IdlecloudX commited on
Commit
3553faa
·
verified ·
1 Parent(s): 0535e28

Update app.py

Browse files
Files changed (1) hide show
  1. app.py +83 -247
app.py CHANGED
@@ -7,25 +7,26 @@ import onnxruntime as rt
7
  import pandas as pd
8
  from PIL import Image
9
  from huggingface_hub import login
10
-
11
- from translator import translate_texts
12
 
13
  # ------------------------------------------------------------------
14
- # 模型配置
15
  # ------------------------------------------------------------------
16
  MODEL_REPO = "SmilingWolf/wd-eva02-large-tagger-v3"
17
  MODEL_FILENAME = "model.onnx"
18
  LABEL_FILENAME = "selected_tags.csv"
 
19
 
20
  HF_TOKEN = os.environ.get("HF_TOKEN", "")
21
  if HF_TOKEN:
22
- login(token=HF_TOKEN)
 
 
 
 
23
  else:
24
  print("⚠️ 未检测到 HF_TOKEN,私有模型可能下载失败")
25
 
26
- # ------------------------------------------------------------------
27
- # Tagger 类 (全局实例化)
28
- # ------------------------------------------------------------------
29
  class Tagger:
30
  def __init__(self):
31
  self.hf_token = HF_TOKEN
@@ -59,7 +60,6 @@ class Tagger:
59
  raise RuntimeError(f"模型初始化失败: {e}")
60
 
61
 
62
- # ------------------------- preprocess -------------------------
63
  def _preprocess(self, img: Image.Image) -> np.ndarray:
64
  if img is None:
65
  raise ValueError("输入图像不能为空")
@@ -70,9 +70,8 @@ class Tagger:
70
  canvas.paste(img, ((size - img.width) // 2, (size - img.height) // 2))
71
  if size != self.input_size:
72
  canvas = canvas.resize((self.input_size, self.input_size), Image.BICUBIC)
73
- return np.array(canvas)[:, :, ::-1].astype(np.float32) # to BGR
74
 
75
- # --------------------------- predict --------------------------
76
  def predict(self, img: Image.Image, gen_th: float = 0.35, char_th: float = 0.85):
77
  if self.model is None:
78
  raise RuntimeError("模型未成功加载,无法进行预测。")
@@ -111,125 +110,41 @@ class Tagger:
111
 
112
  return res, tag_categories_for_translation
113
 
114
- # 全局 Tagger 实例
115
  try:
116
  tagger_instance = Tagger()
117
  except RuntimeError as e:
118
  print(f"应用启动时Tagger初始化失败: {e}")
119
- tagger_instance = None # 允许应用启动,但在处理时会失败
120
 
121
- # ------------------------------------------------------------------
122
- # Gradio UI
123
- # ------------------------------------------------------------------
124
  custom_css = """
125
- .label-container {
126
- max-height: 300px;
127
- overflow-y: auto;
128
- border: 1px solid #ddd;
129
- padding: 10px;
130
- border-radius: 5px;
131
- background-color: #f9f9f9;
132
- }
133
- .tag-item {
134
- display: flex;
135
- justify-content: space-between;
136
- align-items: center;
137
- margin: 2px 0;
138
- padding: 2px 5px;
139
- border-radius: 3px;
140
- background-color: #fff;
141
- transition: background-color 0.2s;
142
- }
143
- .tag-item:hover {
144
- background-color: #f0f0f0;
145
- }
146
- .tag-en {
147
- font-weight: bold;
148
- color: #333;
149
- cursor: pointer; /* Indicates clickable */
150
- }
151
- .tag-zh {
152
- color: #666;
153
- margin-left: 10px;
154
- }
155
- .tag-score {
156
- color: #999;
157
- font-size: 0.9em;
158
- }
159
- .btn-analyze-container { /* Custom class for analyze button container */
160
- margin-top: 15px;
161
- margin-bottom: 15px;
162
- }
163
  """
164
 
165
  _js_functions = """
166
  function copyToClipboard(text) {
167
- // --- 调试信息 ---
168
- console.log('copyToClipboard function was called.');
169
- console.log('Received text:', text);
170
- // console.trace(); // 如果需要更详细的调用栈信息,可以取消这行注释
171
-
172
- // --- 保护性检查 ---
173
- // 如果 text 未定义或为 null,则不执行后续操作,并打印警告
174
  if (typeof text === 'undefined' || text === null) {
175
- console.warn('copyToClipboard was called with undefined or null text. Aborting this specific copy operation.');
176
- // 在这种情况下,我们不应该尝试复制,也不应该显示“已复制”的提示
177
  return;
178
  }
179
-
180
  navigator.clipboard.writeText(text).then(() => {
181
- // console.log('Tag copied to clipboard: ' + text); // 成功复制的日志(可选)
182
  const feedback = document.createElement('div');
183
-
184
- // 确保 text 是���符串类型,再进行 substring 操作
185
- let displayText = String(text); // 将 text 转换为字符串以防万一
186
  displayText = displayText.substring(0, 30) + (displayText.length > 30 ? '...' : '');
187
-
188
  feedback.textContent = '已复制: ' + displayText;
189
- feedback.style.position = 'fixed';
190
- feedback.style.bottom = '20px';
191
- feedback.style.left = '50%';
192
- feedback.style.transform = 'translateX(-50%)';
193
- feedback.style.backgroundColor = '#4CAF50';
194
- feedback.style.color = 'white';
195
- feedback.style.padding = '10px 20px';
196
- feedback.style.borderRadius = '5px';
197
- feedback.style.zIndex = '10000';
198
- feedback.style.transition = 'opacity 0.5s ease-out';
199
  document.body.appendChild(feedback);
200
  setTimeout(() => {
201
  feedback.style.opacity = '0';
202
- setTimeout(() => {
203
- if (document.body.contains(feedback)) { // 确保元素还在DOM中
204
- document.body.removeChild(feedback);
205
- }
206
- }, 500);
207
  }, 1500);
208
  }).catch(err => {
209
  console.error('Failed to copy tag. Error:', err, 'Attempted to copy text:', text);
210
- // 可以考虑也给用户一个错误提示,但原版 alert 可能体验不佳
211
- // alert('复制失败: ' + err);
212
- const errorFeedback = document.createElement('div');
213
- errorFeedback.textContent = '复制操作失败!'; // 更友好的错误提示
214
- errorFeedback.style.position = 'fixed';
215
- errorFeedback.style.bottom = '20px';
216
- errorFeedback.style.left = '50%';
217
- errorFeedback.style.transform = 'translateX(-50%)';
218
- errorFeedback.style.backgroundColor = '#D32F2F'; // 红色背景表示错误
219
- errorFeedback.style.color = 'white';
220
- errorFeedback.style.padding = '10px 20px';
221
- errorFeedback.style.borderRadius = '5px';
222
- errorFeedback.style.zIndex = '10000';
223
- errorFeedback.style.transition = 'opacity 0.5s ease-out';
224
- document.body.appendChild(errorFeedback);
225
- setTimeout(() => {
226
- errorFeedback.style.opacity = '0';
227
- setTimeout(() => {
228
- if (document.body.contains(errorFeedback)) {
229
- document.body.removeChild(errorFeedback);
230
- }
231
- }, 500);
232
- }, 2500);
233
  });
234
  }
235
  """
@@ -246,13 +161,22 @@ with gr.Blocks(theme=gr.themes.Soft(), title="AI 图像标签分析器", css=cus
246
  with gr.Row():
247
  with gr.Column(scale=1):
248
  img_in = gr.Image(type="pil", label="上传图片", height=300)
249
-
250
  btn = gr.Button("🚀 开始分析", variant="primary", elem_classes=["btn-analyze-container"])
251
 
252
  with gr.Accordion("⚙️ 高级设置", open=False):
253
  gen_slider = gr.Slider(0, 1, value=0.35, step=0.01, label="通用标签阈值", info="越高 → 标签更少更准")
254
  char_slider = gr.Slider(0, 1, value=0.85, step=0.01, label="角色标签阈值", info="推荐保持较高阈值")
255
  show_tag_scores = gr.Checkbox(True, label="在列表中显示标签置信度")
 
 
 
 
 
 
 
 
 
 
256
 
257
  with gr.Accordion("📊 标签汇总设置", open=True):
258
  gr.Markdown("选择要包含在下方汇总文本框中的标签类别:")
@@ -283,209 +207,121 @@ with gr.Blocks(theme=gr.themes.Soft(), title="AI 图像标签分析器", css=cus
283
  show_copy_button=True
284
  )
285
 
286
- # ----------------- 辅助函数 -----------------
287
  def format_tags_html(tags_dict, translations_list, category_name, show_scores=True, show_translation_in_list=True):
288
- if not tags_dict:
289
- return "<p>暂无标签</p>"
290
-
291
  html = '<div class="label-container">'
292
-
293
- if not isinstance(translations_list, list):
294
- translations_list = []
295
-
296
  tag_keys = list(tags_dict.keys())
297
-
298
  for i, tag in enumerate(tag_keys):
299
  score = tags_dict[tag]
300
- escaped_tag = tag.replace("'", "\\'") # Escape for JS
301
-
302
  html += '<div class="tag-item">'
303
  tag_display_html = f'<span class="tag-en" onclick="copyToClipboard(\'{escaped_tag}\')">{tag}</span>'
304
-
305
  if show_translation_in_list and i < len(translations_list) and translations_list[i]:
306
  tag_display_html += f'<span class="tag-zh">({translations_list[i]})</span>'
307
-
308
  html += f'<div>{tag_display_html}</div>'
309
- if show_scores:
310
- html += f'<span class="tag-score">{score:.3f}</span>'
311
  html += '</div>'
312
  html += '</div>'
313
  return html
314
 
315
- def generate_summary_text_content(
316
- current_res, current_translations_dict,
317
- s_gen, s_char, s_rat, s_sep_type, s_show_zh
318
- ):
319
- if not current_res:
320
- return "请先分析图像或选择要汇总的标签类别。"
321
-
322
- summary_parts = []
323
- separators = {"逗号": ", ", "换行": "\n", "空格": " "}
324
- separator = separators.get(s_sep_type, ", ")
325
-
326
  categories_to_summarize = []
327
  if s_gen: categories_to_summarize.append("general")
328
  if s_char: categories_to_summarize.append("characters")
329
  if s_rat: categories_to_summarize.append("ratings")
330
-
331
- if not categories_to_summarize:
332
- return "请至少选择一个标签类别进行汇总。"
333
 
334
  for cat_key in categories_to_summarize:
335
  if current_res.get(cat_key):
336
  tags_to_join = []
337
  cat_tags_en = list(current_res[cat_key].keys())
338
  cat_translations = current_translations_dict.get(cat_key, [])
339
-
340
  for i, en_tag in enumerate(cat_tags_en):
341
  if s_show_zh and i < len(cat_translations) and cat_translations[i]:
342
  tags_to_join.append(f"{en_tag}({cat_translations[i]})")
343
  else:
344
  tags_to_join.append(en_tag)
345
- if tags_to_join: # only add if there are tags for this category
346
- summary_parts.append(separator.join(tags_to_join))
347
-
348
- # Join parts with double newline for readability if multiple categories present and separator is not newline
349
- joiner = "\n\n" if separator != "\n" and len(summary_parts) > 1 else separator if separator == "\n" else " "
350
-
351
  final_summary = joiner.join(summary_parts)
352
  return final_summary if final_summary else "选定的类别中没有找到标签。"
353
 
354
-
355
- # ----------------- 主要处理回调 -----------------
356
- def process_image_and_generate_outputs(
357
- img, g_th, c_th, s_scores, # Main inputs
358
- s_gen, s_char, s_rat, s_sep, s_zh_in_sum
359
- ):
360
  if img is None:
361
- yield (
362
- gr.update(interactive=True, value="🚀 开始分析"),
363
- gr.update(visible=True, value="❌ 请先上传图片。"),
364
- "", "", "", "",
365
- gr.update(placeholder="请先上传图片并开始分析..."),
366
- {}, {}, {}
367
- )
368
  return
369
-
370
  if tagger_instance is None:
371
- yield (
372
- gr.update(interactive=True, value="🚀 开始分析"),
373
- gr.update(visible=True, value="❌ 分析器未成功初始化,请检查控制台错误。"),
374
- "", "", "", "",
375
- gr.update(placeholder="分析器初始化失败..."),
376
- {}, {}, {}
377
- )
378
  return
379
 
380
- yield (
381
- gr.update(interactive=False, value="🔄 处理中..."),
382
- gr.update(visible=True, value="🔄 正在分析图像,请稍候..."),
383
- gr.HTML(value="<p>分析中...</p>"), # General
384
- gr.HTML(value="<p>分析中...</p>"), # Character
385
- gr.HTML(value="<p>分析中...</p>"), # Rating
386
- gr.update(value="分析中,请稍候..."), # Summary
387
- {}, {}, {} # Clear states initially
388
- )
389
 
390
  try:
391
- # 1. Predict tags
392
  res, tag_categories_original_order = tagger_instance.predict(img, g_th, c_th)
393
-
394
- all_tags_to_translate = []
395
- for cat_key in ["general", "characters", "ratings"]:
396
- all_tags_to_translate.extend(tag_categories_original_order.get(cat_key, []))
397
 
398
  all_translations_flat = []
399
  if all_tags_to_translate:
400
- all_translations_flat = translate_texts(all_tags_to_translate, src_lang="auto", tgt_lang="zh")
401
-
 
 
 
 
 
 
 
 
 
 
402
  current_translations_dict = {}
403
  offset = 0
404
  for cat_key in ["general", "characters", "ratings"]:
405
- cat_original_tags = tag_categories_original_order.get(cat_key, [])
406
- num_tags_in_cat = len(cat_original_tags)
407
- if num_tags_in_cat > 0:
408
- current_translations_dict[cat_key] = all_translations_flat[offset : offset + num_tags_in_cat]
409
- offset += num_tags_in_cat
410
- else:
411
- current_translations_dict[cat_key] = []
412
 
413
- general_html = format_tags_html(res.get("general", {}), current_translations_dict.get("general", []), "general", s_scores, True)
414
- char_html = format_tags_html(res.get("characters", {}), current_translations_dict.get("characters", []), "characters", s_scores, True)
415
- rating_html = format_tags_html(res.get("ratings", {}), current_translations_dict.get("ratings", []), "ratings", s_scores, True)
416
-
417
- summary_text = generate_summary_text_content(
418
- res, current_translations_dict,
419
- s_gen, s_char, s_rat, s_sep, s_zh_in_sum
420
- )
421
 
422
- yield (
423
- gr.update(interactive=True, value="🚀 开始分析"),
424
- gr.update(visible=True, value="✅ 分析完成!"),
425
- general_html,
426
- char_html,
427
- rating_html,
428
- gr.update(value=summary_text),
429
- res,
430
- current_translations_dict,
431
- tag_categories_original_order
432
- )
433
 
434
  except Exception as e:
435
  import traceback
436
  tb_str = traceback.format_exc()
437
  print(f"处理时发生��误: {e}\n{tb_str}")
438
- yield (
439
- gr.update(interactive=True, value="🚀 开始分析"),
440
- gr.update(visible=True, value=f"❌ 处理失败: {str(e)}"),
441
- "<p>处理出错</p>", "<p>处理出错</p>", "<p>处理出错</p>",
442
- gr.update(value=f"错误: {str(e)}", placeholder="分析失败..."),
443
- {}, {}, {}
444
- )
445
 
446
- # ----------------- 更新汇总文本的回调 -----------------
447
- def update_summary_display(
448
- s_gen, s_char, s_rat, s_sep, s_zh_in_sum,
449
- current_res_from_state, current_translations_from_state
450
- ):
451
- if not current_res_from_state:
452
- return gr.update(placeholder="请先完成一次图像分析以生成汇总。", value="")
453
-
454
- new_summary_text = generate_summary_text_content(
455
- current_res_from_state, current_translations_from_state,
456
- s_gen, s_char, s_rat, s_sep, s_zh_in_sum
457
- )
458
  return gr.update(value=new_summary_text)
459
 
460
- # ----------------- 绑定事件 -----------------
 
 
 
 
 
 
 
 
 
461
  btn.click(
462
  process_image_and_generate_outputs,
463
- inputs=[
464
- img_in, gen_slider, char_slider, show_tag_scores,
465
- sum_general, sum_char, sum_rating, sum_sep, sum_show_zh
466
- ],
467
- outputs=[
468
- btn, processing_info,
469
- out_general, out_char, out_rating,
470
- out_summary,
471
- state_res, state_translations_dict, state_tag_categories_for_translation
472
- ],
473
- # show_progress="full" # Gradio's built-in progress
474
  )
475
 
476
  summary_controls = [sum_general, sum_char, sum_rating, sum_sep, sum_show_zh]
477
  for ctrl in summary_controls:
478
- ctrl.change(
479
- fn=update_summary_display,
480
- inputs=summary_controls + [state_res, state_translations_dict],
481
- outputs=[out_summary],
482
- # show_progress=False # Typically fast, no need for progress indicator
483
- )
484
 
485
- # ------------------------------------------------------------------
486
- # 启动
487
- # ------------------------------------------------------------------
488
  if __name__ == "__main__":
489
  if tagger_instance is None:
490
  print("CRITICAL: Tagger 未能初始化,应用功能将受限。请检查之前的错误信息。")
491
- demo.launch(server_name="0.0.0.0", server_port=7860)
 
7
  import pandas as pd
8
  from PIL import Image
9
  from huggingface_hub import login
10
+ from translator import translate_texts, translate_texts_with_dynamic_keys
 
11
 
12
  # ------------------------------------------------------------------
13
+ # 模型与认证配置
14
  # ------------------------------------------------------------------
15
  MODEL_REPO = "SmilingWolf/wd-eva02-large-tagger-v3"
16
  MODEL_FILENAME = "model.onnx"
17
  LABEL_FILENAME = "selected_tags.csv"
18
+ OWNER_USERNAME = os.environ.get("OWNER_USERNAME", "")
19
 
20
  HF_TOKEN = os.environ.get("HF_TOKEN", "")
21
  if HF_TOKEN:
22
+ try:
23
+ login(token=HF_TOKEN)
24
+ print("✅ HF_TOKEN 登录成功")
25
+ except Exception as e:
26
+ print(f"⚠️ HF_TOKEN 登录失败: {e}")
27
  else:
28
  print("⚠️ 未检测到 HF_TOKEN,私有模型可能下载失败")
29
 
 
 
 
30
  class Tagger:
31
  def __init__(self):
32
  self.hf_token = HF_TOKEN
 
60
  raise RuntimeError(f"模型初始化失败: {e}")
61
 
62
 
 
63
  def _preprocess(self, img: Image.Image) -> np.ndarray:
64
  if img is None:
65
  raise ValueError("输入图像不能为空")
 
70
  canvas.paste(img, ((size - img.width) // 2, (size - img.height) // 2))
71
  if size != self.input_size:
72
  canvas = canvas.resize((self.input_size, self.input_size), Image.BICUBIC)
73
+ return np.array(canvas)[:, :, ::-1].astype(np.float32)
74
 
 
75
  def predict(self, img: Image.Image, gen_th: float = 0.35, char_th: float = 0.85):
76
  if self.model is None:
77
  raise RuntimeError("模型未成功加载,无法进行预测。")
 
110
 
111
  return res, tag_categories_for_translation
112
 
 
113
  try:
114
  tagger_instance = Tagger()
115
  except RuntimeError as e:
116
  print(f"应用启动时Tagger初始化失败: {e}")
117
+ tagger_instance = None
118
 
 
 
 
119
  custom_css = """
120
+ .label-container { max-height: 300px; overflow-y: auto; border: 1px solid #ddd; padding: 10px; border-radius: 5px; background-color: #f9f9f9; }
121
+ .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; }
122
+ .tag-item:hover { background-color: #f0f0f0; }
123
+ .tag-en { font-weight: bold; color: #333; cursor: pointer; }
124
+ .tag-zh { color: #666; margin-left: 10px; }
125
+ .tag-score { color: #999; font-size: 0.9em; }
126
+ .btn-analyze-container { margin-top: 15px; margin-bottom: 15px; }
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
127
  """
128
 
129
  _js_functions = """
130
  function copyToClipboard(text) {
 
 
 
 
 
 
 
131
  if (typeof text === 'undefined' || text === null) {
132
+ console.warn('copyToClipboard was called with undefined or null text.');
 
133
  return;
134
  }
 
135
  navigator.clipboard.writeText(text).then(() => {
 
136
  const feedback = document.createElement('div');
137
+ let displayText = String(text);
 
 
138
  displayText = displayText.substring(0, 30) + (displayText.length > 30 ? '...' : '');
 
139
  feedback.textContent = '已复制: ' + displayText;
140
+ feedback.style.cssText = 'position: fixed; bottom: 20px; left: 50%; transform: translateX(-50%); background-color: #4CAF50; color: white; padding: 10px 20px; border-radius: 5px; z-index: 10000; transition: opacity 0.5s ease-out;';
 
 
 
 
 
 
 
 
 
141
  document.body.appendChild(feedback);
142
  setTimeout(() => {
143
  feedback.style.opacity = '0';
144
+ setTimeout(() => { if (document.body.contains(feedback)) document.body.removeChild(feedback); }, 500);
 
 
 
 
145
  }, 1500);
146
  }).catch(err => {
147
  console.error('Failed to copy tag. Error:', err, 'Attempted to copy text:', text);
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
148
  });
149
  }
150
  """
 
161
  with gr.Row():
162
  with gr.Column(scale=1):
163
  img_in = gr.Image(type="pil", label="上传图片", height=300)
 
164
  btn = gr.Button("🚀 开始分析", variant="primary", elem_classes=["btn-analyze-container"])
165
 
166
  with gr.Accordion("⚙️ 高级设置", open=False):
167
  gen_slider = gr.Slider(0, 1, value=0.35, step=0.01, label="通用标签阈值", info="越高 → 标签更少更准")
168
  char_slider = gr.Slider(0, 1, value=0.85, step=0.01, label="角色标签阈值", info="推荐保持较高阈值")
169
  show_tag_scores = gr.Checkbox(True, label="在列表中显示标签置信度")
170
+
171
+ with gr.Group(visible=False) as guest_api_group:
172
+ gr.Markdown("### 访客翻译API配置\n由于您不是本空间所有者,需要提供自己的翻译API密钥才能使用翻译功能。")
173
+ guest_tencent_id = gr.Textbox(label="腾讯云 Secret ID", type="password")
174
+ guest_tencent_key = gr.Textbox(label="腾讯云 Secret Key", type="password")
175
+ guest_baidu_json = gr.TextArea(
176
+ label="百度翻译凭证 JSON",
177
+ placeholder='[{"app_id": "...", "secret_key": "..."}, ...]',
178
+ lines=3
179
+ )
180
 
181
  with gr.Accordion("📊 标签汇总设置", open=True):
182
  gr.Markdown("选择要包含在下方汇总文本框中的标签类别:")
 
207
  show_copy_button=True
208
  )
209
 
 
210
  def format_tags_html(tags_dict, translations_list, category_name, show_scores=True, show_translation_in_list=True):
211
+ if not tags_dict: return "<p>暂无标签</p>"
 
 
212
  html = '<div class="label-container">'
 
 
 
 
213
  tag_keys = list(tags_dict.keys())
 
214
  for i, tag in enumerate(tag_keys):
215
  score = tags_dict[tag]
216
+ escaped_tag = tag.replace("'", "\\'")
 
217
  html += '<div class="tag-item">'
218
  tag_display_html = f'<span class="tag-en" onclick="copyToClipboard(\'{escaped_tag}\')">{tag}</span>'
 
219
  if show_translation_in_list and i < len(translations_list) and translations_list[i]:
220
  tag_display_html += f'<span class="tag-zh">({translations_list[i]})</span>'
 
221
  html += f'<div>{tag_display_html}</div>'
222
+ if show_scores: html += f'<span class="tag-score">{score:.3f}</span>'
 
223
  html += '</div>'
224
  html += '</div>'
225
  return html
226
 
227
+ def generate_summary_text_content(current_res, current_translations_dict, s_gen, s_char, s_rat, s_sep_type, s_show_zh):
228
+ if not current_res: return "请先分析图像或选择要汇总的标签类别。"
229
+ summary_parts, separator = [], {"逗号": ", ", "换行": "\n", "空格": " "}.get(s_sep_type, ", ")
 
 
 
 
 
 
 
 
230
  categories_to_summarize = []
231
  if s_gen: categories_to_summarize.append("general")
232
  if s_char: categories_to_summarize.append("characters")
233
  if s_rat: categories_to_summarize.append("ratings")
234
+ if not categories_to_summarize: return "请至少选择一个标签类别进行汇总。"
 
 
235
 
236
  for cat_key in categories_to_summarize:
237
  if current_res.get(cat_key):
238
  tags_to_join = []
239
  cat_tags_en = list(current_res[cat_key].keys())
240
  cat_translations = current_translations_dict.get(cat_key, [])
 
241
  for i, en_tag in enumerate(cat_tags_en):
242
  if s_show_zh and i < len(cat_translations) and cat_translations[i]:
243
  tags_to_join.append(f"{en_tag}({cat_translations[i]})")
244
  else:
245
  tags_to_join.append(en_tag)
246
+ if tags_to_join: summary_parts.append(separator.join(tags_to_join))
247
+ joiner = "\n\n" if separator != "\n" and len(summary_parts) > 1 else separator
 
 
 
 
248
  final_summary = joiner.join(summary_parts)
249
  return final_summary if final_summary else "选定的类别中没有找到标签。"
250
 
251
+ def process_image_and_generate_outputs(img, g_th, c_th, s_scores, s_gen, s_char, s_rat, s_sep, s_zh_in_sum, guest_tc_id, guest_tc_key, guest_bd_json, request: gr.Request):
 
 
 
 
 
252
  if img is None:
253
+ yield (gr.update(interactive=True, value="🚀 开始分析"), gr.update(visible=True, value="❌ 请先上传图片。"), "", "", "", "", gr.update(placeholder="请先上传图片并开始分析..."), {}, {}, {})
 
 
 
 
 
 
254
  return
 
255
  if tagger_instance is None:
256
+ yield (gr.update(interactive=True, value="🚀 开始分析"), gr.update(visible=True, value="❌ 分析器未成功初始化,请检查控制台错误。"), "", "", "", "", gr.update(placeholder="分析器初始化失败..."), {}, {}, {})
 
 
 
 
 
 
257
  return
258
 
259
+ yield (gr.update(interactive=False, value="🔄 处理中..."), gr.update(visible=True, value="🔄 正在分析图像..."), gr.HTML(value="<p>分析中...</p>"), gr.HTML(value="<p>分析中...</p>"), gr.HTML(value="<p>分析中...</p>"), gr.update(value="分析中..."), {}, {}, {})
 
 
 
 
 
 
 
 
260
 
261
  try:
 
262
  res, tag_categories_original_order = tagger_instance.predict(img, g_th, c_th)
263
+ all_tags_to_translate = [tag for cat_key in ["general", "characters", "ratings"] for tag in tag_categories_original_order.get(cat_key, [])]
 
 
 
264
 
265
  all_translations_flat = []
266
  if all_tags_to_translate:
267
+ is_owner = request.username and request.username == OWNER_USERNAME
268
+ if is_owner:
269
+ print("- [Auth] 所有者身份,使用预置密钥进行翻译。")
270
+ all_translations_flat = translate_texts(all_tags_to_translate)
271
+ else:
272
+ print("- [Auth] 访客身份,使用用户提供的密钥进行翻译。")
273
+ if not guest_tc_id and not guest_bd_json:
274
+ print(" - [Warning] 访客未提供任何API密钥,将跳过翻译。")
275
+ all_translations_flat = all_tags_to_translate
276
+ else:
277
+ all_translations_flat = translate_texts_with_dynamic_keys(all_tags_to_translate, guest_tc_id, guest_tc_key, guest_bd_json)
278
+
279
  current_translations_dict = {}
280
  offset = 0
281
  for cat_key in ["general", "characters", "ratings"]:
282
+ num_tags = len(tag_categories_original_order.get(cat_key, []))
283
+ current_translations_dict[cat_key] = all_translations_flat[offset : offset + num_tags]
284
+ offset += num_tags
 
 
 
 
285
 
286
+ general_html = format_tags_html(res.get("general", {}), current_translations_dict.get("general", []), "general", s_scores)
287
+ char_html = format_tags_html(res.get("characters", {}), current_translations_dict.get("characters", []), "characters", s_scores)
288
+ rating_html = format_tags_html(res.get("ratings", {}), current_translations_dict.get("ratings", []), "ratings", s_scores)
289
+ summary_text = generate_summary_text_content(res, current_translations_dict, s_gen, s_char, s_rat, s_sep, s_zh_in_sum)
 
 
 
 
290
 
291
+ yield (gr.update(interactive=True, value="🚀 开始分析"), gr.update(visible=True, value="✅ 分析完成!"), general_html, char_html, rating_html, gr.update(value=summary_text), res, current_translations_dict, tag_categories_original_order)
 
 
 
 
 
 
 
 
 
 
292
 
293
  except Exception as e:
294
  import traceback
295
  tb_str = traceback.format_exc()
296
  print(f"处理时发生��误: {e}\n{tb_str}")
297
+ yield (gr.update(interactive=True, value="🚀 开始分析"), gr.update(visible=True, value=f"❌ 处理失败: {str(e)}"), "<p>处理出错</p>", "<p>处理出错</p>", "<p>处理出错</p>", gr.update(value=f"错误: {str(e)}", placeholder="分析失败..."), {}, {}, {})
 
 
 
 
 
 
298
 
299
+ def update_summary_display(s_gen, s_char, s_rat, s_sep, s_zh_in_sum, current_res, current_translations):
300
+ if not current_res: return gr.update(placeholder="请先完成一次图像分析以生成汇总。", value="")
301
+ new_summary_text = generate_summary_text_content(current_res, current_translations, s_gen, s_char, s_rat, s_sep, s_zh_in_sum)
 
 
 
 
 
 
 
 
 
302
  return gr.update(value=new_summary_text)
303
 
304
+ def check_user_auth(request: gr.Request):
305
+ if not OWNER_USERNAME: print("⚠️ 警告: 未设置 OWNER_USERNAME 环境变量。所有用户都将被视为访客。")
306
+ if not request.username or request.username != OWNER_USERNAME:
307
+ print(f"- [Auth] 访客 '{request.username}' 已连接,显示 API Key 输入框。")
308
+ return gr.update(visible=True)
309
+ print(f"- [Auth] 所有者 '{request.username}' 已连接。")
310
+ return gr.update(visible=False)
311
+
312
+ demo.load(fn=check_user_auth, inputs=None, outputs=[guest_api_group])
313
+
314
  btn.click(
315
  process_image_and_generate_outputs,
316
+ inputs=[img_in, gen_slider, char_slider, show_tag_scores, sum_general, sum_char, sum_rating, sum_sep, sum_show_zh, guest_tencent_id, guest_tencent_key, guest_baidu_json],
317
+ outputs=[btn, processing_info, out_general, out_char, out_rating, out_summary, state_res, state_translations_dict, state_tag_categories_for_translation]
 
 
 
 
 
 
 
 
 
318
  )
319
 
320
  summary_controls = [sum_general, sum_char, sum_rating, sum_sep, sum_show_zh]
321
  for ctrl in summary_controls:
322
+ ctrl.change(fn=update_summary_display, inputs=summary_controls + [state_res, state_translations_dict], outputs=[out_summary])
 
 
 
 
 
323
 
 
 
 
324
  if __name__ == "__main__":
325
  if tagger_instance is None:
326
  print("CRITICAL: Tagger 未能初始化,应用功能将受限。请检查之前的错误信息。")
327
+ demo.launch(server_name="0.0.0.0", server_port=7860)