Performance for outdoor dataset

[bqm1111]

I wonder what would it take differently for this method to work on outdoor dataset like semanticKITTI? I tried to retrain on semanticKITTI dataset and get IoU of 22.7% and mIoU of 2.3%. Could you provide some guidance to transfer this work to outdoor dataset?

[JuIvyy]

Thank you for your interest, currently we do not have plan on implementing legoocc on semanticKITTI and have no experience on this.

[bqm1111]

Could you tell me your view on what might be the problem when applying your work on outdoor dataset? What might be challenges?

[JuIvyy]

Sorry, I am not very familiar with the SemanticKITTI dataset, so I’m afraid I cannot provide reliable or useful guidance at the moment.

I plan to explore the adaptation of LegoOcc to outdoor datasets in the future. Once I have conducted some experiments and gained a better understanding of the relevant issues, I will share my findings and thoughts here.

Thanks for your understanding!

[bqm1111]

Thank you for your response. It would be great if I could hear back from you in the future. Truly very nice work.

[bqm1111]

Hi. Could you comment on the generalization capability of this work on unseen indoor datasets?

[JuIvyy]

freeocc studies generalization capability and it's open sourced, you can take a look.