RoMa v2: Harder Better Faster Denser Feature Matching
What it is
RoMa v2 is a deep learning model for dense feature matching between image pairs, designed to be faster and denser than its predecessor. It uses a neural network to predict correspondences and overlap maps, which can be sampled to estimate geometric transformations like fundamental matrices.
The problem it solves
Previous feature matching models were either too slow for real-time applications or produced sparse matches that were insufficient for robust geometric estimation.
What you could use it for
- Estimating camera pose and fundamental matrices from image pairs
- Dense correspondence matching for visual localization
- 3D reconstruction and structure-from-motion pipelines
Read the source ↗Back to the library →
Summary written by an AI model from the source; check the source before you rely on it.