search

loicland / superpoint_graph

Large-scale Point Cloud Semantic Segmentation with Superpoint Graphs
MIT License
745 stars 214 forks source link

Question on custom dataset #256

Closed Devoe-97 closed 1 year ago

Devoe-97 commented 3 years ago

I run partition.py to prune my point cloud(63825 points, xmin=-0.78, xmax=-0.0367, ymin=-0.1375, ymax=0.08109, zmin=-0.552, zmax=0.568) and i set voxel_width=0.01 and it shows:

Voxelization into 0 x 0 x 0 grid Reduced from 63825 to 1 points (0%)

How should i fix thie? Thanks

Devoe-97 commented 3 years ago

i use this code xyz, rgb, labels, o = libply_c.prune(xyz, args.voxel_width, rgb, np.array(1,dtype='uint8'), np.zeros(1, dtype='uint8'), 0, 0)

qzq-hit commented 2 years ago

I run partition.py to prune my point cloud(63825 points, xmin=-0.78, xmax=-0.0367, ymin=-0.1375, ymax=0.08109, zmin=-0.552, zmax=0.568) and i set voxel_width=0.01 and it shows:

Voxelization into 0 x 0 x 0 grid Reduced from 63825 to 1 points (0%)

How should i fix thie? Thanks Have you solved this problem? It happened when I was pruning my own data set

antoine-lag commented 2 years ago

I had the same problem and it came from line 192 of ply_c.cpp, the type of xyz was not float but double

loicland commented 1 year ago

Hi!

We are releasing a new version of SuperPoint Graph called SuperPoint Transformer (SPT). It is better in any way:

✨ SPT in numbers ✨
📊 SOTA results: 76.0 mIoU S3DIS 6-Fold, 63.5 mIoU on KITTI-360 Val, 79.6 mIoU on DALES
🦋 212k parameters only!
⚡ Trains on S3DIS in 3h on 1 GPU
⚡ Preprocessing is x7 faster than SPG!
🚀 Easy install (no more boost!)

If you are interested in lightweight, high-performance 3D deep learning, you should check it out. In the meantime, we will finally retire SPG and stop maintaining this repo.