EfficientNet B7 based Deep Learning Technique for Adaptive Gait Recognition under Challenging Covariate Conditions
Authors
Md. Wadud Jahan
Abstract
Gait recognition is a crucial biometric tool for surveillance and security, but its effectiveness is hindered by factors like clothing, carried obj ects, and environmental conditions. This study introduces an adaptive framework that assigns weights to gait components to counter these issues. Using EfficientNet_B7, a deep learning technique, the framework reduces intra-class variations and improves recognition accuracy. The experimental results show a 98.7% accuracy rate, proving the method's robustness in various conditions. This advancement significantly enhances the reliability of gait recognition systems in real-world applications
