Kent Gauen

Log Files from Training

1

Thank you for your awesome code! I am hoping you might open-source the log files you have from training. Maybe the training and validation loss as a function of epoch...

I want to use the gradient in a loss function, so I do not want to use "checkpoint" since I will use the backward pass twice. This code change simply...

Should the check be "greater than 0" rather than "greater than 1"? I would think this condition is to skip the edge, and "0" is the top-left. Maybe the "1"...

My question is about this code's associated paper [[linked here](https://www.cs.bgu.ac.il/~orenfr/BASS/Uziel_ICCV_2019.pdf)]. I am a bit confused with the initial GMM model. Equation 1 models the image $x$ with shape N x...

These are the changes I made to compile the program with CUDA 12.1, C++11, and using std=c++17 #19

This avoids a bit of panic when testing the compiled binary file. The "/" must be included.