Tanaman hias Philodendron memiliki banyak varietas yang seringkali sulit dibedakan karena kemiripan visualnya. Penelitian ini bertujuan mengembangkan sistem untuk klasifikasi jenis tanaman Philodendron berdasarkan citra daun menggunakan algoritma Convolutional Neural Network (CNN) guna membantu proses identifikasi. Dataset penelitian terdiri dari 5000 citra daun dari lima jenis Philodendron (Rhaphidophora tetrasperma, Epipremnum pinnatum, Monstera deliciosa, Mayoi, dan Amydrium), dengan 1000 citra per jenis, yang diperoleh melalui pemotretan langsung. Model CNN diimplementasikan menggunakan arsitektur SSD MobileNetV2 FPNLite dengan input 320x320 piksel yang dilatih selama 50.000 langkah. Model tersebut dioptimalkan untuk perangkat bergerak menggunakan TensorFlow dan dikonversi ke TensorFlow Lite. Data dibagi menjadi data latih (80%), validasi (10%), dan uji (10%). Kinerja sistem dievaluasi menggunakan confusion matrix serta metrik akurasi, presisi, recall, dan F1-score. Hasil penelitian menunjukkan bahwa model yang dikembangkan mampu melakukan klasifikasi gambar dan klasifikasi real-time jenis Philodendron dengan akurasi keseluruhan mencapai 99,2% berdasarkan pengujian menggunakan data uji. Sistem ini diimplementasikan dalam aplikasi Android dan diharapkan dapat menjadi alat bantu identifikasi bagi pelanggan toko tanaman "Galeri Hejo"

Philodendron ornamental plants have many varieties that are often difficult to distinguish due to their visual similarities. This research aims to develop a system for classifying Philodendron plant species based on leaf images using a Convolutional Neural Network (CNN) algorithm to assist the identification process. The research dataset consists of 5000 leaf images from five Philodendron species (Rhaphidophora tetrasperma, Epipremnum pinnatum, Monstera deliciosa, Mayoi, and Amydrium), with 1000 images per species, obtained through direct photography. A CNN model was implemented using the SSD MobileNetV2 FPNLite architecture with a 320x320 pixel input, trained for 50,000 steps. The model was optimized for mobile devices using TensorFlow and converted to TensorFlow Lite. Data was divided into training (80%), validation (10%), and testing (10%) sets. System performance was evaluated using a confusion matrix along with accuracy, precision, recall, and F1-score metrics. The research results show that the developed model is capable of classifying images and performing real-time classification of Philodendron species with an overall accuracy of 99.2% based on testing using the test data. The system is implemented in an Android application and is expected to serve as an identification tool for customers of the “Galeri Hejo” plant store.