IDENTIFIKASI DAN PREDIKSI UMUR, JENIS KELAMIN SERTA DETEKSI EMOSI BERDASARKAN CITRA WAJAH MENGGUNAKAN ALOGARITMA CONVOLUTIONAL NEURAL NETWORK (CNN)

IDENTIFICATION AND PREDICTION OF AGE, GENDER AND EMOTION DETECTION BASED ON FACIAL IMAGES USING CONVOLUTIONAL NEURAL NETWORK ALGORITHM (CNN)

Fitur unik dan khas dari tubuh manusia adalah wajah. Sangat mudah untuk menentukan jenis kelamin, usia, dan ekspresi seseorang melalui wajah. Kemajuan teknologi kecerdasan buatan mencakup penelitian tentang identifikasi citra wajah. Dalam penelitian ini, konsep CNN, DeepFace, dan InceptionResnet V1 digunakan untuk memeriksa dan menerapkan algoritma pengenalan usia, jenis kelamin, dan emosi berdasarkan citra wajah. Hasil dari pengujian pada variasi umur menghasilkan rata-rata 26 tahun, variasi jenis kelamin menghasilkan deteksi perbandingan 6:2 yaitu 6 woman dan 2 man, dan variasi emosi  menghasilkan perbandingan 5:3 yaitu 5 emosi netral sedangkan 3 terdapat emosi senang dan marah. Model ini mencapai nilai loss terendah 0,9701 dan nilai akurasi tertinggi 0,5068. Atau jika dipersentasekan menghasilkan nilai loss terendah 97,01% dan nilai akurasi tertinggi 50,68%.

Kata Kunci : Pengolahan Citra, Wajah, Umur, Gender, Emosi, Convolutional Neural Network

A unique and distinctive feature of the human body is the face. It is easy to determine a person's gender, age, and expression through the face. The advancement of artificial intelligence technology includes research on facial image identification. In this research, the concepts of CNN, DeepFace, and InceptionResnet V1 are used to examine and implement age, gender, and emotion recognition algorithms based on facial images. The results of the test on age variations resulted in an average of 26 years, gender variations resulted in a 6:2 ratio detection, namely 6 women and 2 men, and emotion variations resulted in a 5:3 ratio, namely 5 neutral emotions while 3 there are happy and angry emotions. This model achieves the lowest loss value of 0.9701 and the highest accuracy value of 0.5068. Or if percented, it produces the lowest loss value of 97.01% and the highest accuracy value of 50.68%.

Keywords : Image Processing, Face, Age, Gender, Emotion, Convolutional Neural Network