About multi-view evaluation #32
Description
I noticed that in multi-view setup, gt 3d keypoints are repeated
pred_keypoints_3d = output['pred_keypoints_3d'].detach()
batch_size = pred_keypoints_3d.shape[0]
num_samples = pred_keypoints_3d.shape[1]
gt_keypoints_3d = batch['keypoints_3d'][:, :, :-1].unsqueeze(1).repeat(1, num_samples, 1, 1)
So I wonder that which coordinate frame are the gt_keypoints_3d in ? In the data processing, I believe gt_keypoints_3d are in camera coordinate frame, so why are they repeated for different views?