Johan Edstedt
Johan Edstedt
> Thanks for the quick response! Also, could you please answer the following questions: > > 1. Are `x_A` and `x_B` lists of key points that are not currently in...
The 2W comes from symmetric warp, :W is from A to B and W: is from B to A. Dense matchers are typically asymmetrical, so getting the symmetric warp requires...
The implementation for keypoints currently assumes asymmetrical, but could be made symmetric. I dont think its obvious what the best way to do symmetric kpt matching with dense warps is...
This is partially supported, see https://github.com/Parskatt/RoMa/issues/24 for additional details.
Hi. The dense warp we compute is unidirectional, and therefore results are asymmetric depnwnding on order of A,B. We can improve results somewhat by running the decoding and refinement twice...
Yes, or rather the refinement stage is repeated but in higher resolution. The is the reason for "upsample_res"
Thats ok, it will fall back to standard pytorch.
It seems that the fp16 implementation for layernorm doesnt exist in your pytorch. Also, are you running on cpu? Could you tell me what torch version youre using?
Sorry don't have it. I'll try training one (also with vit base).
@Dawars I'm not convinced that FeatUp is useful. Would be glad to be proven wrong, but not something I'll spend time on currently.