The initial layer for the rotation classifier (directly applied to z_rot) is RELU -- however as noted in #1 (review) this may not make a lot of sense, since it's just cutting off the negative portion of the gaussian.
Justification or more experiments are needed here.
The initial layer for the rotation classifier (directly applied to z_rot) is RELU -- however as noted in #1 (review) this may not make a lot of sense, since it's just cutting off the negative portion of the gaussian.
Justification or more experiments are needed here.