Qi Zou, Haibin Ling, Yu Pang, Yaping Huang, Mei Tian, Joint Headlight Pairing and Vehicle Tracking by weighted Set Packing in Nighttime Traffic Videos, IEEE Transaction on Intelligent Transportation Systems. 2017  PDF  

About Dataset

1.The dataset includes nine sequences. For each sequence, we provided the video (in jpg images), its groundtruth and our tracking results (reported in our paper in TITS2017). You can download all the sequences except s8. The sequence s8 was bought from a video provider and the copyright belongs to the provider. So for s8, we only provided a sample image, all the groundtruth and our tracking results.

2.Both the groundtruth and our tracking results are in text files. Every trajectory is saved in a separate text file. Each line in a text file records {frame No. , leftup x coordinate of the bounding box , leftup y coordinate of the bounding box, width, height of the bounding box } of a tracked object.

3.The videos are labelled using the tool ‘vbbLabeler’ developed by Piotr Dollar. If you want to visualize the groundtruth, you can also use the functions provided in piotr_toolbox. The groundtruth file got by ‘vbbLabeler’ is originally saved in ‘.vbb’ format, and then transformed into a text file. Both the vbb files and text files are provided in the groundtruth for downloading.

4.Region-of-interest is defined for each sequence, and is provided in the file ‘video’.

5.We use the CLEAR metrics (MOTA, MOTP, MR, FPR, IW) and the MATLAB code developed by H. Possegger for performance evaluation of multiple object tracking algorithms.

Sample frames

Contact Me

Name Qi Zou
Address Beijing Jiaotong University
School of Computer and Information Technology
No.3 Shangyuancun
Haidian District, Beijing
P. R. China 100044
E-Mail qzou@bjtu.edu.cn
© QiZou, 2017.