search

HyeongminLEE / AdaCoF-pytorch

Official source code for our paper "AdaCoF: Adaptive Collaboration of Flows for Video Frame Interpolation" (CVPR 2020)
MIT License
172 stars 27 forks source link

Loss function question #5

Closed issakh closed 3 years ago

issakh commented 4 years ago

Hi, the default loss function is set to 1Charb+0.01g_Spatial+0.005*g_Occlusion'. Charb is the Charbonnier loss, but I can't seem to figure out what g_spatial and g_occlusion are based on the code? For the adversarial loss, of the options stated (GAN, WGAN, WGAN_GP, T_WGAN_GP, FI_GAN) which was the one used to get the best results in the paper? Thanks for your help

HyeongminLEE commented 4 years ago

  1. g_spatial and g_occlusion is the smoothness constraint terms for flow maps and occlusion maps. However, since these terms did not significantly change the result, we removed these terms from the paper. You can freely remove these terms.

  2. FI_GAN is the proposed version of the paper.

Thanks.