Closed haofeng0705 closed 9 months ago
Excellent work! I have a question about code details: regarding 3D HPE, during the training stage. normally, the input of the model is 2D pose coordinates that are lifted to 3D, but the code:
predicted_3d_pos = model_pos(batch_input) # (N, T, 17, 3)
Why is the dimension of batch input not (N, T, 17, 2)
update：it is known that the last dimension is confidence. What is the purpose of this？
Looking forward to your reply
please refer to #56.
Walter0807
commented
9 months ago
Excellent work! I have a question about code details: regarding 3D HPE, during the training stage. normally, the input of the model is 2D pose coordinates that are lifted to 3D, but the code:
predicted_3d_pos = model_pos(batch_input) # (N, T, 17, 3)
Why is the dimension of batch input not (N, T, 17, 2)
update：it is known that the last dimension is confidence. What is the purpose of this？
Looking forward to your reply