BC04: Data Augm (Scooping/MMOneSphere)
After CoRL, trying more data augmentation using the 50 demo case.
Created on July 13|Last edited on July 17
Comment
I am testing data augmentation. The test example point clouds do not have any noise added. It's just for training (and a new randomly sampled noise each time).
Overall
As with pouring we are not seeing much benefit with data augmentation (thus far).
Well, we do get some with training longer but maybe that's just due to training longer?With Gauss 0.05 and training longer we max out at 0.664 / 0.632 = 1.051 which is better than the 0.975 we reported which trains for half as many epochs. I won't read too much from these results.
Results
Transl + Gauss 0.01
5
Gauss 0.01
5
No Data Aug
5
Gauss 0.05
5
Run set
10
Add a comment