A Deep Learning Framework for Real-Time Detection of Rotten Fruits Using Yolov5s Architecture

Maintaining food safety and quality is always a necessity, as food is a vital source of sustenance. In current times, the implementation of computer vision-based technologies is proving to be efficient for real-life implementations, creating a digital world. This research work is focused on the application of a deep learning approach in food processing areas to mitigate the problem of fruit rotting on a large scale. A CNN-based Yolov5s model is proposed for automated detection of rotten and fresh apples using a dataset of 10,228 images. For this model, the map50 and map50-90 values were found to be 0.993 and 0.985, respectively. On testing the model with real data samples, the model is capable of detecting rotten fruits among the fresh fruits with images captured in varying conditions, proving the efficacy of the model.