2018 Driving Policy Prediction Task

1. Overview

The DBNet 2018 Driving Policy Task is intended to push the state of the art in autonomous driving forward. More specifically, DBNet mainly focuses on driving policy learning and hence is largely different from the previous ones (KITTI, Cityscapes, BDDV, etc.). Up to now, DBNet has collected three categories of data by high-sensitive sensors, including point clouds, videos and drivers' behaviors. Our data are acquired under various traffic conditions and multiple professional drivers. For full details of this task, please see the policy prediction evaluation page.

The DBNet train, validation, and test sets, containing more than 200 KM real-world driving data and are available at the download page. More driving data will be released continuously and we will keep iterating this promising and challengeable task.

Note: the policy prediction challenge will be open soon!

2. Dates (not deterministic now)

December 15, 2018 Submission deadline (11:59 UTF+8h)

December 29, 2018 Challenge winners notified

3. Task Guidelines

Participants are recommended to train their algorithms on train/validation sets. The detailed guidelines, including data, description, format and submitting guidelines, will be public soon.

4. Tools and Instructions

We provide a demo for the utilization of images and point clouds. For more details of our models, please check our CVPR paper or visit our GitHub repository. Due to the large size of DBNet and the complexity of this task, the process of provided data may not be simple, which requires many hours. In consequence, we provide prepared data and some tools for the convenience of participants. At the same time, we also provide point clouds of different format (pcd or las) without sampling. To avoid restricting the imagination of participants, the original point clouds are also provided (frame-level).
We do hope the community could contribute their precious methods and help us improve DBNet.

5. Organizers

Yiping Chen
Jingkang Wang
Zhipeng Luo
Cheng Wang

6. Committee

Jonathan Li
Cewu Lu