Sistem Deteksi Ekspresi Siswa Dalam E-Learning Menggunakan Metode Convolutional Neural Network (CNN)

Student Expression Detection System in E-Learning Using the Convolutional Neural Network (CNN) Method

Proses pembelajaran jarak jauh memiliki keunggulan seperti memperoleh fleksibilitas saat belajar dalam waktu dan tempat yang berbeda. Akan tetapi proses pembelajaran dari jarak jauh memiliki kekurangan yaitu guru tidak dapat memantau siswa mengenai antusiasme siswa dalam proses belajar. Dengan menggunakan sistem deteksi ekspresi siswa pada saat proses pembelajaran guru dapat memantau siswa. Sistem deteksi ekspresi siswa dalam e-learning menggunakan metode Convolutional Neural Network (CNN) dan menggunakan dataset dari Kaggle yaitu The Facial Expression Recognition 2013 (FER-2013). FER-2013 terdapat tujuh kategori emosi yaitu marah, jijik, takut, senang, sedih, terkejut dan biasa. Sistem deteksi ekspresi siswa akan menganalisis emosi siswa pada saat proses pembelajaran berlangsung. Kemudian sistem deteksi ekspresi akan menunjukan nilai presentase ekspresi positif ataupun negatif dan kondisi emosi siswa. Setelah itu, sistem akan menyimpan hasil dari deteksi ekspresi berupa video dan dokumen teks. Hasil pengujian dari sistem deteksi ekspresi siswa dalam e-learning menggunakan metode CNN  dengan menggunakan arsitektur AlexNet dapat mengklasifikasi ekspresi wajah  dan didapatkan hasil training accuracy 94,81%, training loss 15,30%. Pada pengujian model CNN menggunakan arsitektur LeNet dapat mengklasifikasi wajah dan didapatkan hasil training accuracy 98,84%, training loss 6,92%.

The distance learning process has advantages such as gaining flexibility when studying in different times and places. However, the distance learning process has a drawback, namely that teachers cannot monitor students regarding their enthusiasm in the learning process. By using a student expression detection system during the learning process teachers can monitor students. The student expression detection system in e-learning uses the Convolutional Neural Network (CNN) method and uses a dataset from Kaggle, namely The Facial Expression Recognition 2013 (FER-2013). FER-2013 has seven categories of emotions, namely anger, disgust, fear, happiness, sadness, surprise and normal. The student expression detection system will analyze students' emotions during the learning process. Then the expression detection system will show the percentage value of positive or negative expressions and the student's emotional condition. After that, the system will save the results of expression detection in the form of videos and text documents. The test results of the student expression detection system in e-learning using the CNN method using the AlexNet architecture were able to classify facial expressions and obtained training accuracy results of 94.81%, training loss of 15.30%. In testing the CNN model using the LeNet architecture it was able to classify faces and obtained training accuracy results of 98.84%, training loss of 6.92%.