| license: apache-2.0 | |
| tags: | |
| - image-matching | |
| - computer-vision | |
| - feature-matching | |
| library_name: image-matching-models | |
| # XoFTR | |
| XoFTR is a cross-view feature matching transformer for image matching. | |
| ## Files | |
| - `xoftr_640.safetensors`: XoFTR model pretrained at 640px resolution | |
| - `xoftr_840.safetensors`: XoFTR model pretrained at 840px resolution | |
| ## Usage | |
| ```python | |
| from matching import get_matcher | |
| # Use 640 variant (default) | |
| matcher = get_matcher("xoftr", device="cuda", pretrained_size=640) | |
| # Or use 840 variant | |
| matcher = get_matcher("xoftr", device="cuda", pretrained_size=840) | |
| mkpts0, mkpts1, _, _, _, _ = matcher(img0, img1) | |
| ``` | |
| ## Links | |
| - [Original repository](https://github.com/OnderT/XoFTR) | |
| - [Image Matching Models](https://github.com/gmberton/image-matching-models) | |