Freshness Detection of Fruits and Vegetables Using Multi-Task CNN and ResNet-50 [Under Review]

In this research, developed a deep multi-task learning CNN model (MTL-CNN) for fruit freshness detection and fruit-type classification, achieving 98.63% accuracy by leveraging shared feature extraction and addressing imbalanced datasets. The model outperforms existing methods like InceptionV3 and CNN_BiLSTM, offering significant implications for automated quality control, food safety, and waste reduction in the agricultural and consumer sectors.

Recommended citation: SK Ray, MA Hossain, N Islam and Mirza AFMRH; Freshness Detection of Fruits and Vegetables Using Multi-Task CNN and ResNet-50.