Daniel Mao

icon company Vision and Image Processing Research Group, University of Waterloo icon location Waterloo, Ontario, Canada icon location Hi. I love judy so so much.

Results 3 issues of Daniel Mao

How was "We train using batches with 2 source images and 2 mixed images for 250k iterations" achieved in the code?

1 comment

How was "We train using batches with 2 source images and 2 mixed images for 250k iterations" achieved in the code? I traced through the training code but did not...

Question about loss

According to the paper the loss L_gan should have `log` operations in the computation but I did not see any `log` function used in the training code in `main.py`?

Code for proposed method

I wasn't able to find code for the proposed method in the original paper, including random crop (Section 3.4.1), inpainting (Section 3.4.2), and copy-and-paste (Section 3.4.3). If these were publically...