<![CDATA[Tempe is a traditional Indonesian food whose fermentation process is highly influenced by temperature, humidity, and soybean quality. Inadequate environmental conditions can lead to fermentation failure, reducing product quality and causing economic losses. Traditionally, quality assessment of tempe has been carried out manually by artisans, which is subjective and inconsistent. This study aims to develop an automatic tempe quality recognition system based on YOLOv8, implemented on a Raspberry Pi 4B with a Logitech C270 camera, monitoring webserver, and buzzer as an early warning system. The YOLOv8 model was trained to detect two main categories, namely good tempe and failed tempe, through real-time visual analysis. Experimental results show system performance with an average accuracy of 93.1%, precision of 93.5%, recall of 91.2%, and mAP@50 of 94.7%. Confidence score analysis indicates that the model is more certain in detecting failed tempe (0.94–0.95) compared to good tempe (0.80–0.86), due to clearer visual differences.]]>
Copyrights © 2025
