Differences in training and evaluation
Hewentao123 opened this issue · comments
Sorry, I am a beginner. I have a problem that bothers me. During the training phase I see that there are two inputs: the image feature vector and the depth values of the 3D points. However, at test time, we only have a 2d image, and there is no corresponding depth value of the 3d point. How does it work? Looking forward to your reply. Thanks!