Abstarct: Using deep learning technique to develop a shelflife prediction model of apricot fruit according to maturity stage harvesting Apricot is a climacteric fruit, and due to its rapid ripening and softening after harvest, it is highly perishable. This increases its susceptibility to mechanical damage and pathogenic contamination, leading to higher post-harvest losses. The appropriate harvest stage can have a significant impact on maintaining post-harvest quality and extending shelf life. To investigate this, the effect of harvest stage on the storage life of ‘Zaferani’ apricot was examined across five different harvest stages following physiological maturity, in accordance with the standards defined by the University of California. The assessment included traits such as firmness, weight loss, color changes, pH, soluble solids content, and titratable acidity during 35 days of storage at 6°C. In parallel, to develop a deep neural network model, images of apricot fruits were captured at different stages of growth on the tree. These images were categorized into five classes, each corresponding to one of the five harvest stages. The results indicated that fruit firmness and titratable acidity decreased over time, while weight loss, color changes, and soluble solids content increased. baxsed on the percentage change in the studied traits, it was found that the least variation occurred in the third harvest stage, which is considered the optimal stage for commercial harvest. The changes in weight loss, firmness, pH, soluble solids, and titratable acidity for this stage were 3.5%, 34.39%, 6.54%, 4.16%, and 20.84%, respectively. Additionally, the deep learning model accurately identified and classified the five growth stages with an 80% accuracy rate. It appears that image capture combined with the use of a deep learning model can serve as a non-destructive method to determine the optimal harvest time for ‘zaferani’ apricots.