wang-jh18-SVM/Bench2DriveZoo

Bench2DriveZoo

Introduction

• This repo contains the training, open-loop evaluation, and closed-loop evaluation code for BEVFormer, UniAD , VAD in Bench2Drive.
• We merge multiple dependencies of UniAD and VAD including mmcv, mmseg, mmdet, and mmdet3d (v0.17.1) into a single library. As a result, it could support latest pytorch and advanced frameworks like deepspeed for acceleration.
• Use "git checkout tcp/admlp" to obtain their corresponding training and evaluation code.

Getting Started

• Installation
• Prepare Dataset
• Train and Open-Loop Eval
• Closed-Loop Eval in CARLA
• Convert Codes from Nuscenes to Bench2Drive

Results and Pre-trained Models

UniAD and VAD

Method	L2 (m) 2s	Driving Score	Success Rate(%)	Config	Download
UniAD-Tiny	0.80	32.00	9.54	config	Hugging Face/Baidu Cloud
UniAD-Base	0.73	37.72	9.54	config	Hugging Face/Baidu Cloud
VAD	0.91	39.42	10.00	config	Hugging Face/Baidu Cloud

BEVFormer

Method	mAP	NDS	Config	Download
BEVFormer-Tiny	0.37	0.43	config	Hugging Face/Baidu Cloud
BEVFormer-Base	0.63	0.67	config	Hugging Face/Baidu Cloud

Failure Cases Analysis

We provide some visualization videos and qualitatively analysis for TCP-traj, UniAD-Base, VAD-Base at here.

Related Resources

• Bench2Drive
• BEVFormer
• UniAD
• VAD