all the points are predicted to be the same class

[mai4567]

hi，thanks for your contribution！
when i using my own s3dis like dataset for training an Instance Segmentation model，i just got an error output：all the points are predicted to be the same class（background）。
here are my trainning logs and my dataset。can you give me some advice？

[Xiaoyang Wu]

Hi, could you provide a more detailed train.log?

[mai4567]

Hi, could you provide a more detailed train.log?

train.log

[Xiaoyang Wu]

Hi, I checked the log,

1. It seem that background should be treated as a ignore segment?
2. Also to ensure that trainining augmentation is aligned with validation alignment, maybe add an centershift after crop is better;
3. Some scale parameter for PointGroup should be adjust, but I am not an expert, just do more experiment.