Performance Analysis of GPU-CPU for The Face Detection

The face detection process is one of the research objects in the field of image processing which aims to recognize faces from an image. Performance and face detection accuracy can be measured based on the platform used. In this research, we propose a face detection approach using the Viola-Jones algorithm to perform face detection that is applied to CPU and GPU programs utilizing the architecture of the Compute Unified Device Architecture (CUDA). The image used in this study is an image in RGB format with 25 different image resolution sizes. Test results for each platform are the detection processing time on the CPU is 1433048.72 ms with an accuracy of 90% while the GPU is 233374.583 ms with 72% of accuracy.