how to use joint angle limit

[yfynb1111]

Thanks to your shareing!
I see the defined joint angle limit,but cant find where you use the joint angle limit.could you share how to use it？

[gmntu]

The joint angle limit self.alim_ was used during the training of the network to regularize the joint angle parameter (alpha)

where L and U are the lower and upper bound of the limit respectively.

For example:

def compute_ang_limit(self, ang):
    # ReLU(x)=max(0,x)
    min_ = self.ReLU(self.alim_[:,0]-ang)
    max_ = self.ReLU(ang-self.alim_[:,1])

    return torch.sum(min_) + torch.sum(max_)

[yfynb1111]

The joint angle limit self.alim_ was used during the training of the network to regularize the joint angle parameter (alpha)

where L and U are the lower and upper bound of the limit respectively.

For example:

def compute_ang_limit(self, ang):
    # ReLU(x)=max(0,x)
    min_ = self.ReLU(self.alim_[:,0]-ang)
    max_ = self.ReLU(ang-self.alim_[:,1])

    return torch.sum(min_) + torch.sum(max_)

Thank you!! Your change from pca to Dof is a excellent idea！now mano model is easier to use.

umm, I try to use optimization based method to estimate 3D hand pose on the dof model，how do you about it？