Spaces:
Sleeping
Sleeping
Update app.py
Browse files
app.py
CHANGED
|
@@ -4,357 +4,382 @@ from PIL import Image
|
|
| 4 |
import easyocr
|
| 5 |
import numpy as np
|
| 6 |
import re
|
| 7 |
-
from datetime import datetime
|
| 8 |
import traceback
|
| 9 |
|
| 10 |
-
class
|
| 11 |
def __init__(self):
|
| 12 |
self.reader = None
|
|
|
|
| 13 |
self.initialize_reader()
|
| 14 |
|
| 15 |
def initialize_reader(self):
|
| 16 |
"""安全初始化EasyOCR"""
|
| 17 |
-
|
| 18 |
-
|
| 19 |
-
|
| 20 |
-
|
| 21 |
-
|
| 22 |
-
|
|
|
|
|
|
|
|
|
|
|
|
|
| 23 |
try:
|
| 24 |
-
|
| 25 |
-
self.reader = easyocr.Reader(
|
| 26 |
-
self.lang_config =
|
| 27 |
-
print("成功初始化:
|
| 28 |
-
|
| 29 |
-
|
| 30 |
-
|
| 31 |
-
|
| 32 |
-
|
| 33 |
-
|
| 34 |
-
|
| 35 |
-
|
| 36 |
-
self.reader = None
|
| 37 |
-
self.lang_config = "初始化失敗"
|
| 38 |
|
| 39 |
-
def
|
| 40 |
-
"""
|
| 41 |
if self.reader is None:
|
| 42 |
-
return self.
|
|
|
|
|
|
|
|
|
|
| 43 |
|
| 44 |
try:
|
| 45 |
-
#
|
| 46 |
if isinstance(image, np.ndarray):
|
| 47 |
-
|
| 48 |
-
|
| 49 |
-
|
| 50 |
|
| 51 |
-
|
| 52 |
-
img_array = np.array(image)
|
| 53 |
|
| 54 |
# 執行OCR
|
| 55 |
-
print("開始執行OCR...")
|
| 56 |
results = self.reader.readtext(img_array)
|
| 57 |
-
print(f"OCR完成,識別到 {len(results)} 個文字區塊")
|
| 58 |
|
| 59 |
-
|
| 60 |
-
processed_data = []
|
| 61 |
-
for i, (bbox, text, confidence) in enumerate(results):
|
| 62 |
-
try:
|
| 63 |
-
top_left = bbox[0]
|
| 64 |
-
bottom_right = bbox[2]
|
| 65 |
-
center_y = (top_left[1] + bottom_right[1]) / 2
|
| 66 |
-
center_x = (top_left[0] + bottom_right[0]) / 2
|
| 67 |
-
|
| 68 |
-
processed_data.append({
|
| 69 |
-
'序號': i + 1,
|
| 70 |
-
'識別文字': text.strip(),
|
| 71 |
-
'信心度': round(float(confidence), 3),
|
| 72 |
-
'中心X': int(center_x),
|
| 73 |
-
'中心Y': int(center_y),
|
| 74 |
-
'左上角X': int(top_left[0]),
|
| 75 |
-
'左上角Y': int(top_left[1]),
|
| 76 |
-
'右下角X': int(bottom_right[0]),
|
| 77 |
-
'右下角Y': int(bottom_right[1]),
|
| 78 |
-
'文字類型': self.classify_text(text.strip())
|
| 79 |
-
})
|
| 80 |
-
except Exception as e:
|
| 81 |
-
print(f"處理第{i+1}個結果時出錯:{e}")
|
| 82 |
-
continue
|
| 83 |
|
| 84 |
-
|
| 85 |
-
|
| 86 |
-
|
| 87 |
-
|
| 88 |
-
|
| 89 |
-
|
| 90 |
-
#
|
| 91 |
-
|
|
|
|
| 92 |
|
| 93 |
-
|
| 94 |
-
|
| 95 |
-
|
| 96 |
-
|
| 97 |
-
|
| 98 |
-
|
| 99 |
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 100 |
except Exception as e:
|
| 101 |
-
error_msg = f"處理圖片時
|
| 102 |
print(error_msg)
|
| 103 |
print(traceback.format_exc())
|
| 104 |
-
return self.
|
| 105 |
|
| 106 |
-
def
|
| 107 |
-
"""
|
| 108 |
if not text:
|
| 109 |
-
return
|
| 110 |
|
| 111 |
-
|
| 112 |
-
|
| 113 |
-
|
| 114 |
-
|
| 115 |
-
|
| 116 |
-
|
| 117 |
-
|
| 118 |
-
|
| 119 |
-
|
| 120 |
-
|
| 121 |
-
|
| 122 |
-
|
| 123 |
-
|
| 124 |
-
|
| 125 |
-
|
| 126 |
-
|
| 127 |
-
|
| 128 |
-
|
| 129 |
-
|
| 130 |
-
|
| 131 |
-
|
| 132 |
-
|
| 133 |
-
return
|
|
|
|
|
|
|
| 134 |
|
| 135 |
-
def
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 136 |
"""重建表格結構"""
|
| 137 |
if df.empty:
|
| 138 |
return pd.DataFrame()
|
| 139 |
|
| 140 |
try:
|
| 141 |
-
# 按Y座標分組
|
| 142 |
-
y_threshold = 30
|
| 143 |
rows = []
|
|
|
|
|
|
|
| 144 |
current_row = []
|
| 145 |
last_y = None
|
|
|
|
| 146 |
|
| 147 |
-
for _,
|
| 148 |
-
|
| 149 |
-
|
| 150 |
-
if last_y is None or abs(current_y - last_y) <= y_threshold:
|
| 151 |
-
current_row.append(row)
|
| 152 |
else:
|
| 153 |
if current_row:
|
|
|
|
| 154 |
current_row.sort(key=lambda x: x['中心X'])
|
| 155 |
rows.append(current_row)
|
| 156 |
-
current_row = [
|
| 157 |
-
|
| 158 |
-
last_y = current_y
|
| 159 |
|
|
|
|
| 160 |
if current_row:
|
| 161 |
current_row.sort(key=lambda x: x['中心X'])
|
| 162 |
rows.append(current_row)
|
| 163 |
|
| 164 |
-
# 構建表格
|
| 165 |
if not rows:
|
| 166 |
return pd.DataFrame()
|
| 167 |
|
| 168 |
-
|
| 169 |
table_data = []
|
|
|
|
| 170 |
|
| 171 |
-
for
|
| 172 |
-
|
| 173 |
-
|
| 174 |
-
|
|
|
|
|
|
|
| 175 |
|
| 176 |
# 填充空列
|
| 177 |
-
for
|
| 178 |
-
|
|
|
|
| 179 |
|
| 180 |
-
table_data.append(
|
| 181 |
|
| 182 |
return pd.DataFrame(table_data)
|
| 183 |
|
| 184 |
except Exception as e:
|
| 185 |
print(f"重建表格失敗:{e}")
|
| 186 |
-
return pd.DataFrame([{
|
| 187 |
|
| 188 |
-
def
|
| 189 |
-
"""創建
|
| 190 |
-
|
| 191 |
-
|
| 192 |
-
return "未識別到任何內容"
|
| 193 |
-
|
| 194 |
-
avg_confidence = df['信心度'].mean()
|
| 195 |
-
high_conf = (df['信心度'] >= 0.8).sum()
|
| 196 |
-
total_count = len(df)
|
| 197 |
-
|
| 198 |
-
summary = f"""
|
| 199 |
-
🔍 OCR識別報告
|
| 200 |
-
═══════════════════════════════════
|
| 201 |
-
📊 基本統計:
|
| 202 |
-
• 語言配置:{self.lang_config}
|
| 203 |
-
• 識別區塊:{total_count} 個
|
| 204 |
-
• 平均信心度:{avg_confidence:.3f}
|
| 205 |
-
• 高信心度區塊:{high_conf} 個 ({high_conf/total_count*100:.1f}%)
|
| 206 |
-
|
| 207 |
-
📝 文字類型分布:
|
| 208 |
-
"""
|
| 209 |
-
type_counts = df['文字類型'].value_counts()
|
| 210 |
-
for text_type, count in type_counts.items():
|
| 211 |
-
percentage = (count / total_count) * 100
|
| 212 |
-
summary += f"• {text_type}:{count} 個 ({percentage:.1f}%)\n"
|
| 213 |
-
|
| 214 |
-
# 品質評估
|
| 215 |
-
if avg_confidence >= 0.8:
|
| 216 |
-
summary += "\n✅ 識別品質:優秀"
|
| 217 |
-
elif avg_confidence >= 0.6:
|
| 218 |
-
summary += "\n⚠️ 識別品質:良好"
|
| 219 |
-
else:
|
| 220 |
-
summary += "\n❌ 識別品質:需改進"
|
| 221 |
-
|
| 222 |
-
return summary
|
| 223 |
-
|
| 224 |
-
except Exception as e:
|
| 225 |
-
return f"生成報告時發生錯誤:{str(e)}"
|
| 226 |
-
|
| 227 |
-
def create_error_response(self, error_msg):
|
| 228 |
-
"""創建錯誤響應"""
|
| 229 |
-
error_df = pd.DataFrame([{'錯誤': error_msg}])
|
| 230 |
-
return error_df, error_df, f"❌ 錯誤:{error_msg}"
|
| 231 |
|
| 232 |
-
def
|
| 233 |
-
"""創建空
|
| 234 |
empty_df = pd.DataFrame()
|
| 235 |
-
return empty_df, empty_df, f"ℹ️ {
|
| 236 |
|
| 237 |
-
def
|
| 238 |
-
"""創建
|
| 239 |
-
processor = StableOCRProcessor()
|
| 240 |
|
| 241 |
-
|
| 242 |
-
|
|
|
|
|
|
|
|
|
|
| 243 |
try:
|
| 244 |
-
|
|
|
|
| 245 |
|
| 246 |
# 應用信心度過濾
|
| 247 |
if not df.empty and '信心度' in df.columns:
|
| 248 |
-
|
| 249 |
-
|
| 250 |
-
|
| 251 |
-
|
|
|
|
|
|
|
| 252 |
|
| 253 |
return summary, df, table_df
|
| 254 |
|
| 255 |
except Exception as e:
|
| 256 |
error_msg = f"處理失敗:{str(e)}"
|
| 257 |
-
error_df = pd.DataFrame([{
|
| 258 |
return error_msg, error_df, error_df
|
| 259 |
|
| 260 |
-
# 創建界面
|
| 261 |
-
with gr.Blocks(
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 262 |
gr.HTML("""
|
| 263 |
-
<div style="text-align: center; padding: 20px; background:
|
| 264 |
-
<h1>🔍
|
| 265 |
-
<p>
|
|
|
|
|
|
|
| 266 |
</div>
|
| 267 |
""")
|
| 268 |
|
|
|
|
| 269 |
with gr.Row():
|
|
|
|
| 270 |
with gr.Column(scale=1):
|
| 271 |
image_input = gr.Image(
|
| 272 |
label="📤 上傳圖片",
|
| 273 |
-
type="pil"
|
| 274 |
-
height=400
|
| 275 |
)
|
| 276 |
|
| 277 |
-
|
| 278 |
minimum=0.0,
|
| 279 |
maximum=1.0,
|
| 280 |
value=0.3,
|
| 281 |
step=0.1,
|
| 282 |
-
label="🎯 最低信心度"
|
| 283 |
-
info="過濾掉信心度低於此值的結果"
|
| 284 |
)
|
| 285 |
|
| 286 |
-
|
| 287 |
"🚀 開始識別",
|
| 288 |
-
variant="primary"
|
| 289 |
-
size="lg"
|
| 290 |
)
|
| 291 |
|
|
|
|
| 292 |
with gr.Column(scale=1):
|
| 293 |
-
|
| 294 |
label="📊 識別報告",
|
| 295 |
-
lines=
|
| 296 |
-
max_lines=
|
| 297 |
)
|
| 298 |
|
|
|
|
| 299 |
with gr.Row():
|
| 300 |
with gr.Column():
|
| 301 |
gr.Markdown("### 📋 詳細識別結果")
|
| 302 |
-
|
| 303 |
-
|
| 304 |
-
|
| 305 |
-
height=300
|
| 306 |
)
|
| 307 |
|
| 308 |
with gr.Column():
|
| 309 |
gr.Markdown("### 🔄 重建表格")
|
| 310 |
-
|
| 311 |
-
|
| 312 |
-
|
| 313 |
-
height=300
|
| 314 |
)
|
| 315 |
|
| 316 |
-
# 說明
|
| 317 |
-
with gr.Accordion("
|
| 318 |
gr.Markdown("""
|
| 319 |
-
###
|
| 320 |
-
|
| 321 |
-
|
| 322 |
-
|
| 323 |
-
|
| 324 |
|
| 325 |
-
###
|
| 326 |
-
|
| 327 |
-
|
| 328 |
-
|
| 329 |
-
4. **結果導出**:可以複製表格數據到其他應用程序
|
| 330 |
|
| 331 |
-
###
|
| 332 |
-
-
|
| 333 |
-
-
|
| 334 |
-
-
|
| 335 |
-
-
|
| 336 |
""")
|
| 337 |
|
| 338 |
# 綁定事件
|
| 339 |
-
|
| 340 |
-
fn=
|
| 341 |
-
inputs=[image_input,
|
| 342 |
-
outputs=[
|
| 343 |
)
|
| 344 |
|
|
|
|
| 345 |
image_input.change(
|
| 346 |
-
fn=
|
| 347 |
-
inputs=[image_input,
|
| 348 |
-
outputs=[
|
| 349 |
)
|
| 350 |
|
| 351 |
-
return
|
| 352 |
|
| 353 |
if __name__ == "__main__":
|
| 354 |
-
|
| 355 |
-
|
| 356 |
server_name="0.0.0.0",
|
| 357 |
server_port=7860,
|
| 358 |
-
share=True
|
| 359 |
-
show_error=True
|
| 360 |
)
|
|
|
|
| 4 |
import easyocr
|
| 5 |
import numpy as np
|
| 6 |
import re
|
|
|
|
| 7 |
import traceback
|
| 8 |
|
| 9 |
+
class FinalOCRProcessor:
|
| 10 |
def __init__(self):
|
| 11 |
self.reader = None
|
| 12 |
+
self.lang_config = "未初始化"
|
| 13 |
self.initialize_reader()
|
| 14 |
|
| 15 |
def initialize_reader(self):
|
| 16 |
"""安全初始化EasyOCR"""
|
| 17 |
+
print("正在初始化OCR引擎...")
|
| 18 |
+
|
| 19 |
+
# 嘗試不同的語言配置
|
| 20 |
+
configs = [
|
| 21 |
+
(['ch_tra', 'en'], "繁體中文+英文"),
|
| 22 |
+
(['ch_sim', 'en'], "簡體中文+英文"),
|
| 23 |
+
(['en'], "僅英文")
|
| 24 |
+
]
|
| 25 |
+
|
| 26 |
+
for lang_list, description in configs:
|
| 27 |
try:
|
| 28 |
+
print(f"嘗試配置:{description}")
|
| 29 |
+
self.reader = easyocr.Reader(lang_list, gpu=False)
|
| 30 |
+
self.lang_config = description
|
| 31 |
+
print(f"✓ 成功初始化:{description}")
|
| 32 |
+
return
|
| 33 |
+
except Exception as e:
|
| 34 |
+
print(f"✗ 配置 {description} 失敗:{str(e)}")
|
| 35 |
+
continue
|
| 36 |
+
|
| 37 |
+
print("❌ 所有配置都失敗了")
|
| 38 |
+
self.reader = None
|
| 39 |
+
self.lang_config = "初始化失敗"
|
|
|
|
|
|
|
| 40 |
|
| 41 |
+
def process_image(self, image):
|
| 42 |
+
"""處理圖片的主要函數"""
|
| 43 |
if self.reader is None:
|
| 44 |
+
return self.create_error_result("OCR引擎未正確初始化")
|
| 45 |
+
|
| 46 |
+
if image is None:
|
| 47 |
+
return self.create_error_result("請上傳圖片")
|
| 48 |
|
| 49 |
try:
|
| 50 |
+
# 轉換圖片格式
|
| 51 |
if isinstance(image, np.ndarray):
|
| 52 |
+
img_array = image
|
| 53 |
+
else:
|
| 54 |
+
img_array = np.array(image)
|
| 55 |
|
| 56 |
+
print("開始OCR識別...")
|
|
|
|
| 57 |
|
| 58 |
# 執行OCR
|
|
|
|
| 59 |
results = self.reader.readtext(img_array)
|
|
|
|
| 60 |
|
| 61 |
+
print(f"識別完成,找到 {len(results)} 個文字區塊")
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 62 |
|
| 63 |
+
if not results:
|
| 64 |
+
return self.create_empty_result("圖片中未找到任何文字")
|
| 65 |
+
|
| 66 |
+
# 處理OCR結果
|
| 67 |
+
data_list = []
|
| 68 |
+
for i, (bbox, text, confidence) in enumerate(results):
|
| 69 |
+
# 計算位置信息
|
| 70 |
+
x_coords = [point[0] for point in bbox]
|
| 71 |
+
y_coords = [point[1] for point in bbox]
|
| 72 |
|
| 73 |
+
left = int(min(x_coords))
|
| 74 |
+
right = int(max(x_coords))
|
| 75 |
+
top = int(min(y_coords))
|
| 76 |
+
bottom = int(max(y_coords))
|
| 77 |
+
center_x = int((left + right) / 2)
|
| 78 |
+
center_y = int((top + bottom) / 2)
|
| 79 |
|
| 80 |
+
data_list.append({
|
| 81 |
+
'序號': i + 1,
|
| 82 |
+
'識別文字': text.strip(),
|
| 83 |
+
'信心度': round(float(confidence), 3),
|
| 84 |
+
'中心X': center_x,
|
| 85 |
+
'中心Y': center_y,
|
| 86 |
+
'左': left,
|
| 87 |
+
'上': top,
|
| 88 |
+
'右': right,
|
| 89 |
+
'下': bottom,
|
| 90 |
+
'寬度': right - left,
|
| 91 |
+
'高度': bottom - top,
|
| 92 |
+
'類型': self.get_text_type(text.strip())
|
| 93 |
+
})
|
| 94 |
+
|
| 95 |
+
# 創建DataFrame並排序
|
| 96 |
+
df = pd.DataFrame(data_list)
|
| 97 |
+
df = df.sort_values(['中心Y', '中心X']).reset_index(drop=True)
|
| 98 |
+
|
| 99 |
+
# 重新編號
|
| 100 |
+
df['序號'] = range(1, len(df) + 1)
|
| 101 |
+
|
| 102 |
+
# 生成摘要
|
| 103 |
+
summary = self.generate_summary(df)
|
| 104 |
+
|
| 105 |
+
# 嘗試重建表格
|
| 106 |
+
table_df = self.reconstruct_table(df)
|
| 107 |
+
|
| 108 |
+
return df, table_df, summary
|
| 109 |
+
|
| 110 |
except Exception as e:
|
| 111 |
+
error_msg = f"處理圖片時出錯:{str(e)}"
|
| 112 |
print(error_msg)
|
| 113 |
print(traceback.format_exc())
|
| 114 |
+
return self.create_error_result(error_msg)
|
| 115 |
|
| 116 |
+
def get_text_type(self, text):
|
| 117 |
+
"""判斷文字類型"""
|
| 118 |
if not text:
|
| 119 |
+
return "空白"
|
| 120 |
|
| 121 |
+
# 日期格式
|
| 122 |
+
if re.search(r'\d+月\d+日', text):
|
| 123 |
+
return "日期"
|
| 124 |
+
|
| 125 |
+
# 時間格式
|
| 126 |
+
if re.search(r'\d{1,2}[-:]\d{1,2}', text):
|
| 127 |
+
return "時間"
|
| 128 |
+
|
| 129 |
+
# 純數字
|
| 130 |
+
if re.match(r'^\d+$', text):
|
| 131 |
+
return "數字"
|
| 132 |
+
|
| 133 |
+
# 中文姓名或詞語
|
| 134 |
+
if re.match(r'^[\u4e00-\u9fff]{1,6}$', text):
|
| 135 |
+
return "中文"
|
| 136 |
+
|
| 137 |
+
# 包含中文的混合內容
|
| 138 |
+
if re.search(r'[\u4e00-\u9fff]', text):
|
| 139 |
+
return "中文混合"
|
| 140 |
+
|
| 141 |
+
# 英文
|
| 142 |
+
if re.match(r'^[a-zA-Z\s]+$', text):
|
| 143 |
+
return "英文"
|
| 144 |
+
|
| 145 |
+
return "其他"
|
| 146 |
|
| 147 |
+
def generate_summary(self, df):
|
| 148 |
+
"""生成識別摘要"""
|
| 149 |
+
if df.empty:
|
| 150 |
+
return "沒有識別到任何內容"
|
| 151 |
+
|
| 152 |
+
total = len(df)
|
| 153 |
+
avg_conf = df['信心度'].mean()
|
| 154 |
+
high_conf = (df['信心度'] >= 0.8).sum()
|
| 155 |
+
|
| 156 |
+
summary = f"""🔍 OCR識別報告
|
| 157 |
+
{'='*40}
|
| 158 |
+
📊 基本統計
|
| 159 |
+
• 引擎配置:{self.lang_config}
|
| 160 |
+
• 識別區塊:{total} 個
|
| 161 |
+
• 平均信心度:{avg_conf:.3f}
|
| 162 |
+
• 高信心度區塊:{high_conf} 個 ({high_conf/total*100:.1f}%)
|
| 163 |
+
|
| 164 |
+
📝 文字類型統計
|
| 165 |
+
"""
|
| 166 |
+
|
| 167 |
+
# 統計各類型數量
|
| 168 |
+
type_counts = df['類型'].value_counts()
|
| 169 |
+
for text_type, count in type_counts.items():
|
| 170 |
+
percentage = count / total * 100
|
| 171 |
+
summary += f"• {text_type}:{count} 個 ({percentage:.1f}%)\n"
|
| 172 |
+
|
| 173 |
+
# 品質評估
|
| 174 |
+
if avg_conf >= 0.8:
|
| 175 |
+
quality = "優秀 ✅"
|
| 176 |
+
elif avg_conf >= 0.6:
|
| 177 |
+
quality = "良好 ⚠️"
|
| 178 |
+
else:
|
| 179 |
+
quality = "待改進 ❌"
|
| 180 |
+
|
| 181 |
+
summary += f"\n🎯 整體品質:{quality}"
|
| 182 |
+
|
| 183 |
+
return summary
|
| 184 |
+
|
| 185 |
+
def reconstruct_table(self, df):
|
| 186 |
"""重建表格結構"""
|
| 187 |
if df.empty:
|
| 188 |
return pd.DataFrame()
|
| 189 |
|
| 190 |
try:
|
| 191 |
+
# 按Y座標分組形成行
|
|
|
|
| 192 |
rows = []
|
| 193 |
+
sorted_df = df.sort_values('中心Y')
|
| 194 |
+
|
| 195 |
current_row = []
|
| 196 |
last_y = None
|
| 197 |
+
y_threshold = 25 # Y座標閾值
|
| 198 |
|
| 199 |
+
for _, item in sorted_df.iterrows():
|
| 200 |
+
if last_y is None or abs(item['中心Y'] - last_y) <= y_threshold:
|
| 201 |
+
current_row.append(item)
|
|
|
|
|
|
|
| 202 |
else:
|
| 203 |
if current_row:
|
| 204 |
+
# 按X座標排序
|
| 205 |
current_row.sort(key=lambda x: x['中心X'])
|
| 206 |
rows.append(current_row)
|
| 207 |
+
current_row = [item]
|
| 208 |
+
last_y = item['中心Y']
|
|
|
|
| 209 |
|
| 210 |
+
# 添加最後一行
|
| 211 |
if current_row:
|
| 212 |
current_row.sort(key=lambda x: x['中心X'])
|
| 213 |
rows.append(current_row)
|
| 214 |
|
|
|
|
| 215 |
if not rows:
|
| 216 |
return pd.DataFrame()
|
| 217 |
|
| 218 |
+
# 構建表格數據
|
| 219 |
table_data = []
|
| 220 |
+
max_cols = max(len(row) for row in rows)
|
| 221 |
|
| 222 |
+
for row_idx, row_items in enumerate(rows):
|
| 223 |
+
row_dict = {'行號': row_idx + 1}
|
| 224 |
+
|
| 225 |
+
for col_idx, item in enumerate(row_items):
|
| 226 |
+
col_name = f'第{col_idx + 1}列'
|
| 227 |
+
row_dict[col_name] = item['識別文字']
|
| 228 |
|
| 229 |
# 填充空列
|
| 230 |
+
for col_idx in range(len(row_items), max_cols):
|
| 231 |
+
col_name = f'第{col_idx + 1}列'
|
| 232 |
+
row_dict[col_name] = ""
|
| 233 |
|
| 234 |
+
table_data.append(row_dict)
|
| 235 |
|
| 236 |
return pd.DataFrame(table_data)
|
| 237 |
|
| 238 |
except Exception as e:
|
| 239 |
print(f"重建表格失敗:{e}")
|
| 240 |
+
return pd.DataFrame([{"錯誤": "表格重建失敗"}])
|
| 241 |
|
| 242 |
+
def create_error_result(self, message):
|
| 243 |
+
"""創建錯誤結果"""
|
| 244 |
+
error_df = pd.DataFrame([{"錯誤信息": message}])
|
| 245 |
+
return error_df, error_df, f"❌ {message}"
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 246 |
|
| 247 |
+
def create_empty_result(self, message):
|
| 248 |
+
"""創建空結果"""
|
| 249 |
empty_df = pd.DataFrame()
|
| 250 |
+
return empty_df, empty_df, f"ℹ️ {message}"
|
| 251 |
|
| 252 |
+
def main():
|
| 253 |
+
"""主函數,創建Gradio應用"""
|
|
|
|
| 254 |
|
| 255 |
+
# 初始化處理器
|
| 256 |
+
processor = FinalOCRProcessor()
|
| 257 |
+
|
| 258 |
+
def process_image_wrapper(image, min_confidence):
|
| 259 |
+
"""包裝處理函數"""
|
| 260 |
try:
|
| 261 |
+
# 處理圖片
|
| 262 |
+
df, table_df, summary = processor.process_image(image)
|
| 263 |
|
| 264 |
# 應用信心度過濾
|
| 265 |
if not df.empty and '信心度' in df.columns:
|
| 266 |
+
original_count = len(df)
|
| 267 |
+
df = df[df['信心度'] >= min_confidence].reset_index(drop=True)
|
| 268 |
+
filtered_count = len(df)
|
| 269 |
+
|
| 270 |
+
if filtered_count < original_count:
|
| 271 |
+
summary += f"\n\n🔍 信心度過濾:保留 {filtered_count}/{original_count} 個結果"
|
| 272 |
|
| 273 |
return summary, df, table_df
|
| 274 |
|
| 275 |
except Exception as e:
|
| 276 |
error_msg = f"處理失敗:{str(e)}"
|
| 277 |
+
error_df = pd.DataFrame([{"錯誤": error_msg}])
|
| 278 |
return error_msg, error_df, error_df
|
| 279 |
|
| 280 |
+
# 創建Gradio界面
|
| 281 |
+
with gr.Blocks(
|
| 282 |
+
title="中文OCR識別系統",
|
| 283 |
+
theme=gr.themes.Default()
|
| 284 |
+
) as app:
|
| 285 |
+
|
| 286 |
+
# 標題和說明
|
| 287 |
gr.HTML("""
|
| 288 |
+
<div style="text-align: center; padding: 20px; background: #f0f2f6; border-radius: 10px; margin-bottom: 20px;">
|
| 289 |
+
<h1 style="color: #1f2937; margin-bottom: 10px;">🔍 中文OCR識別系統</h1>
|
| 290 |
+
<p style="color: #6b7280; font-size: 16px;">
|
| 291 |
+
支持中文、英文、數字識別,自動重建表格結構
|
| 292 |
+
</p>
|
| 293 |
</div>
|
| 294 |
""")
|
| 295 |
|
| 296 |
+
# 主要操作區域
|
| 297 |
with gr.Row():
|
| 298 |
+
# 左側:圖片上傳和控制
|
| 299 |
with gr.Column(scale=1):
|
| 300 |
image_input = gr.Image(
|
| 301 |
label="📤 上傳圖片",
|
| 302 |
+
type="pil"
|
|
|
|
| 303 |
)
|
| 304 |
|
| 305 |
+
confidence_slider = gr.Slider(
|
| 306 |
minimum=0.0,
|
| 307 |
maximum=1.0,
|
| 308 |
value=0.3,
|
| 309 |
step=0.1,
|
| 310 |
+
label="🎯 最低信心度閾值"
|
|
|
|
| 311 |
)
|
| 312 |
|
| 313 |
+
process_button = gr.Button(
|
| 314 |
"🚀 開始識別",
|
| 315 |
+
variant="primary"
|
|
|
|
| 316 |
)
|
| 317 |
|
| 318 |
+
# 右側:摘要報告
|
| 319 |
with gr.Column(scale=1):
|
| 320 |
+
summary_output = gr.Textbox(
|
| 321 |
label="📊 識別報告",
|
| 322 |
+
lines=15,
|
| 323 |
+
max_lines=20
|
| 324 |
)
|
| 325 |
|
| 326 |
+
# 結果展示區域
|
| 327 |
with gr.Row():
|
| 328 |
with gr.Column():
|
| 329 |
gr.Markdown("### 📋 詳細識別結果")
|
| 330 |
+
detail_output = gr.Dataframe(
|
| 331 |
+
label="所有識別的文字內容",
|
| 332 |
+
interactive=True
|
|
|
|
| 333 |
)
|
| 334 |
|
| 335 |
with gr.Column():
|
| 336 |
gr.Markdown("### 🔄 重建表格")
|
| 337 |
+
table_output = gr.Dataframe(
|
| 338 |
+
label="按表格結構重新組織的數據",
|
| 339 |
+
interactive=True
|
|
|
|
| 340 |
)
|
| 341 |
|
| 342 |
+
# 使用說明
|
| 343 |
+
with gr.Accordion("📖 使用說明", open=False):
|
| 344 |
gr.Markdown("""
|
| 345 |
+
### 🚀 快速開始
|
| 346 |
+
1. **上傳圖片**:點擊上傳包含文字的圖片
|
| 347 |
+
2. **調整參數**:設置最低信心度(建議0.3-0.7)
|
| 348 |
+
3. **開始識別**:點擊按鈕或直接上傳圖片自動處理
|
| 349 |
+
4. **查看結果**:在兩個表格中查看識別結果
|
| 350 |
|
| 351 |
+
### 📊 結果說明
|
| 352 |
+
- **詳細識別結果**:每個文字區塊的完整信息
|
| 353 |
+
- **重建表格**:嘗試還原原始表格的行列結構
|
| 354 |
+
- **識別報告**:統計信息和品質評估
|
|
|
|
| 355 |
|
| 356 |
+
### 💡 優化建議
|
| 357 |
+
- 使用清晰、高對比度的圖片
|
| 358 |
+
- 確保文字大小適中(不要太小)
|
| 359 |
+
- 避免圖片傾斜或變形
|
| 360 |
+
- 調整信心度閾值獲得最佳結果
|
| 361 |
""")
|
| 362 |
|
| 363 |
# 綁定事件
|
| 364 |
+
process_button.click(
|
| 365 |
+
fn=process_image_wrapper,
|
| 366 |
+
inputs=[image_input, confidence_slider],
|
| 367 |
+
outputs=[summary_output, detail_output, table_output]
|
| 368 |
)
|
| 369 |
|
| 370 |
+
# 圖片上傳時自動處理
|
| 371 |
image_input.change(
|
| 372 |
+
fn=process_image_wrapper,
|
| 373 |
+
inputs=[image_input, confidence_slider],
|
| 374 |
+
outputs=[summary_output, detail_output, table_output]
|
| 375 |
)
|
| 376 |
|
| 377 |
+
return app
|
| 378 |
|
| 379 |
if __name__ == "__main__":
|
| 380 |
+
app = main()
|
| 381 |
+
app.launch(
|
| 382 |
server_name="0.0.0.0",
|
| 383 |
server_port=7860,
|
| 384 |
+
share=True
|
|
|
|
| 385 |
)
|