anibali/dsntnn

Use generated 2d-guassion heatmap as the regularization.

Closed this issue · 2 comments

I took a look at the paper, and curious about the regularization part, such as KL, JS..etc. I am wondering if it is possible to use 2d-guassion heatmap generated from ground truth as the regularization. What do you think of that?

That is essentially how my code works already. For the divergence-based regularisation terms (KL and JS), the divergence is calculated between the ideal, ground truth-based 2D Gaussian and the predicted heatmap.

thank you for your answers.