# mit-han-lab/bevfusion

[ICRA'23] BEVFusion: Multi-Task Multi-Sensor Fusion with Unified Bird's-Eye View Representation

Repository: https://github.com/mit-han-lab/bevfusion
Canonical: https://ross.abutalabs.com/products/bevfusion
Homepage: https://bevfusion.mit.edu
Language: Python
License: Apache-2.0
License Family: permissive
Topics: 3d-perception, camera, lidar, object-detection, pytorch, semantic-segmentation, sensor-fusion
Archived: true
Last push: 2024-07-31T04:04:52+00:00

## Health v2 (maintenance only)
Score: 10/100 (v2, computed 2026-09-03T02:20:16.233290+00:00)
- activity 0, release rhythm 35, longevity 100
- inputs: {"age_days": 1558, "days_push": 763, "days_rel": null, "gap_med": null, "n_releases_24m": 0}
- flags: no_releases, archived
- formula: round(0.45*activity + 0.35*rhythm + 0.20*longevity); archived -> min(score, 10)

## Adoption (not part of the score)
Stars 3230, forks 614 (observed 2026-08-28T04:07:50.101955+00:00)

## What it is
BEVFusion is a PyTorch-based multi-task multi-sensor fusion framework that unifies camera and LiDAR features in a shared bird's-eye view representation for 3D perception. It achieves state-of-the-art results on nuScenes, Waymo, and Argoverse 3D object detection and BEV map segmentation benchmarks.

## Use cases
- fuse lidar and camera data for 3d object detection
- run bird's-eye view perception for autonomous driving
- benchmark multi-sensor fusion on nuScenes or Waymo
- train a 3d semantic segmentation model with camera and lidar
- deploy sensor fusion models to embedded hardware with TensorRT
- research multi-modal 3d perception architectures

## When to choose
- you need top-accuracy camera-LiDAR fusion for 3D detection or BEV segmentation
- you want a research-proven framework with leaderboard-winning results on nuScenes, Waymo, and Argoverse
- you need a task-agnostic BEV backbone supporting multiple 3D perception heads
- you plan to deploy on NVIDIA Jetson via the official TensorRT/CUDA solutions

## When to avoid
- you only have camera data with no LiDAR
- you need a lightweight production pipeline rather than a research codebase
- you lack GPU resources, since training requires significant compute
- you need active feature development - the repo is primarily a paper artifact with releases focused on maintenance

## Facets
- artifact type: library
- maturity: stable
- function: machine-learning, deep-learning, computer-vision, image-processing
- domain: autonomous-vehicles, computer-vision, deep-learning, robotics
- platform: python
- tags: sensor-fusion, lidar, camera, bird-eye-view, 3d-object-detection, semantic-segmentation, autonomous-driving, pytorch, nuscenes, research-code, gpu, linux

## Member repositories
- mit-han-lab/bevfusion (main) score 10

## Provenance
- Observed fields: from GitHub, fetched 2026-08-28T04:07:50.101955+00:00.
- Health v2: computed from the inputs above; adoption is never an input.
- Inferred fields (summary, facets, guidance): AI-extracted, prompt v1, taxonomy v1, on 2026-08-30T07:24:39.909502+00:00, confidence not recorded.
  - readme: https://github.com/mit-han-lab/bevfusion (fetched 2026-08-28T04:07:50.101955+00:00, sha f9cb652335f1)
  - homepage: https://bevfusion.mit.edu (fetched 2026-08-29T09:37:48.658047+00:00, sha f45ac15dc2bb)
  - site_page: https://hanlab.mit.edu/ (fetched 2026-08-29T09:37:48.667168+00:00, sha 4be659389a86)
- Data as of 2026-08-30T08:39:29.467469+00:00.