Thien Tran

I notice this also. timm only applies SE in the last block. I didn't see the papers mentioned any of these. https://github.com/rwightman/pytorch-image-models/blob/07eb2f12005f75be3ed6c2394f3512e7a8ac640a/timm/models/vovnet.py#L244-L256 Have you figured out whether it is a...

Seems like SE is only applied to stage 2 and 3, since they have only 1 block? Still, it is pretty strange

I can't guarantee that the model trained here can be used in the original centernet repo, or any repo. I haven't extensively tested onnx export, so nothing is guaranteed, again....

as mentioned in #2, I started a refactoring work on this repo but never finished. The backbones were moved to a separate repo [gau-nernst/vision-toolbox](https://github.com/gau-nernst/vision-toolbox). To use a fully working repo...

The inputs and targets should have the same shape, since quality focal loss treats the inputs and targets like binary cross entropy does. In particular, it is for multi-label classification.

The shapes for inputs (model outputs) and targets for Modified Focal Loss (used in CenterNet) and Quality Focal Loss are the same. They both must have the same shape. CenterNet...

Hello, I saw the "Make kornia fully-runnable on Apple Silicon chips" project on Google Summer of Code page and I'm interested to help. May I know how I can help...

I was looking for bilateral filter and saw this issue and the unfinished PR. I took a try, following ideas from sunny2109 implementations. ```python import torch import torch.nn.functional as F...

Input is 512x512x3. CPU is Intel W-2295, GPU is RTX3090. Time measured to process 1 image. For batch 32, time is divided by 32. Implementation | kernel_size=3 | kernel_size=5 |...

I just added speed test with `torch.compile()`. It's unbelievable! @ducha-aiki Yes I agree, correctness first before speed optimization.