Towards Accurate Binary Convolutional Neural Network

TL;DR

This paper presents ABC-Net, a binary CNN training scheme that uses multiple binary weight bases and activations to significantly improve accuracy while maintaining the efficiency benefits of binarization.

Contribution

It introduces a novel approach combining multiple binary weight bases and activations to reduce accuracy loss in binary CNNs.

Findings

Achieves comparable accuracy to full-precision CNNs on ImageNet.

Reduces memory and computation costs with binary operations.

Improves prediction accuracy over previous binary CNN methods.

Abstract

We introduce a novel scheme to train binary convolutional neural networks (CNNs) -- CNNs with weights and activations constrained to {-1,+1} at run-time. It has been known that using binary weights and activations drastically reduce memory size and accesses, and can replace arithmetic operations with more efficient bitwise operations, leading to much faster test-time inference and lower power consumption. However, previous works on binarizing CNNs usually result in severe prediction accuracy degradation. In this paper, we address this issue with two major innovations: (1) approximating full-precision weights with the linear combination of multiple binary weight bases; (2) employing multiple binary activations to alleviate information loss. The implementation of the resulting binary CNN, denoted as ABC-Net, is shown to achieve much closer performance to its full-precision counterpart, and even reach the comparable prediction accuracy on ImageNet and forest trail datasets, given adequate binary weight bases and activations.