In Hyperparameter says，beta is the initial value of Beta in Eq. 3 of the paper.
but in code, beta seem to be 'gamma' of absolute pose, and gamma seem to be 'gamma' of relative pose.
In train.py:

saq = section.getfloat('beta')
srq = section.getfloat('gamma')

In common/criterion.py:

#absolute pose loss
abs_loss =
torch.exp(-self.sax) * self.t_loss_fn(pred.view(-1, *s[2:])[:, :3],
targ.view(-1, *s[2:])[:, :3]) +
self.sax +
torch.exp(-self.saq) * self.q_loss_fn(pred.view(-1, *s[2:])[:, 3:],
targ.view(-1, *s[2:])[:, 3:]) +
self.saq
#VO loss
vo_loss =
torch.exp(-self.srx) * self.t_loss_fn(pred_vos.view(-1, *s[2:])[:, :3],
targ_vos.view(-1, *s[2:])[:, :3]) +
self.srx +
torch.exp(-self.srq) * self.q_loss_fn(pred_vos.view(-1, *s[2:])[:, 3:],
targ_vos.view(-1, *s[2:])[:, 3:]) +
self.srq