Duplicate Instance Identification In Multiview And Multiscale Systems

Visual content understanding has seen significant advancement in recent years, with researchers having easy access to superior computing power and infrastructure to train deep learning models. Organizations are further using multiple cameras to collect additional data for developing robust image content analysis solutions, across domains. However, such initiatives have introduced a lot of duplicate data and redundancy. This poses a challenge when a consolidated decision must be taken at an object level. In this paper, we propose a solution to identify duplicate instances in a multicamera system which captures images of the same instance, at varied viewing angles and scales. The duplicate instances are identified at a pixel level by using a combination of deep learning algorithms and computer vision geometry. We also devise a similarity index to quantify the extent of similarity between a pair of instances. Further, we present our solution customized for duplicate damage instance identification in vehicles, for the auto finance and auto insurance industry.