<![CDATA[Abstrak - Penyortiran kualitas tomat masih banyak dilakukan secara manual sehingga prosesnya lambat, membutuhkan tenaga kerja besar, dan sering menghasilkan penilaian yang tidak konsisten. Penelitian ini bertujuan merancang sistem penyortiran tomat otomatis berbasis Internet of Things (IoT) dan Computer Vision menggunakan ESP32-CAM dan model YOLOv8. Sistem terdiri dari ESP32 DevKit untuk membaca sensor ultrasonik dan mengendalikan servo, ESP32-CAM untuk mengambil citra, serta server Flask yang memproses gambar menggunakan YOLOv8. Dataset diperoleh dari Roboflow dan dilatih melalui proses augmentasi dan preprocessing sehingga model dapat mengklasifikasikan tomat menjadi tiga kategori, yaitu matang, mentah, dan busuk. Hasil pengujian menunjukkan bahwa sensor ultrasonik mampu mendeteksi tomat secara stabil pada jarak 10 cm, ESP32-CAM berhasil mengirim gambar ke server, dan servo dapat menyortir tomat sesuai hasil prediksi. Sistem web monitoring yang dibangun mampu menampilkan prediksi terbaru, grafik statistik, serta riwayat prediksi harian secara real-time. Beberapa kendala ditemukan, seperti kesulitan model membedakan tomat merah busuk dan tomat matang yang memiliki kemiripan visual, serta motor conveyor yang kurang kuat pada kecepatan rendah. Secara keseluruhan, sistem berhasil berfungsi sebagai prototipe penyortiran tomat otomatis yang terintegrasi dan dapat dikembangkan lebih lanjut untuk penggunaan skala industri.Kata kunci : IoT; ESP32-CAM; YOLOv8; Penyortiran Tomat; Computer Vision; Abstract - Manual tomato quality sorting is still widely practiced, resulting in slow processing, high labor requirements, and inconsistent assessment outcomes. This study aims to design an automatic tomato sorting system based on the Internet of Things (IoT) and Computer Vision using ESP32-CAM and the YOLOv8 model. The system consists of an ESP32 DevKit to read ultrasonic sensor data and control servo motors, an ESP32-CAM to capture tomato images, and a Flask server to process images using the YOLOv8 model. The dataset was obtained from Roboflow and trained through augmentation and preprocessing processes to enable the model to classify tomatoes into three categories: ripe, unripe, and rotten. Experimental results show that the ultrasonic sensor can stably detect tomatoes at a distance of 10 cm, the ESP32-CAM successfully transmits images to the server, and the servo motor can sort tomatoes according to the prediction results. The developed web-based monitoring system is capable of displaying real-time predictions, statistical graphs, and daily prediction history. Several limitations were identified, including the model’s difficulty in distinguishing between rotten red tomatoes and ripe tomatoes due to visual similarity, as well as insufficient conveyor motor strength at low speeds. Overall, the proposed system functions effectively as an integrated automatic tomato sorting prototype and can be further developed for industrial-scale applications.Keywords: IoT; ESP32-CAM; YOLOv8; Tomato Sorting; Computer Vision;]]>
