[CLIFF] About J2d Full Projection

[Timkeeper2018]

Sorry to bother you, my own code level is not very good, and I have some questions about how to calculate the projection of 2D joint points
`def projection(pred_joints, pred_camera, center, scale, f_cliff, imge_shape):

# t full
pred_cam_t = cam_crop2full(pred_camera, center, scale, imge_shape, f_cliff)

batch_size = pred_joints.shape[0]
camera_center = torch.zeros(batch_size, 2)
pred_keypoints_2d = perspective_projection(pred_joints,
                                           rotation=torch.eye(3).unsqueeze(0).expand(batch_size, -1, -1).to(pred_joints.device),
                                           translation=pred_cam_t,
                                           focal_length=5000.,
                                           camera_center=camera_center)
# Normalize keypoints to [-1,1]
pred_keypoints_2d = pred_keypoints_2d / (256. / 2.)
return pred_keypoints_2d`

The perspective_projection function is consistent with that in SPIN
After obtaining t full crop, is the projection calculation method mentioned in the HMR with a focal length of 5000 used when calculating the projection? Is camera_center still (0,0)?

[zhihaolee]

• The focal length is not 5000 anymore, its calculation is described in the paper.
• The camera center is still (0, 0).

[124518]

@Timkeeper2018 Hello, sorry to bother you, how did your experiment work,whether it become better,it is my pleasure if you can tell me.

[wulalah]

• The focal length is not 5000 anymore, its calculation is described in the paper.
• The camera center is still (0, 0).

pred_keypoints_2d = pred_keypoints_2d / (256. / 2.)归一化这里应该不准确吧。不是应该除以不同数据集中图片的长宽的一半嘛

[wulalah]

• The focal length is not 5000 anymore, its calculation is described in the paper.
• The camera center is still (0, 0).

pred_keypoints_2d = pred_keypoints_2d / (256. / 2.)归一化这里应该不准确吧。不是应该除以不同数据集中图片的长宽的一半嘛