hustvl/MapTR

Health v2 · maintenance only

Adoption not part of the score

1558 stars · 251 forks observed · 2026-10-17

What it is AI-extracted, prompt v1, taxonomy v1, 2026-08-30, confidence not recorded

MapTR is an end-to-end transformer-based framework for online vectorized HD map construction from camera imagery in autonomous driving. It performs structured modeling and learning of map elements in bird's-eye view in real time, with an extended MapTRv2 variant.

Use cases

• build vectorized HD maps online from camera sensors
• research bird's-eye-view map perception for autonomous driving
• generate map annotations automatically from driving data
• benchmark real-time HD map construction on nuScenes
• provide map priors for downstream motion planning models

When to choose

• you need state-of-the-art online vectorized HD map construction from multi-view cameras
• you are doing academic research on BEV perception or end-to-end driving
• you want a real-time map head to plug into an autonomous driving stack

When to avoid

• you need production-grade mapping with lidar fusion and map maintenance workflows
• you lack GPU resources for training transformer-based perception models
• you need a turnkey commercial HD mapping product rather than research code

Facets

library · maturity active

machine-learning deep-learning computer-vision image-processing autonomous-vehicles computer-vision deep-learning machine-learning python hd-map-construction bev-perception transformer end-to-end vectorized-maps autonomous-driving research-code iclr2023 linux gpu

Member repositories

For agents

markdown · JSON · MCP: product_card(name="hustvl/MapTR")