Adaptive Deep Learning with Optimization Hybrid Convolutional Neural Network and Recurrent Neural Network for Prediction Lemon Fruit Ripeness

Lemon is a valuable fruit in the citrus family; optimal usage requires careful selection. The study categorized lemon suitability prediction int 4 classes based on image data. A hybrid neural network, combining Convolutional and Recurrent Neural Networks, was optimized with the Particle Swarm Optimization algorithm. Experimental results were compared to using Convolutional Neural Network alone. The prediction yielded 89.83% training accuracy and 66.06% testing accuracy. However, combining the results increased training accuracy to 91.58% and testing accuracy to 86.76%.