Lightweight network improves pedestrian head orientation recognition accuracy
Lightweight Pedestrian Head-Orientation Recognition Network for Safe Pedestrian-Vehicle Interaction
Computer Vision and Pattern Recognition
Summary
Recognizing which way a pedestrian's head is turned helps self-driving cars know if they might cross the street. But it's hard because pedestrian heads often appear small and blurry in traffic camera images. The authors created a new, simple but effective computer program called LRHO-CNN that focuses on recognizing head direction even in low-quality pictures. They also gathered many head images from different datasets and carefully labeled them by direction. Their program performs better than several well-known methods and works well in real-world traffic scenes, helping cars understand pedestrian intentions.
What this means in practice
- •For autonomous vehicle developers: Improve self-driving car systems by reliably identifying pedestrian attention direction from low-resolution images to better predict crossing behavior.
- •For traffic safety system engineers: Enhance pedestrian safety applications with lightweight models capable of real-time head orientation recognition in various lighting and image quality conditions.
Authors
Yuanzhe Li, Yidi Huang, Xiaotong Chang, Hounian Liu
Abstract
Pedestrian head orientation recognition plays an important role in autonomous driving by providing valuable cues for understanding pedestrian attention and anticipating potential crossing behavior. However, reliable recognition in real-world traffic scenes remains challenging because pedestrian head regions are often captured at low resolution. To address this challenge, we propose a lightweight Low-Resolution Head Orientation Convolutional Neural Network (LRHO-CNN) for pedestrian head orientation recognition. We construct a new dataset by extracting pedestrian head images from multiple public datasets and manually annotating them into eight orientation categories. The collected images are systematically preprocessed and augmented to increase data diversity and better represent variations in illumination and image quality. The experimental analysis compares LRHO-CNN with three fine-tuned baseline models, namely ResNet-18, ResNet-34, and VGG-16. The results demonstrate that LRHO-CNN achieves the highest classification accuracy among the evaluated models. LRHO-CNN is further evaluated on the JAAD and PIE datasets, demonstrating its effectiveness in recognizing pedestrian head orientation in real-world traffic scenes and providing informative head-orientation cues that can support downstream pedestrian behavior and intention prediction.