Hello,
I am trying reimplementing some popular methods dealing with semantic segmentation for point cloud such as PointNet and PointNet++. But my metrics has some distance between your baseline metrics. Is it possible to share some training setting details? Some hyperparameters such as weight for cross entropy seems to be quite tricky. Thanks!
Hello, I am trying reimplementing some popular methods dealing with semantic segmentation for point cloud such as PointNet and PointNet++. But my metrics has some distance between your baseline metrics. Is it possible to share some training setting details? Some hyperparameters such as weight for cross entropy seems to be quite tricky. Thanks!