Hand pose frame

[taewookim0812]

Hello

It seems the optimization function "solver.retarget()" accepts two inputs of hand_joint_seq and hand_frame_seq.
The hand_joint_seq can simply be acquired by 3rd-party libraries such as mediapipe, however, I wonder how did you get the hand_frame_seq, which represents the transformations between hand landmarks.

It would be appreciate if you let me know the way you get the hand_frame_seq, whether you get it from other libraries or any other implementation code is inside this repository.

Best regards

[yzqin]

Hello @taewookim0812,

hand_frame_seq refers to the sequence of hand poses derived from detectors that utilize the MANO framework. Since MediaPipe operates independently of the MANO/SMPL-X model, it doesn't provide such data.

If you're specifically interested in the retargeting aspect of DexMV, I suggest exploring our new retargeting library. It offers significantly enhanced performance compared to the previous retargeting module found in the DexMV repository.

[taewookim0812]

I appreciate your prompt response. I'll check the library