<![CDATA[This research aims to combine the study of face detection with improvement of image quality in low-light conditions. In this research, we introduce a method that combines Convolutional Neural Networks for image processing to enhance face detection performance in low-light conditions. The proposed method involves pre-processing the images using three image enhancement methods: Deep Lightening Network, Deep Retinex Net, and Signal-to-Noise Ratio Aware. Each of these methods is combined with the face detection method, RetinaFace. The experiment is evaluated using the DARKFACE Dataset, and the performance of each combination is assessed using Average Precision (AP). The combination that yields the best AP value will be determined as the best approach for low-light face detection. The best combination, which utilizes Signal to Noise Ratio Aware for image enhancement and RetinaFace for face detection, achieves an AP score of 52.92%. This result surpasses the face detection performance using the original images from the DARKFACE Dataset, which scored 7.12% in AP. Thus, this experiment demonstrates that image enhancement using Convolutional Neural Networks can significantly improve face detection in low-light conditions]]>
Copyrights © 2024
