Person Re-Identification using Deep Learning

Re-identification of person (re-ID) is a computer vision task, which is trying to find out the same individual’s pictures through a single camera scenario. For addressing Re-ID, we have proposed a deep learning architecture i.e. ResNet, and open Visual Inference and Neural Optimization (Open VINO) toolkit from intel. In this paper, we have utilized the probable feature of learning in the ResNet model that is instructed on enormous datasets and fine-turned on strenuous market-1501 re-ID standard dataset. This process implicates data preprocessing, model selection, training, and optimization steps that are important for acceptable results. We have used data augmentation and triplet loss function to construct implanting more selective as a final step in this pipeline and our developed ResNet model is supported by OpenVINO for hardware-specific optimizations and model conversion. Our Solution on ResNet delivers real-time processing speeds using GPU. This approach shows that is possible and easy to deploy precise person Re-Id for surveillance, forensic identification, and security monitoring.