Deep Unsupervised Learning of Visual Similarities

TL;DR

This paper introduces an unsupervised deep learning method for visual similarity that groups similar samples and learns a unified representation without labels, addressing challenges of data imbalance and unreliable relationships.

Contribution

It proposes a novel optimization framework that uses weak similarity estimates to form consistent sample batches, enabling CNNs to learn visual similarities without supervision.

Findings

Achieved competitive results in posture analysis

Performed well in object classification tasks

Effectively groups similar samples without labels

Abstract

Exemplar learning of visual similarities in an unsupervised manner is a problem of paramount importance to Computer Vision. In this context, however, the recent breakthrough in deep learning could not yet unfold its full potential. With only a single positive sample, a great imbalance between one positive and many negatives, and unreliable relationships between most samples, training of Convolutional Neural networks is impaired. In this paper we use weak estimates of local similarities and propose a single optimization problem to extract batches of samples with mutually consistent relations. Conflicting relations are distributed over different batches and similar samples are grouped into compact groups. Learning visual similarities is then framed as a sequence of categorization tasks. The CNN then consolidates transitivity relations within and between groups and learns a single representation for all samples without the need for labels. The proposed unsupervised approach has shown competitive performance on detailed posture analysis and object classification.