search

ldkong1205 / PointCloud-C

Benchmarking and Analyzing Point Cloud Perception Robustness under Corruptions
https://pointcloud-c.github.io/home
164 stars 22 forks source link

about the code of WOLFMix #6

Closed Roywangj closed 2 years ago

Roywangj commented 2 years ago

Hi @jiawei-ren , i have searched the whole file for codes of WOLFMix but failed, i only found the codes of RSMix and PointWOLF. Would you please release the code of WOLFMix and update corresponding training codes, i think this will be great value to helping understanding the workflow.

Roywangj commented 2 years ago

For the deformation step, we use the default hyper-parameters in PointWOLF (Kim et al., 2021). We set the number of anchors to 4, sampling method to farthest point sampling, kernel bandwidth to 0.5, maximum local rotation range to 10 degrees, maximum local scaling to 3, and maximum local translation to 0.25. AugTune proposed along with PointWOLF is not used in training. For the mixing step, we use the default hyper-parameters in RSMix (Lee et al., 2021). We set RSMix probability to 0.5, β to 1.0, and the maximum number of point modifications to 512. For training, the number of neighbors in k-NN is reduced to 20 and the number of epochs is increased to 500 for all methods.

Even details are provided on paper, it's still confused to follow WOLFMix.

Thanks

jiawei-ren commented 2 years ago

The GDANet sub-directroy includes the implementation of WOLFMix.

To train GDANet+ WOLFMix, please use the following command: python main_cls.py --epochs 500 --beta 1.0 --rdscale --shift --pw

Roywangj commented 2 years ago

hi @jiawei-ren ， i have added PointWOLF to my method. However, i want to reproduce WOLFMix but failed. I want to know more details about it, would you release the codes of WOLFMix ?

jiawei-ren commented 2 years ago

The "PointWOLF" should be "WOLFMix" in the previous comment. Sorry for the typo.

Roywangj commented 2 years ago

hi, @jiawei-ren. Fine, thanks for your patient relpy.