Related Experiment Video
Updated: May 23, 2025
03:31
Author Spotlight: Enhancement of Salient Object Detection for Smart Grid Applications
Published on: December 15, 2023
450
Multi-grained pooling network for age estimation in degraded low-resolution images.
1Sichuan Institute of Computer Sciences, Chengdu, 610041, China.
Scientific Reports
|March 7, 2025
Summary
This study introduces the Multi-Grained Pooling Network (MGP-Net) for accurate age estimation from low-resolution images. The novel architecture preserves crucial details, outperforming existing methods in real-world scenarios.
Area of Science:
- Computer Vision
- Machine Learning
- Biometrics
Background:
- Low-resolution images pose significant challenges for accurate age estimation.
- Existing models degrade performance due to loss of crucial details and weakened feature representations.
Purpose of the Study:
- To develop a novel architecture for robust age estimation from low-resolution images.
- To address the limitations of current models in handling degraded image quality.
Main Methods:
- Proposing the Multi-Grained Pooling Network (MGP-Net) to capture multi-grained information during downsampling.
- Introducing a random shuffle degradation model to simulate realistic low-resolution images for training and evaluation.
Main Results:
- MGP-Net effectively preserves essential features for age estimation despite low image resolution.
- Experimental results on Morph, FG-NET, and CLAP2015 datasets show competitive performance against state-of-the-art models.
- The proposed method demonstrates robustness and applicability in real-world low-resolution scenarios.
Conclusions:
- MGP-Net offers a robust solution for age estimation in challenging low-resolution conditions.
- The approach maintains high performance comparable to models trained on high-resolution data.
- This work advances the applicability of age estimation in practical, real-world settings.
