About Reproduce Results on S3DIS

[Rongtao-Xu]

Thank you for your awesome work! I did not change the original code, and my best mIOU reproduction result on S3DIS is 67.8, which is inconsistent with the official paper. Do I need to change the settings on the original code or change some hyperparameters, specifically which hyperparameters? Thanks a lot!

[chenlinchuang]

Thank you for your awesome work! I did not change the original code, and my best mIOU reproduction result on S3DIS is 67.8, which is inconsistent with the official paper. Do I need to change the settings on the original code or change some hyperparameters, specifically which hyperparameters? Thanks a lot!

Hello man, I'd like to ask if you are willing to share your current checkpoints (model_best.pth / model_last.pth) on this model. I'm currently struggling to train this model because of computing resource limitations. It would help me alot if I could do proper analysis on the pre-trained model. I'm sorry I'm not capable of answering your questions, I hope you could solve it soon. Thank you!