BDD100K Label을 Yolo Label foramt으로 변환하기

python bdd100k_to_yolo.py -l bdd100k\label   -i bdd100k\image           -p bdd100k

                                         label 경로            이미지 경로          image_list.txt 저장 경로

import os
import json
from tqdm import tqdm
import argparse 

categorys = ['bike', 'bus', 'car', 'motor', 'person', 'rider',  'traffic light', 'traffic sign', 'train', 'truck']

def label2det(frames, args, target):
    if target == 'train':
        img_list = open(args.tarin_image_result,"w")
        img_path = args.train_image_dir+"\\"
    elif target == 'val':
        img_list = open(args.val_image_result, "w")
        img_path = args.val_image_dir+"\\"
    else:
        raise Exception('invalid target!!')

    for frame in tqdm(frames):
        iname = frame['name']
        img_list.write(img_path+iname+"\n")
        fname = iname[:-4]
#Creates, opens, and adds to a txt file with the name of each image.jpg
        if target == 'train':
            img_label = open(args.train_image_dir + "\\" + fname + ".txt","w+")
        elif target == 'val':
            img_label = open(args.val_image_dir + "\\" + fname + ".txt","w+")
        else:
            raise Exception('invalid target!!')

#For each sub label of each image, get the box2d variable
#Get the relative center point compared to the image size 1280/720
        for label in frame['labels']:
            if 'box2d' not in label:
                continue
            xy = label['box2d']
            if xy['x1'] >= xy['x2'] or xy['y1'] >= xy['y2']:
                continue
            X = xy['x1']/1280
            Y = xy['y1']/720
            MX = ((xy['x1'] + xy['x2']) / 2)/1280
            MY = ((xy['y1'] + xy['y2']) / 2)/720
            W = xy['x2']/1280
            H = xy['y2']/720
            if X > W or Y > H:
                continue
            lbl = -1
#provide a number corresponding to the category of sub label for darknet format.
            if(label['category'] == "bike"):
                lbl = 0
            if(label['category'] == "bus"):
                lbl = 1
            if(label['category'] == "car"):
                lbl = 2
            if(label['category'] == "motor"):
                lbl = 3
            if(label['category'] == "person"):
                lbl = 4
            if(label['category'] == "rider"):
                lbl = 5
            if(label['category'] == "traffic light"):
                lbl = 6
            if(label['category'] == "traffic sign"):
                lbl = 7
            if(label['category'] == "train"):
                lbl = 8
            if(label['category'] == "truck"):
                lbl = 9
            img_label.write(repr(lbl) + " " + repr(MX) + " " + repr(MY) + " " + repr(W-X) + " " + repr(H-Y) + '\n')
    
    img_list.close()
 
def parseJson(jsonFile):
    '''
      params:
                 jsonFile -- a json tag file for the BDD00K data set
      return:
                 Returns a list of lists that store the coordinates of the boxes in a json file and the classes they belong to.
    '''
    objs = []
    obj = []
    info = jsonFile
    name = info['name']
    objects = info['labels']
    for i in objects:
        if(i['category'] in categorys):
            obj.append(int(i['box2d']['x1']))
            obj.append(int(i['box2d']['y1']))
            obj.append(int(i['box2d']['x2']))
            obj.append(int(i['box2d']['y2']))
            obj.append(i['category'])
            objs.append(obj)
            obj = []
    # print("objs",objs)
    return name, objs
 

if __name__ == "__main__":

    parser = argparse.ArgumentParser(description='convert BDD100k label format to yolo lable format...')
    parser.add_argument('--label', '-l', dest='label_dir', 
                        help='path of label')
    parser.add_argument('--image', '-i', dest='image_dir',
                        help='path of image')
    parser.add_argument('--path', '-p', dest='path_dir',
                        help='path to save image path file')
    args = parser.parse_args()

    args.train_image_dir = os.path.join(args.image_dir, "train")
    args.val_image_dir = os.path.join(args.image_dir, "val")
    args.tarin_label_dir = os.path.join(args.label_dir, "bdd100k_labels_images_train.json")
    args.val_label_dir = os.path.join(args.label_dir, "bdd100k_labels_images_val.json")
    args.tarin_image_result = os.path.join(args.path_dir, "tarin.txt")
    args.val_image_result = os.path.join(args.path_dir, "val.txt")
    
    # 1. validation file convert
    f = open(args.val_label_dir)
    info = json.load(f)
    label2det(info, args, target="val")

    # 2. train file convert
    f = open(args.train_label_dir)
    info = json.load(f)
    label2det(info, args, target="train")

python bdd100k_to_yolo.py -l bdd100k\label -i bdd100k\image -p bdd100k