henry123-boy/SpaTracker
SpatialTracker is a PyTorch implementation that tracks any 2D pixels in 3D space using depth estimation and multi-view geometric constraints, published at CVPR 2024.

SpatialTracker is a deep learning-based computer vision system that tracks arbitrary 2D pixels and reconstructs their 3D positions from video sequences. It leverages monocular depth estimators to obtain metric depth information and uses multi-view geometric constraints for 3D reconstruction and tracking. The official implementation provides inference code, pre-trained checkpoints, and supports dense tracking of thousands of points using a single GPU.
Frequently asked
- What is henry123-boy/SpaTracker?
- SpatialTracker is a PyTorch implementation that tracks any 2D pixels in 3D space using depth estimation and multi-view geometric constraints, published at CVPR 2024.
- Is SpaTracker open source?
- Yes — henry123-boy/SpaTracker is an open-source project tracked on heatdrop.
- What language is SpaTracker written in?
- henry123-boy/SpaTracker is primarily written in Python.
- How popular is SpaTracker?
- henry123-boy/SpaTracker has 1k stars on GitHub.
- Where can I find SpaTracker?
- henry123-boy/SpaTracker is on GitHub at https://github.com/henry123-boy/SpaTracker.