Application of Edge Impulse and ESP32-CAM in Smart Freezer Systems for Object Recognition

The increasing need for efficient refrigeration solutions, especially in food storage and pharmaceutical industries, highlights the importance of advanced monitoring systems. This study presents the development of an IoT-enabled multi-power source freezer system incorporating image recognition for item quality monitoring. The system combines a Vapor Compression Refrigeration System with machine learning-based object detection, utilizing ESP32-CAM and Edge Impulse to classify and track stored items in real time. A dataset of 395 images was gathered and categorized into 272 training samples across five distinct classes: Onion, Tomato, Watermelon, Leaves, and a non-food item (Fan). The trained model achieved an accuracy of 98.1%, while testing accuracy fluctuated between 71% and 99%, depending on object positioning. Compared to previous research, the model demonstrated enhanced precision and reliability. With a cooling capacity of 3,471 BTU/h, the freezer effectively monitored stored items by displaying recognition results via the Arduino IDE Serial Monitor. This research contributes to improved food quality monitoring, inventory management, and storage accountability in refrigeration systems. Future enhancements should focus on integrating image recognition into the user interface to improve accessibility and user experience.