DETEKSI TANAMAN RIMPANG (JAHE, KENCUR, KUNYIT, LENGKUAS) MENGGUNAKAN METODE CNN

Abstrak

Penelitian ini bertujuan untuk mengembangkan model deteksi tanaman rimpang seperti jahe (Zingiber officinale), kencur (Kaempferia galanga), kunyit (Curcuma longa), dan lengkuas (Alpinia galanga) menggunakan metode Convolutional Neural Network (CNN) dasar. CNN dipilih karena kemampuannya mengenali pola visual kompleks secara otomatis tanpa ekstraksi fitur manual. Dataset citra diperoleh dari berbagai sumber publik dan hasil pengambilan langsung di lapangan dengan kondisi pencahayaan berbeda. Arsitektur CNN yang digunakan terdiri atas tiga lapisan konvolusi, dua lapisan pooling, dan dua lapisan fully connected. Proses pelatihan dilakukan menggunakan Adam optimizer dengan learning rate 0,001 dan batch size 32 selama 50 epochs. Hasil pengujian menunjukkan akurasi rata-rata 92,4%, presisi 91,8%, dan recall 92,5%. Hasil ini menunjukkan bahwa CNN dasar mampu menjadi alternatif efisien untuk deteksi otomatis tanaman rimpang, khususnya pada perangkat dengan keterbatasan komputasi.

This study aims to develop a detection model for rhizome plants such as ginger (Zingiber officinale), aromatic ginger (Kaempferia galanga), turmeric (Curcuma longa), and galangal (Alpinia galanga) using a basic Convolutional Neural Network (CNN) method. CNN was chosen due to its ability to automatically recognize complex visual patterns without manual feature extraction. The image dataset was collected from various public sources and direct field captures under different lighting conditions. The CNN architecture used consists of three convolutional layers, two pooling layers, and two fully connected layers. The training process was carried out using the Adam optimizer with a learning rate of 0.001 and a batch size of 32 for 50 epochs. The test results show an average accuracy of 92.4%, precision of 91.8%, and recall of 92.5%. These results indicate that the basic CNN model can serve as an efficient alternative for automatic detection of rhizome plants, especially on devices with limited computational resources.