Cylinder3D
Cylinder3D copied to clipboard
xinge008
Rank 1st in the leaderboard of SemanticKITTI semantic segmentation (both single-scan and multi-scan) (Nov. 2020) (CVPR2021 Oral)
I don't know what happened. There is no such problem on the Internet. Can someone help me？
Is there a C++ version of this project, or is it possible to implement it in C++ in the future?
We modified the point cloud data format of the Nuscenes dataset to the point cloud data format of SemanticKitti. Then we used the pre-trained weights provided by the author for...
I observed that segmentation prediction label must be between 1-16, but you predict 0-16 in nuscenes yaml. So how do you process label 0?
I have trained my model using DDP, but I am unable to achieve the mentioned mIoU 76.1 on the NuScenes dataset. I am wondering if using DDP has negatively affected...
Will it be supported in the future ？ current single card training cost too much time
Hello, first of all thank you for your work, I looked at the code you posted and did not find the weights you set for each category in the weighted...
Hi, I was trying to re-train on SemanticKITTI and I get the following error during validation. Please help in fixing this. Traceback (most recent call last): File "train_cylinder_asym.py", line 167,...
