Pointnet_Pointnet2_pytorch icon indicating copy to clipboard operation
Pointnet_Pointnet2_pytorch copied to clipboard

Validation & testing result doesnt match

Open tapu1996 opened this issue 1 year ago • 2 comments

In a task, I am using exact same validation set as testing. During testing, I am using saved weights. However, every time I run, it provides different logit as output which result in performance mismatch between validation and test. What could be the reason? I have made sure that no augmentation is done during validation/testing. Pointcloud is exactly same as well.

tapu1996 avatar Jul 12 '24 07:07 tapu1996

Hi,bro.Did you solve the problem? I had the same problem.

GH-Fonic avatar Jan 03 '25 08:01 GH-Fonic

I found the cause of the problem was ScannetDatasetWholeScene class, ,I used class S3DISDataset for prediction and it was worked.

GH-Fonic avatar Jan 06 '25 06:01 GH-Fonic