Pre-trained checkpoints

[agoodge]

Thanks for the repo,

Are there pre-trained model checkpoints available?

Thanks

[aldinorizaldy]

I'm training Hengshuang's model now, but only with ModelNet40 and S3DIS data. I'll let you know when it's done.

[Sottix99]

Thanks for the repo,

Are there pre-trained model checkpoints available?

Thanks

I am looking for the same thing, I even wrote to the authors of the paper but they didn't answer me

[aldinorizaldy]

Thanks for the repo,
Are there pre-trained model checkpoints available?
Thanks

I am looking for the same thing, I even wrote to the authors of the paper but they didn't answer me

Which model do you need? I think I have the pre-trained of Hengshuang's model on ModelNet.

Otherwise you can check the original implementation of Point Transformer and some other great models here: https://github.com/Pointcept

[Sottix99]

Thanks for the repo,
Are there pre-trained model checkpoints available?
Thanks

I am looking for the same thing, I even wrote to the authors of the paper but they didn't answer me

Which model do you need? I think I have the pre-trained of Hengshuang's model on ModelNet.

Otherwise you can check the original implementation of Point Transformer and some other great models here: https://github.com/Pointcept

i would need it for my thesis. since it is expensive to train such a model from scratch, if i had some chekpoints i could do finetuning from those. I would be very grateful if you could send me the pre-trained of Hengshuang's model

[aldinorizaldy]

Please download here:
https://drive.google.com/file/d/1WxzsRYmhXQDCc4ctLGeVUM0ReBOi9bIC/view?usp=sharing
I trained the model for 180 epochs (at that time I miscalculated the training time, so I only got about 180 epochs before my machine shutting down). The instance accuracy is 90.16.
The reported accuracy in this repo is 91.7 with 200 epochs. So, today I continued the training but could not achieve higher accuracy. Please let me know if this is sufficient for you. If not, I can give another try by training from scratch, hopefuly to get the similar accuracy.