Hierarchical binary CNNs for landmark localization with limited resources

TL;DR

This paper introduces a novel hierarchical binary CNN architecture for landmark localization that maintains high accuracy while being lightweight, suitable for resource-constrained applications, and demonstrates state-of-the-art results on challenging datasets.

Contribution

It is the first to analyze neural network binarization effects on localization tasks and proposes a new multi-scale residual architecture to improve performance without increasing parameters.

Findings

Achieved state-of-the-art results on human pose estimation datasets.

Demonstrated that binarized networks can perform effectively on localization tasks.

Provided extensive ablation studies to understand the proposed architecture's properties.

Abstract

Our goal is to design architectures that retain the groundbreaking performance of Convolutional Neural Networks (CNNs) for landmark localization and at the same time are lightweight, compact and suitable for applications with limited computational resources. To this end, we make the following contributions: (a) we are the first to study the effect of neural network binarization on localization tasks, namely human pose estimation and face alignment. We exhaustively evaluate various design choices, identify performance bottlenecks, and more importantly propose multiple orthogonal ways to boost performance. (b) Based on our analysis, we propose a novel hierarchical, parallel and multi-scale residual architecture that yields large performance improvement over the standard bottleneck block while having the same number of parameters, thus bridging the gap between the original network and its binarized counterpart. (c) We perform a large number of ablation studies that shed light on the properties and the performance of the proposed block. (d) We present results for experiments on the most challenging datasets for human pose estimation and face alignment, reporting in many cases state-of-the-art performance. (e) We further provide additional results for the problem of facial part segmentation. Code can be downloaded from https://www.adrianbulat.com/binary-cnn-landmark