Deep-Probability-Aggregation-Clustering

PyTorch implementation for ECCV 2024 paper： 'Deep Probability Aggregation Clustering' official code

Introduction

Requirement

-python>=3.7
-pytorch>=1.6.0
-torchvision>=0.8.1
-munkres>=1.1.4
-numpy>=1.19.2
-cudatoolkit>=11.0

Usage

Simply run python pretrain_step.py to strat the contrastive pre-training.
Then run python clustering_step.py to deploy Deep Probability Aggregation Clustering.
We also provide the Python implementation python probability_aggregation_clustering.py. for machine clustering algorithm: 'Probability Aggregation Clustering'.

Results

Method (ACC %)	CIFAR-10	CIFAR-100	STL-10
K-menas + SimCLR	76.8	41.8	66.8
PAC + SimCLR	87.1	43.8	74.9
DPAC	93.4	55.5	93.4