Does it generalize to unseen in-the-wild point clouds?

Question

Does it generalize to unseen in-the-wild point clouds?

coreqode opened this issue 3 years ago · 4 comments

Hi, this is amazing work.
Does the pre-trained model generalise to in-the-wild point clouds? or it is specific to trained datasets only?

Answer 1 · 2021-12-03T04:42:33.000Z

It generalizes well to real-world kinect style scanner based point clouds.

Answer 2 · 2021-12-05T05:24:21.000Z

Is there any codebase for testing the pre-trained models in in-the-wild scenarios?

Answer 3 · 2021-12-06T16:06:04.000Z

This can be achieved by changing the dataset using your custom sensor data, then run the testing.

Answer 4 · 2023-03-22T07:12:22.000Z

@rabbityl when I applied to my own data, I got flow all zeros. R and t are identity and zeros. The confidence matrix has all very small values. What could be the problem?