Dmytro Mishkin

Yes, I did: ```python3 def get_lr_scheduler(optimizer, conf): """Get lr scheduler specified by conf.train.lr_schedule.""" if conf.type == 'OneCycleLR': max_lr = conf.max_lr epochs = conf.epochs steps_per_epoch = conf.steps_per_epoch max_lr = conf.max_lr del...

Thank you for review, I will make changes shortly, where possible. Regarding the "libpng warning: iCCP: known incorrect sRGB profile" - that is an issue with the images themselves, I...

Sounds good, I'll proceed with it

@Phil26AT PR is ready :) https://github.com/cvg/glue-factory/pull/52

There is no way LightGlue is so slow, it usually takes like 0.02sec per 1024 points, unless you run on CPU (even though) or do the compilation.

What is your GPU then? And how exactly do you measure time? Could you please put full code snippet here?

It seems you don't turn the gradients off. How about following: ```python3 torch.cuda.synchronize() with torch.inference_mode(): feats0 = extractor.extract(image0) # auto-resize the image, disable with resize=None feats1 = extractor.extract(image1) matches01 =...

@tcourat are you sure that your machine doesn't do any background process at the time of bechmarking?

@YJonmo or you can use matching interface from the kornia: https://kornia.readthedocs.io/en/latest/feature.html#kornia.feature.LightGlueMatcher

@YJonmo I wonder why would you use `DISK_epipolar`. Is there any task where it is better than depth-based?