Alexandre Kirchmeyer

Hi Andrew, Yes that would be great, thank you! Sure no problem, I could try to complete the gradient versions once fprop is done, this way I'll see if I...

Maybe related: https://github.com/huggingface/diffusers/issues/1429#issuecomment-1328385498 I used to get "mosaic" images similar to what you describe while training a dreambooth model on SD v2, and this solved it for me.

Hi, thank you for your interest in our paper! The formula we chose for $P$ and $Q$ is: $P = \frac{H}{str}$ and $Q=\frac{W}{str}$ so in your notation each strided convolution...

1. Yes that's correct 2. That's right (although I would argue that it is exactly as written in the code but in this file: https://github.com/princeton-vl/Oriented1D/blob/master/models/dwoconv1d/dwoconv1d.py#L133) 3. Definition 1 determines exactly...

1. Yes I agree that my code is not very clear, I apologize for that, I should overhaul everything at some point. 2. That's great thanks for sharing! :) Out...

1. Ok I see that's interesting thank you for sharing! 3. Hmmm I'm not too sure what could be wrong. Can you try using focal loss instead (to help with...

Hi Peiqin, apologies for the delay. You are correct, the rounding formula leads to the plots that you show and it looks like the figure does not accurately represent this...

Hi Peiqin, I just sent you an email!