Combining pretrained CNN feature extractors to enhance clustering of complex natural images

TL;DR

This paper investigates how combining features from different pretrained CNN architectures can improve image clustering, proposing a multi-view clustering approach and a neural network model that achieves state-of-the-art results.

Contribution

It introduces a multi-view clustering framework using multiple CNN features and a neural network architecture trained end-to-end for enhanced image clustering.

Findings

CNN architecture choice significantly impacts clustering quality

Multi-view clustering with combined features improves results

Proposed method achieves state-of-the-art performance on natural image datasets

Abstract

Recently, a common starting point for solving complex unsupervised image classification tasks is to use generic features, extracted with deep Convolutional Neural Networks (CNN) pretrained on a large and versatile dataset (ImageNet). However, in most research, the CNN architecture for feature extraction is chosen arbitrarily, without justification. This paper aims at providing insight on the use of pretrained CNN features for image clustering (IC). First, extensive experiments are conducted and show that, for a given dataset, the choice of the CNN architecture for feature extraction has a huge impact on the final clustering. These experiments also demonstrate that proper extractor selection for a given IC task is difficult. To solve this issue, we propose to rephrase the IC problem as a multi-view clustering (MVC) problem that considers features extracted from different architectures as different "views" of the same data. This approach is based on the assumption that information contained in the different CNN may be complementary, even when pretrained on the same data. We then propose a multi-input neural network architecture that is trained end-to-end to solve the MVC problem effectively. This approach is tested on nine natural image datasets, and produces state-of-the-art results for IC.