating:general"])
|
33 |
+
if not Questionable:
|
34 |
+
rating_filters.extend(["rating:questionable", "rating:sensitive"])
|
35 |
+
if not Explicit:
|
36 |
+
rating_filters.append("rating:explicit")
|
37 |
+
rating_filters = '+'.join(f'-{r}' for r in rating_filters)
|
38 |
+
|
39 |
+
score_filter = f"score:>{score}"
|
40 |
+
|
41 |
+
# Build query
|
42 |
+
base_query = f"tags=sort:random+{Tags}+{exclude_tags}+{score_filter}+{rating_filters}&limit={count}&json=1"
|
43 |
+
base_query = re.sub(r"\++", "+", base_query)
|
44 |
+
|
45 |
+
# Fetch data based on site
|
46 |
+
if site == "Gelbooru":
|
47 |
+
url = f"https://gelbooru.com/index.php?page=dapi&s=post&q=index&{base_query}"
|
48 |
+
response = requests.get(url).json()
|
49 |
+
posts = response.get("post", [])
|
50 |
+
elif site == "Rule34":
|
51 |
+
url = f"https://api.rule34.xxx/index.php?page=dapi&s=post&q=index&{base_query}"
|
52 |
+
response = requests.get(url).json()
|
53 |
+
posts = response
|
54 |
+
elif site == "Xbooru":
|
55 |
+
url = f"https://xbooru.com/index.php?page=dapi&s=post&q=index&{base_query}"
|
56 |
+
response = requests.get(url).json()
|
57 |
+
posts = response
|
58 |
+
else:
|
59 |
+
return [], [], []
|
60 |
+
|
61 |
+
# Extract image URLs, tags, and post URLs
|
62 |
+
image_urls = []
|
63 |
+
tags_list = [post.get("tags", "").replace(" ", ", ").replace("_", " ").replace("(", "\\(").replace(")", "\\)").strip() for post in posts]
|
64 |
+
post_urls = []
|
65 |
+
|
66 |
+
for post in posts:
|
67 |
+
if site in ["Gelbooru", "Rule34", "Xbooru"]:
|
68 |
+
file_url = post.get("file_url")
|
69 |
+
tags = post.get("tags", "").replace(" ", ", ").strip()
|
70 |
+
post_id = post.get("id", "")
|
71 |
+
else:
|
72 |
+
continue
|
73 |
+
|
74 |
+
if file_url:
|
75 |
+
image_urls.append(file_url)
|
76 |
+
tags_list.append(tags)
|
77 |
+
if site == "Gelbooru":
|
78 |
+
post_urls.append(f"https://gelbooru.com/index.php?page=post&s=view&id={post_id}")
|
79 |
+
elif site == "Rule34":
|
80 |
+
post_urls.append(f"https://rule34.xxx/index.php?page=post&s=view&id={post_id}")
|
81 |
+
elif site == "Xbooru":
|
82 |
+
post_urls.append(f"https://xbooru.com/index.php?page=post&s=view&id={post_id}")
|
83 |
+
|
84 |
+
return image_urls, tags_list, post_urls
|
85 |
+
|
86 |
+
# Main function to fetch and return processed images
|
87 |
+
def booru_gradio(Tags, exclude_tags, score, count, Safe, Questionable, Explicit, site):
|
88 |
+
image_urls, tags_list, post_urls = fetch_booru_images(site, Tags, exclude_tags, score, count, Safe, Questionable, Explicit)
|
89 |
+
|
90 |
+
if not image_urls:
|
91 |
+
return [], [], [], []
|
92 |
+
|
93 |
+
image_data = []
|
94 |
+
for url in image_urls:
|
95 |
+
try:
|
96 |
+
image = loadImageFromUrl(url)
|
97 |
+
image = (image * 255).clamp(0, 255).cpu().numpy().astype(np.uint8)[0]
|
98 |
+
image = Image.fromarray(image)
|
99 |
+
image_data.append(image)
|
100 |
+
except Exception as e:
|
101 |
+
print(f"Error loading image from {url}: {e}")
|
102 |
+
continue
|
103 |
+
|
104 |
+
return image_data, tags_list, post_urls, image_urls
|
105 |
+
|
106 |
+
# Update UI on image click
|
107 |
+
def on_select(evt: gr.SelectData, tags_list, post_url_list, image_url_list):
|
108 |
+
idx = evt.index
|
109 |
+
if idx < len(tags_list):
|
110 |
+
return tags_list[idx], post_url_list[idx], image_url_list[idx]
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
111 |
return "No tags", "", ""
|