Real Time Multiple Face Recognition using Intra-Layer, Inter-Layer and Hybrid Parallelism

When it comes to biometric applications with image analysis and computer vision, face recognition ranks high and is a very important process. Face Recognition is a technology that identifies or verifies individuals based on facial features using computer vision and machine learning. In different personal and commercial fields many uses of face recognition applications which required faster face recognition in real time. Recognizing multiple faces in real time on high-resolution CCTVs requires a lot of processing power. Rather than using serial process if we use intra-layer and inter-layer (or both hybrid) parallel process, the desired real-time performance can be attained. This research paper presents a multiple face recognition system for large-scale data sets using intra-layer and inter-layer parallelism with their analysis and comparison on the embedded GPU system. Also, we propose combining parallelism strategy with an intra-layer and inter-layer parallelism known as hybrid parallelism as a research outcome of our research work for optimal performance. Our hybrid parallelism experimental results showed that the proposed system can recognize multiple faces time in real time with 20% reduction in training time compared to intra-layer and improved inference time by 43% compared to inter-layer. Also, accuracy increased marginally by 0.4% compared to intra-layer and inter-layer parallelism due to reduced bottlenecks in computation