Unsupervised Deep Learning by Neighbourhood Discovery

TL;DR

This paper presents an unsupervised deep learning method that discovers sample neighborhoods to learn discriminative features without manual labels, improving image classification performance across multiple benchmarks.

Contribution

It introduces a novel neighborhood discovery approach for unsupervised deep learning that iteratively learns class boundaries without labels.

Findings

Outperforms state-of-the-art unsupervised models on six benchmarks

Effective in both coarse-grained and fine-grained image classification

Demonstrates the viability of neighborhood-based unsupervised learning

Abstract

Deep convolutional neural networks (CNNs) have demonstrated remarkable success in computer vision by supervisedly learning strong visual feature representations. However, training CNNs relies heavily on the availability of exhaustive training data annotations, limiting significantly their deployment and scalability in many application scenarios. In this work, we introduce a generic unsupervised deep learning approach to training deep models without the need for any manual label supervision. Specifically, we progressively discover sample anchored/centred neighbourhoods to reason and learn the underlying class decision boundaries iteratively and accumulatively. Every single neighbourhood is specially formulated so that all the member samples can share the same unseen class labels at high probability for facilitating the extraction of class discriminative feature representations during training. Experiments on image classification show the performance advantages of the proposed method over the state-of-the-art unsupervised learning models on six benchmarks including both coarse-grained and fine-grained object image categorisation.