<![CDATA[Egg incubation and fertility detection are critical processes in poultry production, requiring stable environmental conditions and accurate embryo monitoring. Conventional incubation and candling methods often rely on manual observation, which may reduce efficiency and increase the risk of human error. This study aims to develop an IoT-based automatic egg incubator integrated with a MobileNetV2-based Convolutional Neural Network (CNN) for candling image classification and fertility detection. The proposed system combines an ESP32-based incubation monitoring platform, ESP32-CAM image acquisition module, ESP-NOW wireless communication, and a web-based monitoring interface. Candling images were collected and classified into three categories: live fertile, dead fertile, and infertile eggs. Transfer learning using the MobileNetV2 architecture was employed to train the classification model. The developed CNN model achieved an accuracy of 91.26% on the testing dataset and 82.22% during validation under real-world conditions. Performance evaluation showed precision values up to 0.965, recall values up to 0.967, and F1-scores above 0.89 across the three classes. Furthermore, the integrated system successfully enabled real-time monitoring of incubation conditions, automated image acquisition, and web-based fertility classification. System testing using 45 egg samples demonstrated that the proposed solution could effectively identify embryo development and fertility status. The integration of IoT technology, ESP32-CAM-based candling, and MobileNetV2 CNN provides an effective solution for automated incubation monitoring and egg fertility detection. The developed system improves monitoring efficiency, reduces manual intervention, and supports decision-making during the incubation process.]]>
