Open Starak-x opened 11 months ago
Thank you for your interest. We chose the best checkpoint based on the testing set result. This does lead to some concerns on fairness. However, we believe that our implementation follows the common practice of the point cloud completion evaluation, especially unpaired methods.
Hi Ruikai, I am very interesting in your work. But I find that the ground truth is used for choose the best checkpoints. I wonder is this allowed in self-supervised methods?