Heesoo Yang

Results 4 comments of Heesoo Yang

The meaning of ins_labels and sem_labels

> (1) 预测的instance 数量和gt数量差异大 是因为 blockmerge算法并不能准确的把blocks合并到一起，产生了很多细碎的instances，这一步处理基本上是个bottleneck。当然理想情况是网络可以一次性把一个房间全部test了，直接吐出所有instance， 但当时用的pointnet++ backbone并不能处理大规模点云， 需要新的backbone才行，个人认为这是一个很好future work。 > (2) 如果remove的细碎instances 需要visualize，就随机生成一些颜色保存。 @Yang7879 Hello, I can't read Chinese, but I think I got the gist of the above...

Training with own data

Hi @llijiang, I read in your paper that you tested the model on the S3DIS dataset. Do you have a dataloader and config file for S3DIS that you could provide,...

Make `ir` module public

@emilio are you the right person to tag here? I put up a PR to expose the ir module under a feature flag. I also had a similar use case...

Add `unstable_ir` feature flag that makes the ir pub

> I guess this is reasonable to do, but I don't want to make any change to the ir a breaking change. > > Let's call this `unstable_ir` and document...