search

MengHao666 / Hand-BMC-pytorch

PyTorch reimplementation of hand-biomechanical-constraints (ECCV2020)
MIT License
72 stars 10 forks source link

Normalizing bone length by Reference Bone Length #4

Closed thiagomcoutinho closed 3 years ago

thiagomcoutinho commented 3 years ago

Hi @MengHao666 , first of all thank you for your implementation.

I have a doubt with relation to the reference bone normalization in the BMCLoss Class. When you compute the min and max values for the bone lengths in the calculate_bmc.py code, all bones are normalized by the reference bone, as shown below:

https://github.com/MengHao666/Hand-BMC-pytorch/blob/d9d1a5394ebc84079e9fc25885122c18a9ec1505/calculate_bmc.py#L50-L57

But in the BMCLoss Class this step is not made, which could generate wrong results if the input data isn't normalized. Is this right?

https://github.com/MengHao666/Hand-BMC-pytorch/blob/d9d1a5394ebc84079e9fc25885122c18a9ec1505/weakloss.py#L253-L278

Also, did you think what is the easiest way to use the algorithm for the left hand?

Thanks in advance.

MengHao666 commented 3 years ago

h relation to the reference bone normalization in the BMCLoss Class. When you compute the min and max values for the bone lengths in the calculate_bmc.py code, all bones are normalized by the reference bone, as shown below:

Sure. But as we have strong supervision of 3D groundtruth joint location, I guess this effect is very small. You might do an ablation study to see further.