Copyright © 2026 Authors retain the copyright of this article. This article is an open access article distributed under the Creative Commons Attribution License which permits unrestricted use, distribution, and reproduction in any medium, provided the original work is properly cited.

@article{178888,
        author = {Shreya Shivkumar Chillal and Dr. Renu Kachhoria and Kartik Kumbhar},
        title = {DEEP LEARNING-BASED INTELLIGENT SYSTEM FOR FOOD FRESHNESS ASSESSMENT IN SMART REFRIGERATORS},
        journal = {International Journal of Innovative Research in Technology},
        year = {2025},
        volume = {11},
        number = {12},
        pages = {7324-7331},
        issn = {2349-6002},
        url = {https://ijirt.org/article?manuscript=178888},
        abstract = {Maintaining the safety and freshness of foods and vegetables, especially perishables like fruits and vegetables, is now much more important than ever in today's hectic and busy environment. Using visual inspection and determination, touch, or smell to determine freshness is a traditional method that is typically subjective and unreliable. This task project represents how deep learning might offer us a more intelligent and trustworthy road to assess the quality of perishable food. We present a Convolutional Neural Network (CNN)-powered freshness detection system that focuses on transforming ordinary refrigerators into intelligent food tracking gadgets.
This system examines pictures of fruits and vegetables to determine and convey us there freshness levels and it also makes conclusions of the results and will send out alerts instantly when in need. The performance of five deep learning architectures ResNet50, MobileNetV2, EfficientNetB3, VGG16, and YOLOv8 is compared in our analysis with a carefully chosen dataset of 9,300 labeled images from eight different classes of fruits and vegetables, which are divided into "fresh" and "rotten." We trained the models and evaluated the models in terms of accuracy, precision, recall, and F1-score by applying standard preprocessing, data augmentation, and transfer learning.
YOLOv8 proved to be the best for real-time use with the highest accuracy of 91.5% among them. EfficientNetB3 (91.0%) was second in demonstrating a better balance between computational cost and performance. In some cases, VGG16 and ResNet50 showed promising performance, however MobileNetV2 proved to be weak in this story. Going with the research deep learning can lead to improvement of our overall sustainability and health by lowering waste and improving how we handle our food. AI-assisted freshness determination is likely to become a important component of modern food management systems and smart Kitchen, with applications ranging from home kitchens to industrial food delivery systems.},
        keywords = {data augmentation, transfer learning, Deep Learning, YOLOv8, EfficientNetB3, ResNet50, MobileNetV2, VGG16},
        month = {May},
        }