LirongWu / GCML

Code for WACV 2022 paper "Generalized Clustering and Multi-Manifold Learning with Geometric Structure Preservation"

Generalized Clustering and Multi-Manifold Learning (GCML)

The code includes the following modules:

• Datasets (MNIST-full, MNIST-test, USPS, Fashion-MNIST, Reuters-10k, HAR, and Pendigits)
• Training for GCML
• Evaluation metrics
• Visualisation

Requirements

• pytorch == 1.3.1
• scipy == 1.3.1
• numpy == 1.18.5
• scikit-learn == 0.21.3
• matplotlib == 3.1.1

Description

• main.py

  • pretrain() -- Pretraining the model with self-reconstruction Loss
  • train() -- End-to-end training of the GCML model
  • test() -- Test generalization performance on out-of-sample (testing sample)

• autotrain.py -- Scripts for automatic testing on seven datasets

• dataset.py

  • Dataset() -- Load data of selected dataset

• evaluation.py

  • GetIndicator() -- Auxiliary tool for evaluating metric

• loss.py

  • Loss_calculate() -- Calculate losses: ℒ_LIS, ℒ_rank, ℒ_AE, ℒ_align

• model.py

  • AutoEncoder() -- The architecture used in this work
  • GCML() -- Calculation Q distribution and P distribution

• utils.py

  • visualize() -- Auxiliary tools for visualizing intermediate results
  • Clustering() -- For initializing the clustering centers

Dataset

The datasets used in this paper are available in:

https://drive.google.com/file/d/1nNenJQVBJ-R4B6rs_K_YxGrVyZq4kAfz/view?usp=sharing

Running the code

1. Install the required dependency packages

2. To get the results on seven datasets, run

python autotrain.py

3. To get the metrics and visualisation, refer to

../plots/dataset/pics/

where the dataset is one of the seven datasets (MNIST-full, MNIST-test, USPS, Fashion-MNIST, Reuters-10k, HAR, and Pendigits)

Citation

If you find this project useful for your research, please use the following BibTeX entry.

@inproceedings{wu2022generalized,
  title={Generalized Clustering and Multi-Manifold Learning with Geometric Structure Preservation},
  author={Wu, Lirong and Liu, Zicheng and Xia, Jun and Zang, Zelin and Li, Siyuan and Li, Stan Z},
  booktitle={Proceedings of the IEEE/CVF Winter Conference on Applications of Computer Vision},
  pages={139--147},
  year={2022}
}