VISION-SJTU/PillarNet-LTS

About the number of training epochs

Konstantin5389 opened this issue · 1 comments

why choose 36 epochs for training? Have you tried less epochs like 12 or 24? Does it influence much?

To fairly compare with the baseline CenterPoint-voxelnet, we simply adopt the pre-defined settings. You can try different hyper-parameters for testing.