From 10c66ac1c8f9c97172493944f5ced3e910f382e8 Mon Sep 17 00:00:00 2001
From: BUJIQI <145289312+BUJIQI@users.noreply.github.com>
Date: Wed, 11 Dec 2024 10:47:27 +0800
Subject: [PATCH] =?UTF-8?q?Update=20=E5=8D=B0=E7=AB=A0=E5=BC=AF=E6=9B=B2?=
 =?UTF-8?q?=E6=96=87=E5=AD=97=E8=AF=86=E5=88=AB.md?=
MIME-Version: 1.0
Content-Type: text/plain; charset=UTF-8
Content-Transfer-Encoding: 8bit

---
 ...07\345\255\227\350\257\206\345\210\253.md" | 56 +++++++++++++++++++
 1 file changed, 56 insertions(+)

diff --git "a/docs/applications/\345\215\260\347\253\240\345\274\257\346\233\262\346\226\207\345\255\227\350\257\206\345\210\253.md" "b/docs/applications/\345\215\260\347\253\240\345\274\257\346\233\262\346\226\207\345\255\227\350\257\206\345\210\253.md"
index be9b2f6beaf..0b60efd15c2 100644
--- "a/docs/applications/\345\215\260\347\253\240\345\274\257\346\233\262\346\226\207\345\255\227\350\257\206\345\210\253.md"
+++ "b/docs/applications/\345\215\260\347\253\240\345\274\257\346\233\262\346\226\207\345\255\227\350\257\206\345\210\253.md"
@@ -995,6 +995,62 @@ def run(data_dir, label_file, save_dir):
 
 ```
 
+若训练数据使用**4点标注以外的多点标注**（如8点，16点等不规则形状标注），可以使用以下代码进行印章文字区域剪切
+
+<details>
+  
+```python linenums="1"
+import cv2
+import numpy as np
+import os
+import json
+
+def get_polygon_crop_image(img, points):
+    points = np.array(points, dtype=np.int32)
+    mask = np.zeros_like(img, dtype=np.uint8)
+    cv2.fillPoly(mask, [points], (255, 255, 255))
+    result = cv2.bitwise_and(img, mask)
+    x, y, w, h = cv2.boundingRect(points)
+    cropped_result = result[y:y+h, x:x+w]
+    return cropped_result
+
+
+def run(data_dir, label_file, save_dir, output_txt):
+    if not os.path.exists(output_txt):
+        os.makedirs(os.path.dirname(output_txt), exist_ok=True)
+        open(output_txt, 'w').close()
+
+    with open(output_txt, 'w') as txt_file:
+        datas = open(label_file, 'r').readlines()
+        for line in datas:
+            filename, label = line.strip().split('\t') 
+            img_path = os.path.join(data_dir, filename) 
+            label = json.loads(label)
+            src_im = cv2.imread(img_path) 
+            if src_im is None:
+                continue
+            for i, anno in enumerate(label):
+                txt_boxes = anno['points']
+                crop_im = get_polygon_crop_image(src_im, txt_boxes)
+                crop_img_name = f'{filename.split("/")[-1].split(".")[0]}_crop_{i}.jpg'
+                save_path = os.path.join(save_dir, crop_img_name)
+                if not os.path.exists(save_dir):
+                    os.makedirs(save_dir)
+                cv2.imwrite(save_path, crop_im)
+                txt_file.write(f'crop_img/{crop_img_name}\t{anno["transcription"]}\n')
+
+if __name__ == "__main__":
+    data_dir = "seal" # 图片数据集路径
+    label_file = "Label.txt" # 数据标记结果txt路径
+    save_dir = "crop_img" #导出识别结果，即识别所用训练图片的文件夹路径
+    output_txt = "rec_gt.txt" # 识别训练标记结果txt路径
+    run(data_dir, label_file, save_dir, output_txt)
+    
+```
+
+</details>
+
+
 数据处理完成后，即可配置训练的配置文件。SVTR配置文件选择[configs/rec/PP-OCRv3/ch_PP-OCRv3_rec.yml](https://github.com/PaddlePaddle/PaddleOCR/blob/release/2.6/configs/rec/PP-OCRv3/ch_PP-OCRv3_rec.yml)
 修改SVTR配置文件中的训练数据部分如下：