ENHANCING CNN PERFORMANCE VIA GENETIC ALGORITHMS FOR FATIGUE MONITORING

Abstract

The use of evolutionary algorithms improves convolutional neural networks for the detection of sleepy drivers. By identifying the best hyperparameters, detection improves the model's performance. Using Genetic Algorithms (GAs) and Convolutional Neural Networks (CNNs), this method tweaks learning rates, filter dimensions, and layer structure. Genetic algorithms improve detection performance by iteratively adjusting CNN parameters through mutation, selection, and crossover. In order to detect weariness in real-time, the improved CNN records key facial signals such as eye shutting and yawning patterns. This method not only makes things last longer, but it also decreases handling expenses and overfitting. Convergence is faster and more accurate than with traditional CNNs, according to the data. Through continuous driver monitoring, this device enhances traffic safety. In the future, researchers will mainly concentrate on practical uses and methods to include several sensors.