Validation IoUs are computed on the interpolated sample instead of the subsampled one, which takes longer but gives more meaningful information during training. This means that transforms are defined separately for training, eval (val/test), and prediction.
HDF5 datasets for training/evaluation, and Inference dataset that uses the same code. Overall simplification of data preparation. It now happens on the fly, and input is a folder of LAS + a csv file with basename (xyz.las) x split pairs (split=train/val/test).
Only keep RandLaNet to ease maintenance, as other architectures are not used.
Positions are centered but not normalized, to always work in the metric domain and facilitate configuration of data augmentations.
Preparing for V3.0.0
Key Changes :
basename(xyz.las) xsplitpairs (split=train/val/test).