Exploiting Inductive Bias in Transformer for Point Cloud Classification and Segmentation

This repository contains PyTorch implementation for IBT : Exploiting Inductive Bias in Transformer for Point Cloud Classification and Segmentation. Our code skeleton is borrowed from antao97/dgcnn.pytorch

Our IBT module is as follows:

Requirements

Python >= 3.7 , PyTorch >= 1.2 , CUDA >= 10.0 , Package: glob, h5py, sklearn, plyfile, torch_scatter

Point Cloud Classification

Note: You can choose to implement the classification experiment on the ModelNet40 or ScanObjectNN dataset, just select the corresponding data loading function.

Run the training script:

python main_cls.py --exp_name=cls_1024_scan --num_points=1024 --k=40 

Run the evaluation script after training finished:

python main_cls.py --exp_name=cls_1024_eval_scan --num_points=1024 --k=40 --eval=True --model_path=outputs/cls_1024_scan/models/model.t7

Point Cloud Part Segmentation

Note: There are two options for training on the full dataset and for each class individually. In order to obtain more detailed features, we set the k value to 80 for Part Segmentation.

Run the training script:

· Full dataset

python main_partseg.py --exp_name=partseg 

· With class choice, for example airplane

python main_partseg.py --exp_name=partseg_airplane --class_choice=airplane

Run the evaluation script after training finished:

· Full dataset

python main_partseg.py --exp_name=partseg_eval --eval=True --model_path=outputs/partseg/models/model.t7

· With class choice, for example airplane

python main_partseg.py --exp_name=partseg_airplane_eval --class_choice=airplane --eval=True --model_path=outputs/partseg_airplane/models/model.t7