$ show Thinklab-SJTU/Bench2Drive
This repository provides the Bench2Drive benchmark for closed-loop end-to-end autonomous driving, accepted at NeurIPS 2024 Datasets and Benchmarks Track. It includes a large-scale multi-modal dataset (up to 7.3TB) collected in the CARLA simulator, an evaluation suite with metrics like Driving Score and Success Rate, and baseline results for over 10 methods. The codebase is a hybrid of research data pipeline and application: the dataset generation and evaluation scripts form the core data_research component, while the B2DVisualize Flask app, scenario runner, and leaderboard evaluator provide application-like interfaces for interaction and reproducibility. Data lineage: raw simulation data (sensor recordings, ground truth annotations) is collected via data_collect.py, transformed into compressed archives with JSON manifests, and evaluated using the leaderboard pipeline that computes multiple ability metrics. The repository answers research questions about the effectiveness of end-to-end driving policies under diverse urban scenarios.
The LLM identified 3 files worth highlighting for reuse:
Implements an offline occupancy label generation pipeline that densifies point clouds across frames, segments objects via Hungarian matching, and performs camera-ray visibility checks to produce voxel labels (free/occupied/unknown). This is a reusable data engineering asset for any autonomous driving dataset requiring occupancy ground truth from CARLA recordings. The pipeline is self-contained with configurable parameters (voxel size, range, frame buffer) and can be adapted to different sensor configurations.
Computes driving comfort and smoothness metrics (jerk, yaw rate, longitudinal/lateral accelerations) from logged simulation data. It reads control and state logs, computes derivatives, and outputs aggregated metrics. This script is a standalone, reusable component for evaluating driving quality in any CARLA-based autonomous driving dataset. It is well-structured, parameterized, and can be integrated into other evaluation pipelines.
Extracts HD map elements (lane markings, trigger volumes for traffic lights and stop signs) from a CARLA town and saves them as compressed numpy arrays. This script solves a recurring data preparation problem for autonomous driving datasets that require map information. It is modular, uses CARLA's map API, and outputs a standardized format that can be reused across different towns and datasets.