loicland / superpoint_graph

As per the readme: https://github.com/loicland/superpoint_graph#datasets-without-rgb

If your data does not have RGB values you can easily use SPG. You will need to follow the instructions in partition/partition.ply regarding the pruning.

My data has good calibration and I would like to see the results with rgb values.
Can I use the prune function as used for s3dis for my dataset to include rgb values

superpoint_graph/partition/partition.py

Line 124 in 8428739

    
           xyz, rgb, labels, dump = libply_c.prune(xyz.astype('f4'), args.voxel_width, rgb.astype('uint8'), labels.astype('uint8'), np.zeros(1, dtype='uint8'), n_labels, 0)

Then stack rgb values with geof
features = np.hstack((geof, rgb/255.)).astype('float32')#add rgb as a feature for partitioning

I am not getting good results with these. Are there any additional changes?

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.

Using RGB values for custom dataset SPG partition