This an open-source implementation of a method for 3D model attention based classification proposed by Chen et al.
To install them, run:
pip install -r requirements.txt
First, specify the training configuration and all model/DataModule hyperparameters in their respective .yaml config files under configs/, then run in the root directory chose between mvt_module and mvt_module_with_back_bone in the target name of the model/mvt.yaml config file to train either model
python train.py
After specifying the path to the checkpointed model and its config file (respective target name for example) run in root directory:
python test.py
Project implemented and maintained by Alexandre Lutt, Boris Burkalo, Mohamed Said Derbel, Ludwig Gräf.