File size: 9,607 Bytes
23fa49c 0bd8f65 23fa49c 0bd8f65 0c14216 0bd8f65 0c14216 0bd8f65 0c14216 0bd8f65 23fa49c 0c14216 23fa49c 0c14216 0a8bf1b 0bd8f65 0a8bf1b 0c14216 0a8bf1b 23fa49c 0c14216 23fa49c 0bd8f65 23fa49c 0bd8f65 23fa49c 0bd8f65 23fa49c 0bd8f65 23fa49c 0bd8f65 23fa49c 0a8bf1b 23fa49c 0bd8f65 23fa49c 0a8bf1b 0bd8f65 0c14216 0bd8f65 0a8bf1b 0c14216 0bd8f65 23fa49c 9934ab5 0aa5488 0bd8f65 0aa5488 0c14216 0aa5488 0bd8f65 0aa5488 0c14216 0aa5488 9934ab5 0bd8f65 9934ab5 74bd9c8 0c14216 9934ab5 0aa5488 0bd8f65 0c14216 0bd8f65 0c14216 0bd8f65 9934ab5 0bd8f65 9934ab5 0c14216 9934ab5 0aa5488 8444c60 9934ab5 23fa49c 0a8bf1b 0bd8f65 0a8bf1b 0c14216 0a8bf1b 23fa49c |
1 2 3 4 5 6 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 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 99 100 101 102 103 104 105 106 107 108 109 110 111 112 113 114 115 116 117 118 119 120 121 122 123 124 125 126 127 128 129 130 131 132 133 134 135 136 137 138 139 140 141 142 143 144 145 146 147 148 149 150 151 152 153 154 155 156 157 158 159 160 161 162 163 164 165 166 167 168 169 170 171 172 173 174 175 176 177 178 179 180 181 182 183 184 185 186 187 188 189 190 191 192 193 194 195 196 197 198 199 200 201 202 203 204 205 206 207 208 209 210 211 212 213 214 215 216 217 218 219 220 221 222 223 224 225 226 227 228 229 230 231 232 233 234 235 236 237 238 239 240 241 242 243 244 245 246 247 248 249 250 251 252 253 254 255 256 257 258 259 260 261 262 263 264 265 266 267 268 269 270 271 272 273 274 275 276 277 278 279 280 281 282 283 284 285 286 287 288 289 290 291 292 293 294 295 296 297 298 299 300 301 302 303 304 305 306 307 308 309 310 311 312 313 314 315 |
import json
import faiss
import flax
import gradio as gr
import jax
import numpy as np
import pandas as pd
import requests
from imgutils.tagging import wd14
from Models.CLIP import CLIP
def combine_embeddings(pos_img_embs, pos_tags_embs, neg_img_embs, neg_tags_embs):
pos = pos_img_embs + pos_tags_embs
faiss.normalize_L2(pos)
neg = neg_img_embs + neg_tags_embs
faiss.normalize_L2(neg)
result = pos - neg
faiss.normalize_L2(result)
return result
def danbooru_id_to_url(image_id, selected_ratings, api_username="", api_key=""):
headers = {"User-Agent": "image_similarity_tool"}
ratings_to_letters = {
"General": "g",
"Sensitive": "s",
"Questionable": "q",
"Explicit": "e",
}
acceptable_ratings = [ratings_to_letters[x] for x in selected_ratings]
image_url = f"https://danbooru.donmai.us/posts/{image_id}.json"
if api_username != "" and api_key != "":
image_url = f"{image_url}?api_key={api_key}&login={api_username}"
r = requests.get(image_url, headers=headers)
if r.status_code != 200:
return None
content = json.loads(r.text)
image_url = content["large_file_url"] if "large_file_url" in content else None
image_url = image_url if content["rating"] in acceptable_ratings else None
return image_url
class Predictor:
def __init__(self):
self.loaded_variant = None
self.base_model = "wd-v1-4-convnext-tagger-v2"
self.model = CLIP()
self.tags_df = pd.read_csv("data/selected_tags.csv")
self.images_ids = np.load("index/cosine_ids.npy")
self.knn_index = faiss.read_index("index/cosine_knn.index")
config = json.loads(open("index/cosine_infos.json").read())["index_param"]
faiss.ParameterSpace().set_index_parameters(self.knn_index, config)
def load_params(self, variant):
if self.loaded_variant == variant:
return
if variant == "CLIP":
with open(f"data/{self.base_model}/clip.msgpack", "rb") as f:
data = f.read()
elif variant == "SigLIP":
with open(f"data/{self.base_model}/siglip.msgpack", "rb") as f:
data = f.read()
self.params = flax.serialization.msgpack_restore(data)["model"]
self.loaded_variant = variant
def predict(
self,
pos_img_input,
neg_img_input,
positive_tags,
negative_tags,
selected_model,
selected_ratings,
n_neighbours,
api_username,
api_key,
):
tags_df = self.tags_df
model = self.model
self.load_params(selected_model)
num_classes = len(tags_df)
output_shape = model.out_units
pos_img_embs = np.zeros((1, output_shape), dtype=np.float32)
neg_img_embs = np.zeros((1, output_shape), dtype=np.float32)
pos_tags_embs = np.zeros((1, output_shape), dtype=np.float32)
neg_tags_embs = np.zeros((1, output_shape), dtype=np.float32)
positive_tags = positive_tags.split(",")
negative_tags = negative_tags.split(",")
positive_tags_idxs = tags_df.index[tags_df["name"].isin(positive_tags)].tolist()
negative_tags_idxs = tags_df.index[tags_df["name"].isin(negative_tags)].tolist()
if pos_img_input is not None:
pos_img_embs = wd14.get_wd14_tags(
pos_img_input,
model_name="ConvNext",
fmt=("embedding"),
)
pos_img_embs = np.expand_dims(pos_img_embs, 0)
faiss.normalize_L2(pos_img_embs)
if neg_img_input is not None:
neg_img_embs = wd14.get_wd14_tags(
neg_img_input,
model_name="ConvNext",
fmt=("embedding"),
)
neg_img_embs = np.expand_dims(neg_img_embs, 0)
faiss.normalize_L2(neg_img_embs)
if len(positive_tags_idxs) > 0:
tags = np.zeros((1, num_classes), dtype=np.float32)
tags[0][positive_tags_idxs] = 1
pos_tags_embs = model.apply(
{"params": self.params},
tags,
method=model.encode_text,
)
pos_tags_embs = jax.device_get(pos_tags_embs)
faiss.normalize_L2(pos_tags_embs)
if len(negative_tags_idxs) > 0:
tags = np.zeros((1, num_classes), dtype=np.float32)
tags[0][negative_tags_idxs] = 1
neg_tags_embs = model.apply(
{"params": self.params},
tags,
method=model.encode_text,
)
neg_tags_embs = jax.device_get(neg_tags_embs)
faiss.normalize_L2(neg_tags_embs)
embeddings = combine_embeddings(
pos_img_embs,
pos_tags_embs,
neg_img_embs,
neg_tags_embs,
)
dists, indexes = self.knn_index.search(embeddings, k=n_neighbours)
neighbours_ids = self.images_ids[indexes][0]
neighbours_ids = [int(x) for x in neighbours_ids]
captions = []
image_urls = []
for image_id, dist in zip(neighbours_ids, dists[0]):
current_url = danbooru_id_to_url(
image_id,
selected_ratings,
api_username,
api_key,
)
if current_url is not None:
image_urls.append(current_url)
captions.append(f"{image_id}/{dist:.2f}")
return list(zip(image_urls, captions))
def main():
predictor = Predictor()
with gr.Blocks() as demo:
with gr.Row():
pos_img_input = gr.Image(type="pil", label="Positive input")
neg_img_input = gr.Image(type="pil", label="Negative input")
with gr.Row():
with gr.Column():
positive_tags = gr.Textbox(label="Positive tags")
negative_tags = gr.Textbox(label="Negative tags")
with gr.Column():
selected_model = gr.Radio(
choices=["CLIP", "SigLIP"],
value="CLIP",
label="Model",
)
n_neighbours = gr.Slider(
minimum=1,
maximum=20,
value=5,
step=1,
label="# of images",
)
with gr.Column():
selected_ratings = gr.CheckboxGroup(
choices=["General", "Sensitive", "Questionable", "Explicit"],
value=["General", "Sensitive"],
label="Ratings",
)
with gr.Row():
api_username = gr.Textbox(label="Danbooru API Username")
api_key = gr.Textbox(label="Danbooru API Key")
find_btn = gr.Button("Find similar images")
similar_images = gr.Gallery(label="Similar images", columns=[5])
examples = gr.Examples(
[
[
None,
None,
"marcille_donato",
"",
"CLIP",
["General", "Sensitive"],
5,
"",
"",
],
[
None,
None,
"yellow_eyes,red_horns",
"",
"CLIP",
["General", "Sensitive"],
5,
"",
"",
],
[
None,
None,
"artoria_pendragon_(fate),solo",
"green_eyes",
"CLIP",
["General", "Sensitive"],
5,
"",
"",
],
[
"examples/60378883_p0.jpg",
None,
"fujimaru_ritsuka_(female)",
"solo",
"CLIP",
["General", "Sensitive"],
5,
"",
"",
],
[
"examples/DaRlExxUwAAcUOS-orig.jpg",
"examples/46657164_p1.jpg",
"",
"",
"CLIP",
["General", "Sensitive"],
5,
"",
"",
],
],
inputs=[
pos_img_input,
neg_img_input,
positive_tags,
negative_tags,
selected_model,
selected_ratings,
n_neighbours,
api_username,
api_key,
],
outputs=[similar_images],
fn=predictor.predict,
run_on_click=True,
cache_examples=False,
)
find_btn.click(
fn=predictor.predict,
inputs=[
pos_img_input,
neg_img_input,
positive_tags,
negative_tags,
selected_model,
selected_ratings,
n_neighbours,
api_username,
api_key,
],
outputs=[similar_images],
)
demo.queue()
demo.launch()
if __name__ == "__main__":
main()
|