CASIA-Tencent Road Scene Dataset


The “CASIA-Tencent Road Scene Dataset” (RS10K) was built by the State Key Laboratory of Multimodal Artificial Intelligence Systems (MAIS), Institute of Automation of Chinese Academy of Sciences (CASIA), and T Lab, Tencent Map, Tencent Technology (Beijing) Co., Ltd. The images in our dataset were taken by onboard cameras of various vehicles from 31 cities in China. In each city, we randomly chose some sections of high-level roads (roads above national standard level 6) to collect images. RS10K is a comprehensive dataset, including annotations of various elements and relations, which can be used for road and lane segmentation, traffic sign detection, traffic sign understanding, and visual traffic knowledge graph generation. We hope RS10K can provide support for the development of the community. Due to some images involving sensitive information, they have been removed, so the publicly available dataset slightly differs from the statistical data published in paper [1].


The code used for visual traffic knowledge graph generation and evaluation


There are 10066 images in RS10K, of which 7041 are for training and 3000 are for testing. The element annotations include roads and lanes annotated by masks, signs and components annotated by quadrilateral bounding boxes. The relation annotations include S-S relations between signs, C-C relations between components, and A-T relations between the arrow element and traffic element. The statistics are shown in Table 1.

Tabel.1 Statistics of elements and relations.

All the defined symbols of the components are shown in Table 2.

Tabel.2 Categories of symbols in components.

The visual traffic knowledge graphs are composed of one or several knowledge trees, each of which is organized as Figure 1.

Figure 1. The structure of a knowledge tree.

Condition of Use

  • The CASIA-Tencent Chinese Traffic Sign Understanding Dataset, built by CASIA and Tencent, are released for academic research free of cost under an agreement.
  • Commercial use of the databases is subject to charge. For possible license of commercial use, please contact Fei Yin (
  • Reference

    RS10K was first used in the research work referred to as

          [1] Yunfei Guo, Fei Yin, Xiao-Hui Li, Xudong Yan, Tao Xue, Shuqi Mei and Cheng-Lin Liu. Visual traffic knowledge graph generation from scene images[C]//Proceedings of the IEEE/CVF International Conference on Computer Vision (ICCV). 2023: 21604-21613.


    Cheng-Lin Liu (, Fei Yin (

    National Laboratory of Pattern Recognition (NLPR)

    Institute of Automation of Chinese Academy of Sciences

    95 Zhongguancun East Road, Beijing 100190, P.R. China

    Contact Information

    Haidian | Beijing | China

    Phone : (+86-10)8254-4797

    Fax : (+86-10) 8254-4594