May I know how to sample the pixels which are used in the loss function?

[chaneyddtt]

Hi @jamt9000,
Thanks for the great work. I have one question about the implementation for pixel sampling. How to sample pixels from the original images, which will be used in the loss function? I am not quite understand these two lines of code in the implementation:

Could you please give some explanation about the function random_tps_weights()? What does 'weights1' represent and also the multiplication between 'self.F' and 'weights1'