Should the image and point cloud branches be trained with the training set only or with the training set and validation set together?
Should the image and point cloud branches be trained with the training set only or with the training set and validation set together?