CrowdHuman数据集标注格式转换为YOLOv3可以使用的COCO格式

需要了解CrowdHuman的数据标注格式odgt，YOLOv3需要的COCO格式（不需要使用json文件，只需要图片位置信息和标注信息）

YOLOv3 github地址：https://github.com/eriklindernoren/PyTorch-YOLOv3

保存每一张图片的位置信息

import os
import json


def load_file(fpath):  # fpath是具体的文件 ，作用：#str to list
    assert os.path.exists(fpath)  # assert() raise-if-not
    with open(fpath, 'r') as fid:
        lines = fid.readlines()
    records = [json.loads(line.strip('
')) for line in lines]  # str to list
    return records


def img2txt(odgtpath, respath):
    records = load_file(odgtpath)  # 提取odgt文件数据
    record_list = len(records)  # 获得record的长度，循环遍历所有数据。
    print(os.getcwd())
    # os.mkdir(os.getcwd() + respath)
    with open(respath, 'w') as txt:
        for i in range(record_list):
            file_name = records[i]['ID'] + '.jpg'
            file_name = str("/datasets/crowdhuman/images/val/Image/" + file_name)
            txt.write(file_name + '
')


if __name__ == '__main__':
    odgtpath = "/datasets/crowdhuman/annotation_val.odgt"
    respath = "/datasets/crowdhuman/val_name.txt"
    img2txt(odgtpath, respath)

保存每一张图片标注信息中的全身坐标fbox

import time
import img2txt
from PIL import Image


def tonormlabel(odgtpath, storepath):
    records = img2txt.load_file(odgtpath)
    record_list = len(records)
    print(record_list)
    categories = {}
    # txt = open(respath, 'w')
    for i in range(record_list):
        txt_name = storepath + records[i]['ID'] + '.txt'
        file_name = records[i]['ID'] + '.jpg'
        #print(i)
        im = Image.open("/datasets/crowdhuman/images/train_all/Image/" + file_name)
        height = im.size[1]
        width = im.size[0]
        file = open(txt_name, 'w')
        gt_box = records[i]['gtboxes']
        gt_box_len = len(gt_box)  # 每一个字典gtboxes里，也有好几个记录，分别提取记录。
        for j in range(gt_box_len):
            category = gt_box[j]['tag']
            if category not in categories:  # 该类型不在categories，就添加上去
                new_id = len(categories) + 1  # ID递增
                categories[category] = new_id
            category_id = categories[category]  # 重新获取它的类别ID
            fbox = gt_box[j]['fbox']  # 获得全身框
            norm_x = fbox[0] / width
            norm_y = fbox[1] / height
            norm_w = fbox[2] / width
            norm_h = fbox[3] / height
            '''
            norm_x = 0 if norm_x <= 0 else norm_x
            norm_x = 1 if norm_x >= 1 else norm_x
            norm_y = 0 if norm_y <= 0 else norm_y
            norm_y = 1 if norm_y >= 1 else norm_y
            norm_w = 0 if norm_w <= 0 else norm_w
            norm_w = 1 if norm_w >= 1 else norm_w
            norm_h = 0 if norm_h <= 0 else norm_h
            norm_h = 1 if norm_h >= 1 else norm_h
            '''
            blank = ' '
            if j == gt_box_len-1:
                file.write(str(category_id - 1) + blank + '{:.6f}'.format(norm_x) + blank + '{:.6f}'.format(norm_y) + blank
                           + '{:.6f}'.format(norm_w) + blank + '{:.6f}'.format(norm_h))
            else:
                file.write(str(category_id - 1) + blank + '{:.6f}'.format(norm_x) + blank + '{:.6f}'.format(norm_y) + blank
                           + '{:.6f}'.format(norm_w) + blank + '{:.6f}'.format(norm_h) + '
')


if __name__ == '__main__':
    odgtpath = "/datasets/crowdhuman/annotation_train.odgt"  
    storepath = "/datasets/crowdhuman/labels/train_all/Image/"
    print(time.strftime('%Y-%m-%d %H:%M:%S', time.localtime(time.time())))  # 格式化输出时间
    start = time.time()
    tonormlabel(odgtpath, storepath)
    end = time.time()
    print(time.strftime('%Y-%m-%d %H:%M:%S', time.localtime(time.time())))
    print('已完成转换，共耗时{:.5f}s'.format(end - start))