Is skip_spatial the result of F_L ?

[cnjanus]

I have a question about the local refinement unit.
In the essay, the author writed 'Next, to obtain the local point feature overFL, a common routine is to apply MLPs followed by a max-pooling along theK-dimension.' But what I find in code is that
skip_spatial = tf.reduce_max(grouped_feat, axis=[2]) skip_spatial = tf_util.conv1d(skip_spatial, mlp[-1], 1, padding='VALID', stride=1,bn=use_bn, is_training=is_training, scope='skip',bn_decay=bn_decay, weight_decay=weight_decay)
and grouped_feat = tf.add(grouped_feat, skip_spatial).
So I want to determine wheter skip_spatial is the result of F_L ?

[anshen666]

I have a question about the local refinement unit. In the essay, the author writed 'Next, to obtain the local point feature overFL, a common routine is to apply MLPs followed by a max-pooling along theK-dimension.' But what I find in code is that skip_spatial = tf.reduce_max(grouped_feat, axis=[2]) skip_spatial = tf_util.conv1d(skip_spatial, mlp[-1], 1, padding='VALID', stride=1,bn=use_bn, is_training=is_training, scope='skip',bn_decay=bn_decay, weight_decay=weight_decay) and grouped_feat = tf.add(grouped_feat, skip_spatial). So I want to determine wheter skip_spatial is the result of F_L ?

请问方便留个联系方式吗？这个项目一直没跑通，请问一定要在ubuntu环境下才可以跑通吗？