Enhancing single shot unsupervised domain adaptation for inter-camera person re-identification

Inter-camera person re-identification (re-ID), is the process of identifying people in a surveillance system from various camera perspectives. It involves confirming person identification throughout several cameras and navigating limitations like transferring lighting, converting views, and occlusions all important for safety and monitoring applications. Managing differences in camera angles, occlusions, and illumination may be difficult. These elements may cause mismatches among people, which could decrease the re-ID system's basic efficacy. This research proposed a novel technique to enhance Single Shot Unsupervised Domain Adaptation for Inter-camera Person Re-ID to address this problem, proposed work consists of Preprocessing, and Classification. Initially the preprocessing is applied via Augmentation the use of Cycle GAN, Noise reduction the use of Median Filter, and Enhance Image contrast using Histogram Equalization (HE). Using those preprocessed data, the Siamese Network is trained under the Classification stage. To further enhance the procedure inside the Siamese Network, utilize Conv50 and Conv152. The Python platform is used to develop the suggested model, and performance metrices are used to evaluate the model's effectiveness.