Comparing the performance of yolov10s and ssd300 models in the problem of automatic fruit identification and classification

Tóm tắt

DOI: https://doi.org/10.66195/mtu.2025.12.074

The research focuses on applying deep learning to automate the fruit recognition and classification process, meeting the development needs of modern agriculture. Applying this technology helps improve efficiency and classification quality and reduces labor costs, resulting in lower product prices. The research team used two deep learning models, SSD300 and YOLOv10s, to recognize and classify six types of fruits: apples, bananas, kiwis, lemons, oranges, and strawberries. The dataset consists of 2575 images divided into Train, Validation, and Test sets with a ratio of 87%-8%-4%. The images were resized to 300x300 pixels for SSD300 and 640x640 pixels for YOLOv10s. The experimental results show that the YOLOv10s model achieved higher precision at 96% compared to 93% for SSD300. The research also proposes future improvements to enhance the system’s accuracy and applicability.