question about need_norm and norm

[MengHao666]

Hi ,there are two things confusing me.

1. One is need_norm, what is it used for?
   if need_norm: dmap_norm = tf.norm(dmap_raw, axis=-1, keepdims=True) dmap = dmap_raw / tf.maximum(dmap_norm, 1e-6) else: dmap = dmap_raw

2. the other is in which you define the iknet:
   eps = np.finfo(np.float32).eps norm = tf.maximum(tf.norm(theta_raw, axis=-1, keepdims=True), eps) theta_pos = theta_raw / norm
   in this way, u believe the reference bone mentioned in your paper is the longest bone ,and u used its length to normalize bone vectors.Is that always true?

[CalciferZh]

1. It's merely length normalization.
2. I don't get your point. maximum is to avoid numerical issues.