End to end training

[UCASRMT]

Hi bardiadoosti, I noticed that the currently provided pre-trained model of GraphUNet only converts 2d point coordinates to 3d predictions, but I am more interested in the Resnet part, which converts RGB images to 2d point coordinates in the first place.
It will be greatly appreciated if you could add pretrained Resnet model to github, or the pretrained HOPE-net weight parameters.