Closed caozhenxiang-kouji closed 3 years ago
My bad. There's something wrong with 'mogrify' command so I changed the resize code to the OpenCV version. It seems that the resized images are a lot sharper than the ones you provide, which caused the difference in PSNR.
Thanks for the great work! I have a question about the poses.npy.
When I use the COLMAP to generate the poses of the dataset your provide, the PSNR of the results trained with my generated poses are lower than the PSNR of the results trained with the poses you provide. I ‘ve tried both v3.5 and 3.6 on Windows but get almost the same results.
The details are as follows:
Do you have any idea about how this come out?