This is the code of homography estimation model for my semester project at the ETH-Zurich. HomoRAFT is a deep learning model used to estimate homography between two images. The main structure is based on the RAFT.
If training with Synthetic MS-COCO dataset, please install the latest version of OpenCV, PyTorch and Kornia.
To train the model, you need to first download the MS-COCO dataset, then change the config file to point to the correct path. Then, you can run the following command to train the model:
python train.py --config_file configs/raft-easy.yamlTo evaluate the model, first download the HPatch, then run the following command:
python evaluation/evaluate_hpatch.py --image-data-path <path-to-hpatch> --ckpt <path-to-model>Weight can be downloaded from here