使用Python根据JSON将图像分类为文件夹

Question

我有一个图像数据集，我需要将其转换为TensorFlow格式进行图像分类。数据集JSON元数据格式如下：

{
  "0": {
    "image_filepath": "images/0.jpg",
    "anomaly_class": "A"
  },
  "1": {
    "image_filepath": "images/1.jpg",
    "anomaly_class": "B"
  },
  "2": {
    "image_filepath": "images/2.jpg",
    "anomaly_class": "C"
  },
  "3": {
    "image_filepath": "images/3.jpg",
    "anomaly_class": "D"
  },
  "4": {
    "image_filepath": "images/4.jpg",
    "anomaly_class": "E"
  },
  "5": {
    "image_filepath": "images/5.jpg",
    "anomaly_class": "F"
  },
  "6": {
    "image_filepath": "images/6.jpg",
    "anomaly_class": "G"
  },
  "7": {
    "image_filepath": "images/7.jpg",
    "anomaly_class": "H"
  },
  "8": {
    "image_filepath": "images/8.jpg",
    "anomaly_class": "I"
  },
  "9": {
    "image_filepath": "images/9.jpg",
    "anomaly_class": "J"
  },
  "10": {
    "image_filepath": "images/10.jpg",
    "anomaly_class": "K"
  },
  "11": {
    "image_filepath": "images/11.jpg",
    "anomaly_class": "L"
  },
  "12": {
    "image_filepath": "images/12.jpg",
    "anomaly_class": "M"
  },
.
.
.

"1000": {
    "image_filepath": "images/1000.jpg",
    "anomaly_class": "A"
  }
}

我想编写一个Python脚本，该脚本使用键“ Anomaly_class”的值创建文件夹名称，然后将图像分类/将图像移动到其各自的文件夹中。

有帮助吗？

Answer 1

这应该有效：

import os

json_object = {
  "0": {
    "image_filepath": "images/0.jpg",
    "anomaly_class": "A"
  },
  "1": {
    "image_filepath": "images/1.jpg",
    "anomaly_class": "B"
  },
  "2": {
    "image_filepath": "images/2.jpg",
    "anomaly_class": "C"
  }
}

# Parent Directory path
parent_dir = "D:/PyProject/"

for key in json_object:
    anomaly_class_folder_name = json_object[key]['anomaly_class']
    image_fp = json_object[key]['image_filepath']

    # Strip path from image file
    file_name = os.path.basename(image_fp)

    # Create Folder
    path = os.path.join(parent_dir, anomaly_class_folder_name)
    os.mkdir(path)

    # Move file to target directory
    os.replace(image_fp, path + "/" + file_name)