Learning Effective Binary Visual Representations with Deep Networks

TL;DR

This paper introduces Approximately Binary Clamping (ABC), a novel non-saturating, end-to-end trainable method for generating true binary visual representations that improve accuracy and generalization in recognition, detection, and retrieval tasks.

Contribution

The paper presents ABC, a new binary representation learning method that converges faster and outperforms existing hashing techniques in accuracy and generalization.

Findings

ABC achieves comparable ImageNet classification accuracy to real-valued models.

ABC outperforms existing hashing methods on image retrieval benchmarks.

Binary representations generalize better in object detection tasks.

Abstract

Although traditionally binary visual representations are mainly designed to reduce computational and storage costs in the image retrieval research, this paper argues that binary visual representations can be applied to large scale recognition and detection problems in addition to hashing in retrieval. Furthermore, the binary nature may make it generalize better than its real-valued counterparts. Existing binary hashing methods are either two-stage or hinging on loss term regularization or saturated functions, hence converge slowly and only emit soft binary values. This paper proposes Approximately Binary Clamping (ABC), which is non-saturating, end-to-end trainable, with fast convergence and can output true binary visual representations. ABC achieves comparable accuracy in ImageNet classification as its real-valued counterpart, and even generalizes better in object detection. On benchmark image retrieval datasets, ABC also outperforms existing hashing methods.