SNN-PAR: Energy Efficient Pedestrian Attribute Recognition via Spiking Neural Networks

Haiyang Wang,Qian Zhu,Mowen She,Yabo Li,Haoyu Song,Minghe Xu,Xiao Wang

2024-10-10

Abstract:Artificial neural network based Pedestrian Attribute Recognition (PAR) has been widely studied in recent years, despite many progresses, however, the energy consumption is still high. To address this issue, in this paper, we propose a Spiking Neural Network (SNN) based framework for energy-efficient attribute recognition. Specifically, we first adopt a spiking tokenizer module to transform the given pedestrian image into spiking feature representations. Then, the output will be fed into the spiking Transformer backbone networks for energy-efficient feature extraction. We feed the enhanced spiking features into a set of feed-forward networks for pedestrian attribute recognition. In addition to the widely used binary cross-entropy loss function, we also exploit knowledge distillation from the artificial neural network to the spiking Transformer network for more accurate attribute recognition. Extensive experiments on three widely used PAR benchmark datasets fully validated the effectiveness of our proposed SNN-PAR framework. The source code of this paper is released on \url{<a class="link-external link-https" href="https://github.com/Event-AHU/OpenPAR" rel="external noopener nofollow">this https URL</a>}.

Computer Vision and Pattern Recognition,Artificial Intelligence,Neural and Evolutionary Computing

What problem does this paper attempt to address?

### What problem does this paper attempt to solve?

This paper aims to solve the high - energy - consumption problem in the Pedestrian Attribute Recognition (PAR) task. Although Pedestrian Attribute Recognition based on Artificial Neural Networks (ANN) has made remarkable progress in recent years, it still faces the challenge of high - energy - consumption in practical applications. To solve this problem, the paper proposes a framework based on Spiking Neural Networks (SNN) - SNN - PAR to achieve more energy - efficient Pedestrian Attribute Recognition.

Specifically, the paper solves the problem in the following ways:

1. **Introducing Spiking Neural Networks (SNN)**: SNN has attracted attention because of its biological inspiration and low - energy - consumption characteristics. For the first time, the paper applies SNN to the Pedestrian Attribute Recognition task, taking advantage of its efficient spatio - temporal feature extraction ability while reducing energy consumption.

2. **Designing an energy - efficient SNN architecture**: The paper proposes a framework that includes a spiking tokenizer, a spiking Transformer backbone network and a feed - forward network. These components work together to achieve an energy - efficient processing flow from image to attribute prediction.

3. **Application of knowledge distillation technology**: To further improve the model performance, the paper introduces the Knowledge Distillation technology to transfer knowledge from the pre - trained ANN model to the SNN model. This not only improves the learning efficiency of SNN, but also enhances its robustness in complex environments.

4. **Experimental verification**: The paper has carried out a large number of experiments on three widely - used PAR benchmark datasets to verify the effectiveness of the proposed SNN - PAR framework.

In summary, the main goal of this paper is to build an energy - efficient and high - performance Pedestrian Attribute Recognition system by introducing SNN and knowledge distillation technology, thereby solving the shortcomings of existing methods in terms of energy consumption.