Deep Learning for Sustainable Food Systems: Mitigating Climate Change Through Food Waste Reduction

Abstract

Climate change is a critical global issue, with food waste being a significant contributing factor due to its environmental impacts, including greenhouse gas emissions. Reducing food waste can substantially mitigate climate change. This study introduces a lightweight, nature-inspired multi-channel deep learning technique utilizing Capsule Neural Networks and Particle Swarm Optimization (PSO) for hyperparameter tuning. The multi-channel CapsNet model automatically classifies fresh and rotten fruits. Trained and evaluated on over 23,200 images of apples and bananas, the model achieved perfect test accuracy, precision, F1 score, and ROC score of 100%. This research promotes sustainable practices in the food industry, aligning with global efforts to minimize climate change impacts and enhance resource efficiency.