| Autor: |
Hanifa Fitri, Surfa Yondri, Albar Albar, Ivan Finiel Hotmartua Bagariang, Hendrick, Rahmat Hidayat |
| Rok vydání: |
2020 |
| Předmět: |
|
| Zdroj: |
2020 International Conference on Applied Science and Technology (iCAST). |
| DOI: |
10.1109/icast51016.2020.9557718 |
| Popis: |
Face detection is one of hot issue in image processing combined with Deep Learning methods. The application have influenced in many area such as security, and medical. Mostly face detection only applied in RGB images to locate the face area and continued with face recognition. Thermal camera function is not only to capture the thermal images, but it also applied in security system avoiding the spoofing faces. Nowadays, the temperature measurement is important to do but without any contact with the subject. In this research, we proposed a method to create a face model based on the thermal images. This model will apply in the multi object temperature measurement as real time measurement. YOLO is one of the Deep Learning methods which have best performance in object detection. The main YOLO architecture was formed by Convolutional Neural Network. The YOLO method was applied to create the face model with some modification from previous YOLO architectures. The dataset was built from direct measurement combined with online dataset. FLIR Lepton thermal 3.5 camera was applied in this research to capture subject. The dataset size was extended by using data augmentation to prevent over-fitting during training. By using 1600 images, the face model was successfully created with the average accuracy around 72.7%. |
| Databáze: |
OpenAIRE |
| Externí odkaz: |
|
