tangsankou / dfa

Dynamic Local Feature Aggregation for Learning on Point Clouds

This repository contains PyTorch implementation for DFA : Dynamic Local Feature Aggregation for Learning on Point Clouds. Our code skeleton is borrowed from antao97/dgcnn.pytorch

Our DFA module is as follows:

We use DFA to complete the point cloud classification and segmentation task structure 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 1024 or 2048 points for training and evaluation.

Run the training script:

python main_cls.py --exp_name=cls_1024 --num_points=1024 --k=20 

Run the evaluation script after training finished:

python main_cls.py --exp_name=cls_1024_eval --num_points=1024 --k=20 --eval=True --model_path=outputs/cls_1024/models/model.t7

Point Cloud Part Segmentation

Note: There are two options for training on the full dataset and for each class individually.

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

Point Cloud Semantic Segmentation on the S3DIS Dataset

Note : This task uses 6-fold training.

Run the training script:

python main_semseg_s3dis.py --exp_name=semseg_s3dis_5 --test_area=5 

Run the evaluation script after training finished:

python main_semseg_s3dis.py --exp_name=semseg_s3dis_eval_5 --test_area=5 --eval=True --model_root=outputs/semseg_s3dis/models/