Febrian, Devin Ahmad (2020) Sistem Pemantauan Mahasiswa Dalam Kelas dengan Face Recognition berbasis Deep Neural Network. Other thesis, Institut Teknologi Sepuluh Nopember.
Preview |
Text
02311540000071-Undergraduate_Thesis.pdf Download (2MB) | Preview |
Abstract
Pemantauan mahasiswa di dalam kelas dapat dilakukan dengan mengenali mahasiswa menggunakan Face Recognition Deep Neural Network Sistem terdiri dari model yang akan memprediksi nama mahasiswa berdasarkan video yang diambil oleh kamera di dalam kelas dan komputer yang akan memproses gambar dan menjalankan prediksi model. Model akan dilatih dengan metode Faster Region Convolutional Neural Network dengan arsitektur ResNet-50. Model dilatih dan dites dengan dataset primer berupa gambar dari 6 wajah mahasiswa Teknik Fisika (TF) dan dataset sekunder berupa gambar wajah umum dari WIDER FACE. Model akan dilatih dengan WIDER FACE terlebih dahulu agar model bisa mempelajari bentuk wajah secara umum. Lalu model akan dilatih pada dataset mahasiswa TF dengan menggunakan metode transfer learning. Digunakan 908 gambar untuk pelatihan dan 375 gambar untuk tes dari dataset mahasiswa Teknik Fisika (TF). Dan digunakan 12.880 gambar untuk pelatihan dan 3.226 gambar untuk tes dari dataset WIDER FACE. Model ini mendapatkan mAP 96.52% dan waktu deteksi 1 sekon.
============================================================================================================================
A student monitoring system inside classroom could be done by using Deep Neural Network-based Face Recognition method. This system will consist of model that predict students name by video input taken from camera inside classroom, and computer to process the video and run inference for the model. The model is trained with Faster Region Convolutional Neural Network with ResNet-50 architecture. Training will use data from primary and secondary dataset. Primary dataset contain images of faces from six engineering physics department students. Secondary dataset is taken form WIDER FACE dataset containing images of faces in general. First, model is trained on WIDER FACE dataset to learn faces in general term. Then model will be trained on primary dataset with transfer learning method. Primary dataset use 908 images for training and 375 images for testing. Secondary dataset use 12.880 images for training and 3.226 images for testing. This model achive mAP score of 96.52% with 1 second detection time.
| Item Type: | Thesis (Other) |
|---|---|
| Additional Information: | RSF 006.42 Feb s-1 2020 |
| Uncontrolled Keywords: | Face Recognition, Region Convolution Neural Network, ResNet-50, WIDER FACE |
| Subjects: | T Technology > TA Engineering (General). Civil engineering (General) > TA1650 Face recognition. Optical pattern recognition. |
| Divisions: | Faculty of Industrial Technology and Systems Engineering (INDSYS) > Physics Engineering > 30201-(S1) Undergraduate Thesis |
| Depositing User: | Devin Ahmad Febrian |
| Date Deposited: | 06 Mar 2025 07:23 |
| Last Modified: | 06 Mar 2025 07:23 |
| URI: | http://repository.its.ac.id/id/eprint/74713 |
Actions (login required)
 |
View Item |