You signed in with another tab or window. Reload to refresh your session.You signed out in another tab or window. Reload to refresh your session.You switched accounts on another tab or window. Reload to refresh your session.Dismiss alert
No, the data distributions of these two datasets are very different. If you do this, the result will get a big performance loss. You should retrain this model on KITTI.
@hyangwinter could you explain the reason why the generalization of this model is not good for kitti? In the flownet3d paper, they seem to have a well generalization on the KITTI even without re-train. Is it due to your implementation?
Well, I have not tested this model on the KITTI dataset. If you are interested, you can refer to the official implementation to replace FlyingThings3D dataloader with KITTI dataloader.
Can pertained model be applied to kitti dataset?
The text was updated successfully, but these errors were encountered: