BiHRNet: A Binary high-resolution network for Human Pose Estimation

TL;DR

BiHRNet introduces a binary neural network for human pose estimation that maintains high accuracy while significantly reducing computational costs, making it suitable for resource-limited devices.

Contribution

The paper proposes a novel binary HRNet-based pose estimator with specialized loss functions and network structures to minimize accuracy loss due to binarization.

Findings

Achieves 87.9 PCKh on MPII, outperforming other binary networks.

Attains 70.8 mAP on COCO, surpassing many lightweight full-precision models.

Uses fewer resources with minimal accuracy compromise.

Abstract

Human Pose Estimation (HPE) plays a crucial role in computer vision applications. However, it is difficult to deploy state-of-the-art models on resouce-limited devices due to the high computational costs of the networks. In this work, a binary human pose estimator named BiHRNet(Binary HRNet) is proposed, whose weights and activations are expressed as $\pm$1. BiHRNet retains the keypoint extraction ability of HRNet, while using fewer computing resources by adapting binary neural network (BNN). In order to reduce the accuracy drop caused by network binarization, two categories of techniques are proposed in this work. For optimizing the training process for binary pose estimator, we propose a new loss function combining KL divergence loss with AWing loss, which makes the binary network obtain more comprehensive output distribution from its real-valued counterpart to reduce information loss caused by binarization. For designing more binarization-friendly structures, we propose a new information reconstruction bottleneck called IR Bottleneck to retain more information in the initial stage of the network. In addition, we also propose a multi-scale basic block called MS-Block for information retention. Our work has less computation cost with few precision drop. Experimental results demonstrate that BiHRNet achieves a PCKh of 87.9 on the MPII dataset, which outperforms all binary pose estimation networks. On the challenging of COCO dataset, the proposed method enables the binary neural network to achieve 70.8 mAP, which is better than most tested lightweight full-precision networks.