时间戳

首页 > 技术文章 > 将彩色RGB分割标注图像数据集转换为COCO格式的JSON文件

将彩色RGB分割标注图像数据集转换为COCO格式的JSON文件

xiaochouk 2022-03-13 10:20 原文

https://www.freesion.com/article/95661187982/

 

将彩色RGB分割标注图像数据集转换为COCO格式的JSON文件

 
 

由于很多检测、分割网络对coco格式的数据集都兼容支持，有时候需要将自己的数据集转化为coco格式的json文件，写一篇博客记录一下自己将彩色RGB的分割标注图像转换为coco格式文件的过程。

彩色 label：

转换为单个物体的黑白mask：

生成最终的 json 文件：

具体流程：

1. 安装pycococreator（先安装pycocotools, cython）：

# windows
pip install git+https://github.com/philferriere/cocoapi.git#egg=pycocotools^&subdirectory=PythonAPI
# Linux
pip install git+https://github.com/waleedka/cocoapi.git#egg=pycocotools&subdirectory=PythonAPI
 
 
pip install cython
pip install git+git://github.com/waspinator/coco.git@2.1.0

2. 创建文件目录格式如下 ：

shapes
│
|
|———labels
|        | <image_id>.png
|        | ...
|
└───train
    │ 
    └───annotations
    │    │ <image_id>_<object_class_name>_<annotation_id>.png
    │    │ ...
    │   
    └───<subset><year>
         │   <image_id>.png
         │   ...
 
e.g
 
shapes
│
|
|———labels
|        | 5.png
|        | ...
|
└───train
    │ 
    └───annotations
    │    │ 5_leaf_0.png        # 第5张图片的第0个叶片
    │    │ ...
    │   
    └───shapes_train2017
         │   5.png             # 第5张图片
         │   ...

3. 将彩色RGB的标注图像转化为黑白图像，命名格式为 ：

<image_id>_<object_class_name>_<annotation_id>.png
# e.g  第5张图片的第0个叶片
./shapes/train/annotations/5_leaf_0.png

 RGB标注彩图转换为单个物体的黑白mask图像代码为：

import cv2
import numpy as np
import os, glob
 
 
def rgb2masks(label_name):
    lbl_id = os.path.split(label_name)[-1].split('.')[0]
    lbl = cv2.imread(label_name, 1)
    h, w = lbl.shape[:2]
    leaf_dict = {}
    idx = 0
    white_mask = np.ones((h, w, 3), dtype=np.uint8) * 255
    for i in range(h):
        for j in range(w):
            if tuple(lbl[i][j]) in leaf_dict or tuple(lbl[i][j]) == (0, 0, 0):
                continue
            leaf_dict[tuple(lbl[i][j])] = idx
            mask = (lbl == lbl[i][j]).all(-1)
            # leaf = lbl * mask[..., None]      # colorful leaf with black background
            # np.repeat(mask[...,None],3,axis=2)    # 3D mask
            leaf = np.where(mask[..., None], white_mask, 0)
            mask_name = './shapes/train/annotations/' + lbl_id + '_leaf_' + str(idx) + '.png'
            cv2.imwrite(mask_name, leaf)
            idx += 1
 
 
label_dir = './labels'
label_list = glob.glob(os.path.join(label_dir, '*.png'))
for label_name in label_list:
    rgb2masks(label_name)

4. 利用pycococreator和得到的黑白masks生成coco json格式的数据集，代码如下：

import datetime
import json
import os
import re
import fnmatch
from PIL import Image
import numpy as np
from pycococreatortools import pycococreatortools

ROOT_DIR = 'C:/Users/1/Desktop/333/tokyo/tokyo/'
IMAGE_DIR = os.path.join(ROOT_DIR, "img")
ANNOTATION_DIR = os.path.join(ROOT_DIR, "anno")

INFO = {
    "description": "Leaf Dataset",
    "url": "https://github.com/waspinator/pycococreator",
    "version": "0.1.0",
    "year": 2017,
    "contributor": "Francis_Liu",
    "date_created": datetime.datetime.utcnow().isoformat(' ')
}

LICENSES = [
    {
        "id": 1,
        "name": "Attribution-NonCommercial-ShareAlike License",
        "url": "http://creativecommons.org/licenses/by-nc-sa/2.0/"
    }
]

# 根据自己的需要添加种类
CATEGORIES = [
    {
        'id': 1,
        'name': 'tokyo',
        'supercategory': 'tokyo',
    }
]


def filter_for_jpeg(root, files):
    file_types = ['*.jpeg', '*.jpg', '*.png']
    file_types = r'|'.join([fnmatch.translate(x) for x in file_types])
    files = [os.path.join(root, f) for f in files]
    files = [f for f in files if re.match(file_types, f)]
    return files


def filter_for_annotations(root, files, image_filename):
    file_types = ['*.png']
    file_types = r'|'.join([fnmatch.translate(x) for x in file_types])
    basename_no_extension = os.path.splitext(os.path.basename(image_filename))[0]
    file_name_prefix = basename_no_extension + '.*'
    files = [os.path.join(root, f) for f in files]
    files = [f for f in files if re.match(file_types, f)]
    files = [f for f in files if re.match(file_name_prefix, os.path.splitext(os.path.basename(f))[0])]
    return files


def main():
    coco_output = {
        "info": INFO,
        "licenses": LICENSES,
        "categories": CATEGORIES,
        "images": [],
        "annotations": []
    }

    image_id = 1
    segmentation_id = 1

    # filter for jpeg images
    for root, _, files in os.walk(IMAGE_DIR):
        image_files = filter_for_jpeg(root, files)

        # go through each image
        for image_filename in image_files:
            image = Image.open(image_filename)
            image_info = pycococreatortools.create_image_info(
                image_id, os.path.basename(image_filename), image.size)
            coco_output["images"].append(image_info)

            # filter for associated png annotations
            for root, _, files in os.walk(ANNOTATION_DIR):
                annotation_files = filter_for_annotations(root, files, image_filename)

                # go through each associated annotation
                for annotation_filename in annotation_files:

                    print(annotation_filename)
                    class_id = [x['id'] for x in CATEGORIES if x['name'] in annotation_filename]

                    category_info = {'id': class_id, 'is_crowd': 'tokyo' in image_filename}
                    binary_mask = np.asarray(Image.open(annotation_filename)
                                             .convert('1')).astype(np.uint8)

                    annotation_info = pycococreatortools.create_annotation_info(
                        segmentation_id, image_id, category_info, binary_mask,
                        image.size, tolerance=2)

                    if annotation_info is not None:
                        coco_output["annotations"].append(annotation_info)

                    segmentation_id = segmentation_id + 1

            image_id = image_id + 1

    with open('{}/instances_leaf_train2017.json'.format(ROOT_DIR), 'w') as output_json_file:
        json.dump(coco_output, output_json_file)


if __name__ == "__main__":
    main()

 

 

 

 

 

 

 

 

 

 

 

 

 

推荐阅读