search

facebookresearch / ContrastiveSceneContexts

Code for CVPR 2021 oral paper "Exploring Data-Efficient 3D Scene Understanding with Contrastive Scene Contexts"
MIT License
218 stars 29 forks source link

I want to ask about running my own pointcloud in S3DIS Fine-tuning #36

Open Freedomcls opened 2 years ago

Freedomcls commented 2 years ago

Hello, excuse me. I am a novice to 3d semantic segmentation. And it's nice to see your work. I'd like to ask you a few questions. ①I would like to run my own point cloud in the semantic segmentation task of S3DIS Fine-tuning. May I ask whether could I convert my point cloud into the same file structure and format as the Stanford3dDataset_v1.2_Aligned_Version dataset, and then use it as input. In addition, whether the weight of the network (PRETRAIN=PATH_CHECKPOINT) is necessary in the ./scripts/train_stanford3d.sh. ②when I run the code following the README, I encountered problems in ./scripts/train_stanford3d.sh, it reported the following error: image I wonder if you know how to solve it, looking forward to your reply!!

Sekunde commented 2 years ago

have you got the preprocessed point cloud? can you try to load them by torch.load() to see if there are similar errors?

Freedomcls commented 2 years ago

Thanks！Now everything works successfully. And I want to ask how can I visualize the result of segmentation, because I don't think I saw it. Thanks again for your help.