search

hansen7 / OcCo

[ICCV '21] "Unsupervised Point Cloud Pre-training via Occlusion Completion"
https://hansen7.github.io/OcCo/
MIT License
228 stars 26 forks source link

Can't reproduce Linear SVM result with DGCNN backbone. #21

Closed JasonNie96 closed 2 years ago

JasonNie96 commented 2 years ago

Hi Hansen,

Thanks for your work!

I'm trying to reproduce your result in Table 10. It said with the DGCNN backbone, OcCo has 89.2% accuracy. But I loaded your pre-trained model: dgcnn_occo_cls.pth and train the svm with the following command:

python train_svm.py --gpu=0 --model=dgcnn_util --dataset=modelnet40 --restore_path=./data/pretrained/dgcnn_occo_cls.pth.

But it only gets 88.4% accuracy. May I ask did I do anything wrong?

hansen7 commented 2 years ago

Hi, @JasonNie96 . Thanks for asking.

The provided checkpoints are mainly for classification tasks (with MLP output heads), for the results reported in the appendix, I have rerun the pre-training with different configurations many times, and reported the best results I have in hand.

JasonNie96 commented 2 years ago

Hi Hansen,

Thanks for your quick reply!

Got it. Is that possible to provide the best result checkpoint? That would be really helpful! If not, I totally understand.