AdaLAM is a handcrafted outlier detection framework for local feature matching in computer vision pipelines like SfM and SLAM. It offers a robust and efficient alternative to deep learning methods for filtering outlier correspondences, targeting researchers and practitioners in computer vision.

How It Works

AdaLAM detects inliers by searching for significant local affine patterns within image correspondences. This handcrafted approach integrates several best practices into a single, efficient framework, proving competitive with deep learning methods.

Quick Start & Requirements

• Install via pip: pip install git+https://github.com/cavalli1234/AdaLAM
• Requires Python 3.7+, PyTorch, and tqdm.
• For best performance, a CUDA-enabled GPU is recommended, though CPU execution is supported.
• An example adalam.yml is provided for setting up a conda environment.
• See kornia integration: https://kornia.readthedocs.io/en/latest/feature.match.html

Highlighted Details

• Achieved second place in the Image Matching Challenge at CVPR 2020 (8000 keypoints category).
• Fully PyTorch implementation provided.
• Integrates with the kornia library (version 0.6.7+).
• Includes example scripts for COLMAP reconstruction and image pair matching.

Maintenance & Community

• The project is associated with ECCV 2020.
• Citation details are provided in the README.

Licensing & Compatibility

• The repository does not explicitly state a license.

Limitations & Caveats

• The README does not specify a license, which may impact commercial use or closed-source integration.
• Example scripts require additional dependencies like opencv-python-nonfree and a COLMAP installation.