Prechádzať zdrojové kódy

[feat] publaynet 数据集处理

WangChao 2 mesiacov pred
rodič
commit
576d2b2218
1 zmenil súbory, kde vykonal 71 pridanie a 0 odobranie
  1. 71 0
      data_collection/publaynet.py

+ 71 - 0
data_collection/publaynet.py

@@ -0,0 +1,71 @@
+import argparse
+import ijson
+import os
+import shutil
+from tqdm import tqdm
+
+
+def copy_file_to_directory(source_dir, target_dir, filename):
+    # 构建源文件路径
+    source_file = os.path.join(source_dir, filename)
+
+    # 检查文件是否存在
+    if os.path.isfile(source_file):
+        # 构建目标文件路径
+        target_file = os.path.join(target_dir, filename)
+
+        # 确保目标目录存在
+        os.makedirs(target_dir, exist_ok=True)
+
+        # 拷贝文件
+        shutil.copy(source_file, target_file)
+    else:
+        print(f"文件 {source_file} 不存在")
+
+
+def find_filenames_for_category(json_file, category_id, image_dir, output_dir):
+    # Step 1: 找出 category_id 为 3 的 image_id 列表
+    matching_image_ids = set()
+
+    # 首先计算 annotations 的总项数,方便在 tqdm 进度条中使用
+    print('load annotations...')
+    with open(json_file, 'r', encoding='utf-8') as f:
+        total_annotations = sum(1 for _ in ijson.items(f, 'annotations.item'))
+    print('annotations count : ', total_annotations)
+
+    with open(json_file, 'r', encoding='utf-8') as f:
+        # 使用 tqdm 包装 ijson.items,显示解析 annotations 的进度
+        for annotation in tqdm(ijson.items(f, 'annotations.item'), total=total_annotations,
+                               desc="Processing annotations"):
+            if annotation['category_id'] == category_id:
+                matching_image_ids.add(annotation['image_id'])
+
+    # Step 2: 根据 image_id 找到对应的 file_name
+    matching_filenames = []
+
+    print('load images...')
+    # 计算 images 的总项数,方便在 tqdm 进度条中使用
+    with open(json_file, 'r', encoding='utf-8') as f:
+        total_images = sum(1 for _ in ijson.items(f, 'images.item'))
+    print('images count : ', total_images)
+
+    count = 0
+    with open(json_file, 'r', encoding='utf-8') as f:
+        # 使用 tqdm 包装 ijson.items,显示解析 images 的进度
+        for image in tqdm(ijson.items(f, 'images.item'), total=total_images, desc="Processing images"):
+            if image['id'] in matching_image_ids:
+                count += 1
+                copy_file_to_directory(image_dir, output_dir, image['file_name'])
+
+    print(count, 'files matched')
+
+
+# 主程序入口,处理命令行参数
+if __name__ == '__main__':
+    parser = argparse.ArgumentParser("处理数据集\n")
+    parser.add_argument("json_file", help="publaynet json 文件路径")
+    parser.add_argument('image_dir', help='图片目录')
+    parser.add_argument('output', help='输出目录')
+
+    args = parser.parse_args()
+    find_filenames_for_category(args.json_file, 3, args.image_dir, args.output)