RoMa
RoMa copied to clipboard
Parskatt
[CVPR 2024] RoMa: Robust Dense Feature Matching; RoMa is the robust dense feature matcher capable of estimating pixel-dense warps and reliable certainties for almost any image pair.
``` from roma import roma_indoor xFormers not available xFormers not available ``` I get this error message.
Matching 2 images with 20000 matches works with commit https://github.com/Parskatt/RoMa/commit/69cefb130b3ec7586da792ed897adf6e801acb1a and doesn't with current`main`. I guess that is because of removing ``` torch.backends.cuda.matmul.allow_tf32 = True # allow tf32 on matmul...
Hello! I try to training roma myself, i wonder if you can upload your training plot. In addition the training process contains several loss(delta_regression_loss_1,delta_certainty_loss_16, delta_certainty_loss_4,gm_cls_loss_16..) and i am not sure...
hello! I wonder if there is additional version of dino (for example ViT-S/14) that can be used as backbone in this model
batch size
hello i noticed that when i used different batch size the certainty value is changing did you any idea why?
Hello, your results are very promising! I wonder if you've tried other backbones instead of DinoV2. Indeed, I wonder how much of the good results you're getting come from the...
Adding this feature to the library would benefit users by enabling them to preprocess images before passing them directly to the method.
