Sapota Ripeness Detection and Shelf Life Prediction Using Thermal

Sapota, a tropical fruit, undergoes significant changes in ripeness and quality over time, impacting its shelf life and consumer appeal. This project explores the application of thermal imaging for sapota ripeness detection and shelf-life prediction, along with bruising classification. The dataset comprises images categorized into rotten and not rotten classes, accompanied by the number of days since the image was captured. Additionally, another dataset includes labels indicating bruises or no bruises on the fruit. Various convolutional neural network (CNN) architectures such as MobileNet, ResNet, and VGG16 are employed for both tasks. The effectiveness of each model is evaluated for sapota ripeness classification and shelf-life prediction, as well as bruising detection. Results indicate the potential of thermal imaging coupled with deep learning techniques for non-destructive quality assessment of sapota fruit, aiding in reducing food waste and enhancing consumer satisfaction.

Keywords: Sapota, thermal imaging, ripeness detection, shelf-life prediction, bruising classification, convolutional neural networks (CNNs), MobileNet, ResNet, VGG16, fruit quality assessment, food waste reduction