Model Details

Important

Check the LICENSE of each model/original code base before use in your application. Some are heavily restricted.

Most models can run on both CPU and GPU. If a runtime shows ❌, that model cannot run on that device. MPS (Apple Silicon) is not tested.

Model Code Paper GPU Runtime (s/img) CPU Runtime (s/img)
2026
LoMa-R (CVPR IMW '26) Official arxiv
LoMa (arXiv '26) Official arxiv
2025
RoMa v2 (arXiv '25) Official arxiv
UFM (NeurIPS '25) Official arxiv
RIPE (ICCV '25) Official arxiv 2.772
RDD (CVPR '25) Official Project page 6.73 / 7.85 / 9.18 / 15.75
EDM (arXiv '25) Official arxiv 0.493
LiftFeat (ICRA '25) Official arxiv 0.514 2.577
MINIMA (CVPR '25) Official arxiv 11.32 / 7.355 / 96.6 / 0.561
2024
XoFTR (CVPR IMW '24) Official pdf 8.954
Affine Steerers (ECCV '24) Official arxiv 0.677
Keypt2Subpx* (ECCV '24) Official arxiv 0.055 /0.164 / 0.033 / 0.291 --
MASt3R (ECCV '24) Official arxiv 0.699 22.27
Efficient-LoFTR (CVPR '24) Official pdf 0.1026 2.117
OmniGlue (CVPR '24) Official arxiv 6.351
xFeat-Steerers Official Discussion 0.037 0.124
xFeat (CVPR '24) Official arxiv 0.027 0.048
GIM* (ICLR '24) Official arxiv 0.077 / 1.627 5.321 / 20.301
RoMa / Tiny-RoMa (CVPR '24) Official arxiv 0.453 / 0.0456 18.950
DUSt3R (CVPR '24) Official arxiv 3.639 26.813
DeDoDe (3DV '24) Official arxiv 0.311 (+MNN)/ 0.218 (+LG)
Steerers (CVPR '24) Official arxiv 0.150 26.425
2023
TopicFM[+] (AAAI '23) Official arxiv
ZippyPoint (CVPR '23) Official pdf
SphereGlue* (CVPRW '23) Official pdf 0.548 / 0.070 0.804 / 7.407
LightGlue* (ICCV '23) Official arxiv 0.417 / 0.093 / 0.184 / 0.128 2.828 / 8.852 / 8.100 / 8.128
2022
SE2-LoFTR (CVPRW '22) Official arxiv 0.133 2.378
Aspanformer (ECCV '22) Official arxiv 0.384 11.73
Matchformer (ACCV '22) Official arxiv 0.232 6.101
2021
LoFTR (CVPR '21) Official / Kornia arxiv 0.722 2.36
Patch2Pix (CVPR '21) Official / IMT arxiv 0.145 4.97
2020
LISRD (ECCV '20) Official arxiv 12.319 / 3.823/ 12.305 0.187 / 1.191 / 0.165
SuperGlue (CVPR '20) Official / IMT arxiv 0.0894 2.178
2019
R2D2 (NeurIPS '19) Official / IMT arxiv 0.429 6.79
D2Net (CVPR '19) Official / IMT arxiv 0.600 1.324
pre 2019
DoGHardNet (NeurIPS '17) IMT / Kornia arxiv 2.697 (+NN) / 0.526 (+LG) 2.438(+NN) / 4.528 (+LG)
ORB-NN (ICCV '11) OpenCV ResearchGate 0.088 0.092
SIFT-NN (IJCV '04) OpenCV pdf 0.124 0.117


Our implementation of Patch2Pix (+ Patch2PixSuperGlue), R2D2, and D2Net are based on the Image Matching Toolbox (IMT). LoFTR and DeDoDe-Lightglue are from Kornia. Other models are based on the offical repos above.

Runtime benchmark is the average of 5 iterations over the 5 pairs of examples in the assets/example_pairs folder at image size 512x512. Benchmark is done using vismatch_test.py on an NVIDIA RTX A4000 GPU. Results rounded to the hundredths place.

  • SphereGlue requires torch-geometric and torch-cluster. Install with: pip install .[all] -f https://data.pyg.org/whl/torch-2.5.0+cu124.html (replace cu124 with cpu for CPU). See PyG installation docs.

  • SphereGlue model runtimes are listed in the order: SIFT, SuperPoint

  • GIM model runtimes are listed in the order: LG, DKMv3

  • LightGlue model runtimes are listed in the order: SIFT, SuperPoint, Disk, ALIKED

  • Keypt2Subpx model runtimes are listed in the order: superpoint-lightglue, aliked-lightglue, xfeat, dedode

  • MINIMA model runtimes are listed in the order: superpoint-lightglue, LoFTR, RoMa (large), RoMa (tiny)

  • RDD model runtimes are listed in the order: sparse, star (semi-dense), lightglue, aliked