tangshitao
tangshitao
It's MIT license. You can copy the codes.
Thanks for your interests. It’s MiT license. You can copy the codes. Regrads, Shitao. > 在 2022年7月12日，下午1:19，Dmytro Mishkin ***@***.***> 写道： > > > Hi, > > Could you please...
Can you post the error messages? Usually the problem is caused by cuda and pytorch version mismatch. The codes can be compiled under both cuda 10 or cuda 11. Please...
No. The codebase of stereo matching is messy.
把所有channel减少到1/2
1. k=8的显存消耗不可能比k=16大。 2. QuadTree-B∗ 就是这个repo里面release的。
1. K=16时候确实显存消耗会大一些， 你可以试试小一点的K。 2. 只是VIT不一样，Quadtree-B用的是原来loftr的结构，B*是ViT结构
Not exactly. There are 2 reasons: 1) we implement the quadtree attention with raw cuda without much optimization. We expect a speedup if implemented with torch.geometry. 2) The sparsity nature...
The code of notebook is from the original LoFTR. Please don't use that.
Yes, you can use fewer iterations. The one provided in configuration is not optimal.