Image_Inversion / app.py
IdlecloudX's picture
Update app.py
01d7dca verified
raw
history blame
16.7 kB
import os
import json
import gradio as gr
import huggingface_hub
import numpy as np
import onnxruntime as rt
import pandas as pd
from PIL import Image
from huggingface_hub import whoami, HfApi
from translator import translate_texts
# ------------------------------------------------------------------
# Model Configuration
# ------------------------------------------------------------------
MODEL_REPO = "SmilingWolf/wd-eva02-large-tagger-v3"
MODEL_FILENAME = "model.onnx"
LABEL_FILENAME = "selected_tags.csv"
# It's recommended to manage the token within the HF Spaces secrets
HF_TOKEN = os.environ.get("HF_TOKEN")
# A more robust way to get the space owner
SPACE_ID = os.environ.get("SPACE_ID")
SPACE_OWNER = SPACE_ID.split('/')[0] if SPACE_ID else None
# ------------------------------------------------------------------
# Tagger Class (Global Instance)
# ------------------------------------------------------------------
class Tagger:
def __init__(self):
self.hf_token = HF_TOKEN
self.tag_names = []
self.categories = {}
self.model = None
self.input_size = 0
self._load_model_and_labels()
def _load_model_and_labels(self):
try:
label_path = huggingface_hub.hf_hub_download(
MODEL_REPO, LABEL_FILENAME, token=self.hf_token, resume_download=True
)
model_path = huggingface_hub.hf_hub_download(
MODEL_REPO, MODEL_FILENAME, token=self.hf_token, resume_download=True
)
tags_df = pd.read_csv(label_path)
self.tag_names = tags_df["name"].tolist()
self.categories = {
"rating": np.where(tags_df["category"] == 9)[0],
"general": np.where(tags_df["category"] == 0)[0],
"character": np.where(tags_df["category"] == 4)[0],
}
self.model = rt.InferenceSession(model_path)
self.input_size = self.model.get_inputs()[0].shape[1]
print("✅ Model and labels loaded successfully.")
except Exception as e:
print(f"❌ Failed to load model or labels: {e}")
raise RuntimeError(f"Model initialization failed: {e}")
# ------------------------- preprocess -------------------------
def _preprocess(self, img: Image.Image) -> np.ndarray:
if img is None: raise ValueError("Input image cannot be None.")
if img.mode != "RGB": img = img.convert("RGB")
size = max(img.size)
canvas = Image.new("RGB", (size, size), (255, 255, 255))
canvas.paste(img, ((size - img.width) // 2, (size - img.height) // 2))
if size != self.input_size:
canvas = canvas.resize((self.input_size, self.input_size), Image.BICUBIC)
return np.array(canvas)[:, :, ::-1].astype(np.float32)
# --------------------------- predict --------------------------
def predict(self, img: Image.Image, gen_th: float = 0.35, char_th: float = 0.85):
if self.model is None: raise RuntimeError("Model not loaded, cannot predict.")
inp_name = self.model.get_inputs()[0].name
outputs = self.model.run(None, {inp_name: self._preprocess(img)[None, ...]})[0][0]
res = {"ratings": {}, "general": {}, "characters": {}}
tag_categories_for_translation = {"ratings": [], "general": [], "characters": []}
for cat_key, cat_indices in self.categories.items():
sub_res = {}
if cat_key == "rating":
for idx in cat_indices:
tag_name = self.tag_names[idx].replace("_", " ")
sub_res[tag_name] = float(outputs[idx])
else:
threshold = char_th if cat_key == "character" else gen_th
for idx in cat_indices:
if outputs[idx] > threshold:
tag_name = self.tag_names[idx].replace("_", " ")
sub_res[tag_name] = float(outputs[idx])
res_key = "characters" if cat_key == "character" else cat_key
res[res_key] = dict(sorted(sub_res.items(), key=lambda kv: kv[1], reverse=True))
tag_categories_for_translation[res_key] = list(res[res_key].keys())
return res, tag_categories_for_translation
# Global Tagger instance
try:
tagger_instance = Tagger()
except RuntimeError as e:
print(f"Tagger initialization failed on app startup: {e}")
tagger_instance = None
# ------------------------------------------------------------------
# Gradio UI
# ------------------------------------------------------------------
custom_css = """
.label-container { max-height: 300px; overflow-y: auto; border: 1px solid #ddd; padding: 10px; border-radius: 5px; background-color: #f9f9f9; }
.tag-item { display: flex; justify-content: space-between; align-items: center; margin: 2px 0; padding: 2px 5px; border-radius: 3px; background-color: #fff; transition: background-color 0.2s; }
.tag-item:hover { background-color: #f0f0f0; }
.tag-en { font-weight: bold; color: #333; cursor: pointer; }
.tag-zh { color: #666; margin-left: 10px; }
.tag-score { color: #999; font-size: 0.9em; }
.btn-analyze-container { margin-top: 15px; margin-bottom: 15px; }
"""
_js_functions = """
function copyToClipboard(text) {
if (typeof text === 'undefined' || text === null) {
console.warn('copyToClipboard was called with undefined or null text.');
return;
}
navigator.clipboard.writeText(text).then(() => {
const feedback = document.createElement('div');
let displayText = String(text).substring(0, 30) + (String(text).length > 30 ? '...' : '');
feedback.textContent = '已复制: ' + displayText;
Object.assign(feedback.style, {
position: 'fixed', bottom: '20px', left: '50%', transform: 'translateX(-50%)',
backgroundColor: '#4CAF50', color: 'white', padding: '10px 20px',
borderRadius: '5px', zIndex: '10000', transition: 'opacity 0.5s ease-out'
});
document.body.appendChild(feedback);
setTimeout(() => {
feedback.style.opacity = '0';
setTimeout(() => { if (document.body.contains(feedback)) document.body.removeChild(feedback); }, 500);
}, 1500);
}).catch(err => {
console.error('Failed to copy tag. Error:', err, 'Attempted to copy text:', text);
});
}
"""
with gr.Blocks(theme=gr.themes.Soft(), title="AI 图像标签分析器", css=custom_css, js=_js_functions) as demo:
gr.Markdown("# 🖼️ AI 图像标签分析器")
gr.Markdown("上传图片自动识别标签,支持中英文显示和一键复制。[NovelAI在线绘画](https://nai.idlecloud.cc/)")
with gr.Row():
with gr.Column(scale=1):
login_button = gr.LoginButton(value="🤗 通过 Hugging Face 登录")
user_status_md = gr.Markdown("ℹ️ 正在检查登录状态...")
state_res = gr.State({})
state_translations_dict = gr.State({})
with gr.Row():
with gr.Column(scale=1):
img_in = gr.Image(type="pil", label="上传图片", height=300)
btn = gr.Button("🚀 开始分析", variant="primary", elem_classes=["btn-analyze-container"])
with gr.Accordion("⚙️ 高级设置", open=False):
gen_slider = gr.Slider(0, 1, value=0.35, step=0.01, label="通用标签阈值")
char_slider = gr.Slider(0, 1, value=0.85, step=0.01, label="角色标签阈值")
show_tag_scores = gr.Checkbox(True, label="在列表中显示标签置信度")
with gr.Accordion("🔑 自定义翻译密钥 (可选)", open=False, visible=False) as api_key_accordion:
gr.Markdown("如果你不是空间所有者,需要在这里提供自己的API密钥才能使用翻译功能。")
tencent_id_in = gr.Textbox(label="腾讯云 Secret ID", lines=1)
tencent_key_in = gr.Textbox(label="腾讯云 Secret Key", lines=1, type="password")
baidu_json_in = gr.Textbox(label="百度翻译凭证 (JSON 格式)", lines=3, placeholder='[{"app_id": "...", "secret_key": "..."}]')
with gr.Accordion("📊 标签汇总设置", open=True):
sum_cats = gr.CheckboxGroup(["通用标签", "角色标签", "评分标签"], value=["通用标签", "角色标签"], label="汇总类别")
sum_sep = gr.Dropdown(["逗号", "换行", "空格"], value="逗号", label="标签分隔符")
sum_show_zh = gr.Checkbox(False, label="在汇总中显示中文翻译")
processing_info = gr.Markdown("", visible=False)
with gr.Column(scale=2):
with gr.Tabs():
with gr.TabItem("🏷️ 通用标签"): out_general = gr.HTML(label="General Tags")
with gr.TabItem("👤 角色标签"): out_char = gr.HTML(label="Character Tags")
with gr.TabItem("⭐ 评分标签"): out_rating = gr.HTML(label="Rating Tags")
gr.Markdown("### 标签汇总结果")
out_summary = gr.Textbox(label="标签汇总", lines=5, show_copy_button=True)
def get_token_from_request(request: gr.Request) -> str | None:
auth_header = request.headers.get("authorization")
if auth_header and auth_header.startswith("Bearer "):
return auth_header.split(" ")[1]
return None
def is_user_space_owner(user_info: dict | None) -> bool:
"""
Robustly checks if the user is the owner of the space by parsing SPACE_ID.
"""
if not user_info or not SPACE_OWNER:
if not SPACE_OWNER:
print("⚠️ Warning: SPACE_ID environment variable not found.")
return False
user_name = user_info.get("name")
user_orgs = [org.get("name") for org in user_info.get("orgs", [])]
print(f"ℹ️ [Auth Check] Space Owner: '{SPACE_OWNER}', User: '{user_name}', User Orgs: {user_orgs}")
is_owner = (user_name == SPACE_OWNER) or (SPACE_OWNER in user_orgs)
return is_owner
def check_user_status(request: gr.Request):
token = get_token_from_request(request)
if token:
try:
user_info = whoami(token=token)
if is_user_space_owner(user_info):
return f"✅ 以所有者 **{user_info.get('fullname', user_info.get('name'))}** 身份登录,将使用空间配置的密钥。", gr.update(visible=False)
else:
return f"👋 你好, **{user_info.get('fullname', '用户')}**!请在下方提供你自己的翻译 API 密钥。", gr.update(visible=True, open=True)
except Exception as e:
print(f"Error getting user info: {e}")
return "⚠️ 无法验证您的登录状态。请提供 API 密钥。", gr.update(visible=True, open=True)
return "ℹ️ **访客模式**。如需使用翻译功能,请<a href='/login?redirect=/'>登录</a>或提供 API 密钥。", gr.update(visible=True, open=True)
def format_tags_html(tags_dict, translations_list, show_scores):
if not tags_dict: return "<p>暂无标签</p>"
html = '<div class="label-container">'
for i, (tag, score) in enumerate(tags_dict.items()):
escaped_tag = tag.replace("'", "\\'")
html += '<div class="tag-item">'
tag_display_html = f'<span class="tag-en" onclick="copyToClipboard(\'{escaped_tag}\')">{tag}</span>'
if i < len(translations_list) and translations_list[i]:
tag_display_html += f'<span class="tag-zh">({translations_list[i]})</span>'
html += f'<div>{tag_display_html}</div>'
if show_scores: html += f'<span class="tag-score">{score:.3f}</span>'
html += '</div>'
return html + '</div>'
def generate_summary_text_content(current_res, translations, sum_cats, sep_type, show_zh):
if not current_res: return "请先分析图像。"
parts, sep = [], {"逗号": ", ", "换行": "\n", "空格": " "}.get(sep_type, ", ")
cat_map = {"通用标签": "general", "角色标签": "characters", "评分标签": "ratings"}
for cat_name in sum_cats:
cat_key = cat_map.get(cat_name)
if cat_key and current_res.get(cat_key):
tags_en, trans = list(current_res[cat_key].keys()), translations.get(cat_key, [])
tags_to_join = [f"{en}({zh})" if show_zh and i < len(trans) and trans[i] else en for i, en in enumerate(tags_en)]
if tags_to_join: parts.append(sep.join(tags_to_join))
return "\n".join(parts) if parts else "选定的类别中没有找到标签。"
def process_image_and_generate_outputs(
img, g_th, c_th, s_scores,
user_tencent_id, user_tencent_key, user_baidu_json,
sum_cats, s_sep, s_zh_in_sum,
request: gr.Request
):
if img is None:
raise gr.Error("请先上传图片。")
if tagger_instance is None:
raise gr.Error("分析器未成功初始化,请检查后台错误。")
yield gr.update(interactive=False, value="🔄 处理中..."), gr.update(visible=True, value="🔄 正在分析..."), *["<p>分析中...</p>"]*3, "分析中...", {}, {}
token = get_token_from_request(request)
is_owner = False
if token:
try:
user_info = whoami(token=token)
if is_user_space_owner(user_info):
is_owner = True
except Exception: pass
final_tencent_id, final_tencent_key, baidu_json_str = (
(os.environ.get("TENCENT_SECRET_ID"), os.environ.get("TENCENT_SECRET_KEY"), os.environ.get("BAIDU_CREDENTIALS_JSON", "[]"))
if is_owner else (user_tencent_id, user_tencent_key, user_baidu_json)
)
final_baidu_creds_list = []
if baidu_json_str and baidu_json_str.strip():
try:
parsed_data = json.loads(baidu_json_str)
if isinstance(parsed_data, list): final_baidu_creds_list = parsed_data
except json.JSONDecodeError: print("提供的百度凭证JSON无效。")
try:
res, tag_cats_original = tagger_instance.predict(img, g_th, c_th)
all_tags = [tag for cat in tag_cats_original.values() for tag in cat]
translations_flat = translate_texts(
all_tags,
tencent_secret_id=final_tencent_id,
tencent_secret_key=final_tencent_key,
baidu_credentials_list=final_baidu_creds_list
) if all_tags else []
translations, offset = {}, 0
for cat_key, tags in tag_cats_original.items():
translations[cat_key] = translations_flat[offset : offset + len(tags)]
offset += len(tags)
outputs_html = {k: format_tags_html(res.get(k, {}), translations.get(k, []), s_scores) for k in ["general", "characters", "ratings"]}
summary = generate_summary_text_content(res, translations, sum_cats, s_sep, s_zh_in_sum)
yield gr.update(interactive=True, value="🚀 开始分析"), gr.update(visible=True, value="✅ 分析完成!"), outputs_html["general"], outputs_html["characters"], outputs_html["ratings"], summary, res, translations
except Exception as e:
import traceback
traceback.print_exc()
raise gr.Error(f"处理时发生错误: {e}")
demo.load(fn=check_user_status, inputs=None, outputs=[user_status_md, api_key_accordion], queue=False)
btn.click(
process_image_and_generate_outputs,
inputs=[
img_in, gen_slider, char_slider, show_tag_scores,
tencent_id_in, tencent_key_in, baidu_json_in,
sum_cats, sum_sep, sum_show_zh
],
outputs=[
btn, processing_info,
out_general, out_char, out_rating,
out_summary,
state_res, state_translations_dict
],
)
summary_controls = [sum_cats, sum_sep, sum_show_zh]
for ctrl in summary_controls:
ctrl.change(
fn=lambda r, t, c, s, z: generate_summary_text_content(r, t, c, s, z),
inputs=[state_res, state_translations_dict] + summary_controls,
outputs=[out_summary],
)
if __name__ == "__main__":
if tagger_instance is None:
print("CRITICAL: Tagger failed to initialize. App functionality will be limited.")
demo.launch(server_name="0.0.0.0", server_port=7860)