【数据集】LaRa交通灯数据集解析

交通标志和信号灯数据集有TT100K(清华腾讯100K数据集)、中国交通标志检测数据集CCTSDB、 Bosch Small Traffic Lights Dataset(BSTLD).

交通信号灯检测系统的数据集有很多，其中包括：

1. LISA: Traffic Light Dataset：由美国加州大学伯克利分校的交通实验室（LISA）提供的数据集，包括交通信号灯的图像和视频，涵盖多种天气和时间条件。

2. Bosch Small Traffic Lights Dataset: 由德国博世公司提供的数据集，包括城市街道上的交通信号灯图像和视频，注重复杂的交通场景。

3. KAIST Multi-Spectral Pedestrian Detection Benchmark：由韩国科学技术院（KAIST）提供的多光谱数据集，其中包括红、黄、绿三种颜色的交通信号灯图像。

Lara转yolov5可用的数据集；

code： Lara2coco128；

# YOLOv5 🚀 by Ultralytics, GPL-3.0 license
# Lara dataset http://www.lara.prd.fr/benchmarks/trafficlightsrecognition
# Example usage: python train.py --data VisDrone.yaml
# parent
# ├── yolov5
# └── datasets
#     └── VisDrone  ← downloads here (2.3 GB)


# Train/val/test sets as 1) dir: path/to/imgs, 2) file: path/to/imgs.txt, or 3) list: [path/to/imgs1, path/to/imgs2, ..]
# path: /media/uuuuu/mobile_disk/workspace/zzz/dataset/traffic_light/Lara_UrbanSeq1_JPG # dataset root dir
# train: images  # train images (relative to 'path')   images
# val: images  # val images (relative to 'path')   images
# test: images  # test images (optional)   images
# 
# # Classes
# nc: 4  # number of classes
# names: ['go', 'stop', 'warning', 'ambiguous']
# # names: ['green', 'red', 'orange', 'ambiguous']

#Lara File format is as follows:
#Timestamp / frameindex x1 y1 x2 y2 id 'type' 'subtype'

import os
# from pathlib import Path
import shutil

def convert_box(size, box):
    # Convert Lara box to YOLO xywh box
    dw = 1. / size[0]
    dh = 1. / size[1]
    xc = ((box[0] + box[2])/2)*dw
    yc = ((box[1] + box[3])/2)*dh
    w = (box[2]-box[0])*dw
    h = (box[3]-box[1])*dh
    return xc, yc, w, h

def lara2yolo(path):
    # Path(dir / 'labels').rmdir()
    # Path(dir / 'images').rmdir()
    # (dir / 'labels').mkdir(parents=True, exist_ok=True)  # make labels directory
    # (dir / 'images').mkdir(parents=True, exist_ok=True)  # make images directory
    shutil.rmtree(os.path.join(path, 'labels'), ignore_errors=True)
    shutil.rmtree(os.path.join(path, 'images'), ignore_errors=True)
    os.mkdir(os.path.join(path, 'labels'))
    os.mkdir(os.path.join(path, 'images'))

    gt = open('Lara_UrbanSeq1_GroundTruth_GT.txt', 'rt')
    lines = gt.readlines()
    img_size = (640, 480) # (w, h)
    for i in range(13, len(lines)): 
        line = lines[i].split()
        index = line[2]
        imgname = 'frame_' + index.zfill(6) + '.jpg'
        # cls = int(line[7]) # id
        subcls = line[10]
        global cls
        if subcls == 'go':
            cls = 0
        elif subcls == 'stop':
            cls =1 
        elif subcls == 'warning':
            cls =2 
        elif subcls == 'ambiiguous':
            cls =3 
        else:
            cls =3 
        box = convert_box(img_size, tuple(map(int, line[3:7])))
        info = f"{cls} {' '.join(f'{x:.6f}' for x in box)}\n"
        labelname = './labels/' + imgname.replace('jpg', 'txt')
        imagepath = './Lara_UrbanSeq1_JPG/Lara3D_UrbanSeq1_JPG/' + imgname
        # copy images
        shutil.copy(imagepath, './images/')
        labelfile = open(labelname, 'a+')
        labelfile.write(info)
        labelfile.close()

if __name__ == "__main__":
    # Convert
    path = './dataset/traffic_light/Lara_UrbanSeq1_JPG'
    # dir = Path(path)
    lara2yolo(path)  # convert Lara_Urban annotations to YOLO labels

 

 

参考

1. Lara_dataset;

2. 【Datasets】LaRa交通灯数据集解析;

3. github_yolov5;

完