大数据集按间隔抽取，做成小数据集

import os
import shutil


def extract_subset(img_dir, labels_dir, output_dir, interval=15, image_ext='.jpg', label_ext='.txt'):
    # Create output directories if they don't exist
    output_img_dir = os.path.join(output_dir, 'img')
    output_labels_dir = os.path.join(output_dir, 'labels')
    os.makedirs(output_img_dir, exist_ok=True)
    os.makedirs(output_labels_dir, exist_ok=True)

    # List all images assuming they end with the specified image extension
    images = [f for f in os.listdir(img_dir) if f.endswith(image_ext)]

    # Sort images to maintain a consistent order
    images.sort()

    # Filter images to pick every 15th image
    selected_images = images[::interval]

    # Copy selected images and their corresponding YOLO label files
    for img in selected_images:
        # Construct the full path for image and label
        img_path = os.path.join(img_dir, img)
        label_path = os.path.join(labels_dir, os.path.splitext(img)[
            0] + label_ext)  # Assumes label has same base name with .txt extension

        # Check if corresponding label file exists
        if os.path.exists(label_path):
            # Copy image
            shutil.copy(img_path, output_img_dir)
            # Copy label
            shutil.copy(label_path, output_labels_dir)
        else:
            print(f"Label file for {img} not found. Skipping this file.")

    print(f"Extracted {len(selected_images)} images and their labels to {output_dir}")


# Example usage
if __name__ == "__main__":
    img_dir = "D:\Myself文档\Dataset\官方Dataset\spain\original\original\imgs"
    labels_dir = "D:\Myself文档\Dataset\官方Dataset\spain\original\original\yolo_labels"
    new_dataset_dir = "D:\Myself文档\Dataset\官方Dataset\spain\original\original_select_15"
    extract_subset(img_dir, labels_dir, new_dataset_dir)

img_dir，原数据集img目录地址

labels_dir，原数据集labels目录地址（YOLO格式）

new_dataset_dir，形成新数据集的目录地址，脚本会自动生产两个子目录img/labels  注意yolo数据集命名必须为images/labels。所以这里最终生成文件名称img的需要改一下

interval=15，控制提取间隔