hustvl/MapTR
End-to-end transformer-based framework for constructing high-definition vectorized maps from camera inputs for autonomous driving.
Not currently ranked — collecting fresh signals.
star history
MapTR proposes a structured modeling and learning approach for online HD map construction using bird’s eye view (BEV) perception with transformer architectures. It detects and localizes road elements (lanes, boundaries, intersections) from multi-view camera inputs and outputs them as structured vectorized representations. The framework supports real-time inference and has been extended to MapTRv2 with improved performance.
Frequently asked
- What is hustvl/MapTR?
- End-to-end transformer-based framework for constructing high-definition vectorized maps from camera inputs for autonomous driving.
- Is MapTR open source?
- Yes — hustvl/MapTR is open source, released under the MIT license.
- What language is MapTR written in?
- hustvl/MapTR is primarily written in Python.
- How popular is MapTR?
- hustvl/MapTR has 1.5k stars on GitHub.
- Where can I find MapTR?
- hustvl/MapTR is on GitHub at https://github.com/hustvl/MapTR.