# wzzheng/TPVFormer

[CVPR 2023] An academic alternative to Tesla's occupancy network for autonomous driving.

Repository: https://github.com/wzzheng/TPVFormer
Canonical: https://ross.abutalabs.com/products/tpvformer
Homepage: https://wzzheng.net/TPVFormer/
Language: Python
License: Apache-2.0
License Family: permissive
Last push: 2024-09-07T04:25:37+00:00

## Health v2 (maintenance only)
Score: 32/100 (v2, computed 2026-09-02T17:46:02.011165+00:00)
- activity 0, release rhythm 35, longevity 98
- inputs: {"age_days": 1383, "days_push": 725, "days_rel": null, "gap_med": null, "n_releases_24m": 0}
- flags: no_releases
- formula: round(0.45*activity + 0.35*rhythm + 0.20*longevity); archived -> min(score, 10)

## Adoption (not part of the score)
Stars 1364, forks 127 (observed 2026-08-28T04:04:30.541750+00:00)

## What it is
TPVFormer is a CVPR 2023 research implementation of a tri-perspective view transformer for vision-based 3D semantic occupancy prediction, serving as an academic alternative to Tesla's occupancy network. It lifts surround-camera RGB features into 3D space using sparse LiDAR supervision to predict semantic occupancy for all voxels.

## Use cases
- predict 3d semantic occupancy from camera images
- research alternative to tesla occupancy network
- segment 3d voxels for autonomous driving scenes
- train occupancy prediction model on nuScenes
- run semantic scene completion on SemanticKITTI
- compare bev vs tri-perspective view perception

## When to choose
- you need a published, peer-reviewed baseline for camera-based 3D occupancy prediction
- you want to reproduce CVPR 2023 results on nuScenes or SemanticKITTI
- you are researching vision-centric autonomous driving perception without LiDAR at inference

## When to avoid
- you need a production-ready autonomous driving perception stack
- you require dense occupancy prediction rather than a tri-plane representation
- you need actively maintained code with frequent updates

## Facets
- artifact type: library
- maturity: maintenance
- function: machine-learning, deep-learning, computer-vision, image-processing
- domain: autonomous-vehicles, computer-vision, deep-learning, artificial-intelligence
- platform: python
- tags: 3d-occupancy-prediction, autonomous-driving, bev-perception, transformer, semantic-segmentation, nuscenes, cvpr-2023, research-code, linux, gpu

## Member repositories
- wzzheng/TPVFormer (main) score 32

## Provenance
- Observed fields: from GitHub, fetched 2026-08-28T04:04:30.541750+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 2026-08-30T04:41:22.886812+00:00, confidence not recorded.
  - readme: https://github.com/wzzheng/TPVFormer (fetched 2026-08-28T04:04:30.541750+00:00, sha c72550723aa1)
  - homepage: https://wzzheng.net/TPVFormer/ (fetched 2026-08-29T11:58:50.505131+00:00, sha 530625d91abc)
- Data as of 2026-08-30T08:39:29.467469+00:00.