search

qiqihaer / RandLA-Net-pytorch

RandLA-Net's implementation with Pytorch
MIT License
127 stars 20 forks source link

About the performance of certain categories of SemanticKITTI #10

Open Zhen-ao opened 3 years ago

Zhen-ao commented 3 years ago

I found that the results I reproduced are similar to yours, but for the same problem, the result of "motorcyclist" is 0.00, and the result of the same "other-ground" is much lower than the paper. I don’t know the problem, I hope you can help me

qiqihaer commented 3 years ago

I found that the results I reproduced are similar to yours, but for the same problem, the result of "motorcyclist" is 0.00, and the result of the same "other-ground" is much lower than the paper. I don’t know the problem, I hope you can help me

There's no "motorcyclist" in the validation dataset, so the result is 0.00. I have no idea about the "other-ground" .