<![CDATA[Starfruit (Averrhoa carambola L.) is a horticultural commodity with potential for development in export markets. However, manual postharvest sorting may result in errors and inconsistencies in determining fruit ripeness and size. This study aims to develop an automatic system for detecting the ripeness and size of starfruit using an ESP32-CAM, a weight sensor (load cell and HX711), a Convolutional Neural Network (CNN) algorithm, and the Internet of Things (IoT). The system determines the size category based on fruit weight according to SNI 4491:2009, while fruit images are processed using a CNN to classify ripeness into three categories: unripe, partially ripe, and fully ripe. The classification results are transmitted in real time via Telegram. Testing on 30 samples showed that the CNN achieved an accuracy of 86.6%, while the load cell sensor produced an average measurement error of approximately 1–2%. These results indicate that the system is capable of automatically sorting starfruit based on ripeness and size while enabling remote monitoring of the results. The system has the potential to improve sorting consistency and support starfruit quality management in accordance with established standards.]]>
Copyrights © 2026
