Skip to content

Latest commit

 

History

History
84 lines (71 loc) · 3.74 KB

README.md

File metadata and controls

84 lines (71 loc) · 3.74 KB
Raw

ModelNet-40 Classification Experiment

This repo implements the ModelNet-40 classification experiments presented in the paper. The experiment involves six architectures:

  1. PointNet: models/pointnet.py. This is directly copied from Pointnet_Pointnet2_pytorch for comparison with Architecture 2.
  2. PointNet_VecKM: models/pointnet_VecKM.py. We replace the PointNetEncoder in the original architecture with our VecKM module.
  3. PointNet2: models/pointnet2.py. This is directly copied from Pointnet_Pointnet2_pytorch for comparison with Architecture 4.
  4. PointNet2_VecKM: models/pointnet2_VecKM.py. We replace the first set abstraction layer with VecKM.
  5. PCT: Point Cloud Transformer models/PCT.py. This is directly copied from PCT_Pytorch for comparison with Architecture 6.
  6. PCT_VecKM: models/PCT_VecKM.py We replace the initial point embedding module in PCT with our VecKM.

The data augmentation and training strategies are borrowed from PCT_Pytorch. Many thanks to their great codes!

Data Preparation

Please download the data here and unzip the file into ./data/modelnet40_ply_hdf5_2048. The file structure shall look like:

./
├── data
│   └── modelnet40_ply_hdf5_2048
│       ├── ply_data_test0.h5
│       ├── ply_data_test_0_id2file.json
│       ├── ply_data_test1.h5
│       ├── ply_data_test_1_id2file.json
│       ├── ply_data_train0.h5
│       ├── ply_data_train_0_id2file.json
│       ├── ply_data_train1.h5
│       ├── ply_data_train_1_id2file.json
│       ├── ply_data_train2.h5
│       ├── ply_data_train_2_id2file.json
│       ├── ply_data_train3.h5
│       ├── ply_data_train_3_id2file.json
│       ├── ply_data_train4.h5
│       ├── ply_data_train_4_id2file.json
│       ├── shape_names.txt
│       ├── test_files.txt
│       └── train_files.txt
├── data.py
├── main.py
├── models
│   ├── pointnet.py
│   ├── pointnet_VecKM.py
│   ├── pointnet2.py
│   ├── pointnet2_VecKM.py
│   ├── PCT.py
│   ├── PCT_VecKM.py
├── README.md
└── util.py

Requirements

python >= 3.8
pytorch >= 1.9
h5py
scikit-learn
scipy

Models

We get the following accuracies by setting random seed as 0. Different GPUs will produce different results. My result is given by an RTXA5000 GPU.

Instance Accuracy Avg. Class Accuracy Inference Time (1 batch) # parameters
PointNet 90.8% 87.1% 3.03 ms 1.61M
PointNet + VecKM 92.9% (+2.1%) 89.7% (+2.6%) 14.3 ms 9.06M
PointNet2 92.8% 89.4% 117 ms 1.48M
PointNet2 + VecKM 93.0% (+0.2%) 89.7% (+0.3%) 65.8 ms (78% faster) 3.94M
PCT 92.5% 89.2% 149.72 ms 2.88M
PCT + VecKM 93.0% (+0.5%) 90.5% (+1.3%) 21.4 (6x faster) 5.07M

Training

python main.py --model pointnet
python main.py --model pointnet_VecKM
python main.py --model pointnet2
python main.py --model pointnet2_VecKM
python main.py --model PCT
python main.py --model PCT_VecKM