Related Experiment Video
Updated: Jan 10, 2026
08:41
Lensfree On-chip Tomographic Microscopy Employing Multi-angle Illumination and Pixel Super-resolution
Published on: August 16, 2012
11.9K
Low-resolution driver face recognition based on super-resolution and triplet loss.
Zhi Zhang1,2, Bingyu Sun3, Jiuzhen Liang4
1Hefei Institutes of Physical Science, Chinese Academy of Sciences, Hefei, 230031, Anhui, China.
Scientific Reports
|November 27, 2025
Summary
This study introduces a new driver face recognition model (SPFL-DC) for challenging traffic scenarios. It automates dataset creation and uses super-resolution to improve low-resolution face recognition.
Area of Science:
- Computer Vision
- Artificial Intelligence
- Deep Learning
Background:
- Deep neural networks excel at face recognition but struggle in unconstrained, low-resolution scenarios like traffic monitoring.
- Vehicle-based face recognition faces challenges due to image variations and limited computational resources.
Purpose of the Study:
- To develop a robust low-resolution driver face detection and recognition model (SPFL-DC) for complex traffic environments.
- To enhance dataset construction efficiency and model prediction robustness for unconstrained face recognition.
Main Methods:
- Automated dataset construction using triplet loss, guided by license plate information and pre-training on public datasets.
- A super-resolution technique to process and fuse low-resolution images, preserving critical identity information.
- Experimental validation on AR and LFW datasets.
Main Results:
- The SPFL-DC framework demonstrates effective low-resolution driver face detection and recognition.
- Automated dataset construction significantly improves efficiency and model robustness.
- Super-resolution fusion enhances image quality while retaining essential identity features.
Conclusions:
- The SPFL-DC framework offers a competitive solution for face recognition in challenging, resource-constrained traffic monitoring systems.
- The novel dataset construction and super-resolution methods improve the performance and applicability of deep learning models in real-world scenarios.
