Sistema de detecção de cores em uniformes

Abstract

This research is relevant because it proposes to work in a growing field. Therefore, it becomes possible the application of this technology in several areas, such as industry, commerce, among others. In this way, we seek to develop a system to identify uniforms by color in order to contribute to the speed of organizational processes. Furthermore, the project is based on the development of four systems, namely: Training System, Processing System, General System and User Interface. For the training of the machine learning algorithm, it was used transfer learning techniques from a pre-trained model of the DETR for T-shirt identification. Then K-Means, a data clustering algorithm, was used, with the goal of detecting the predominant color of the T-shirt found by the identification model. In addition, an interface was made to provide a way for the user to have contact with the identification processes, and backends systems that handle these processes. Finally, the applications were deployed in a cloud and local environment, enabling users to test their use. Among the research results it was possible to develop a uniform and color identification system. Moreover, it was also observed points for optimization in future works. It is concluded, through the data collected, that the system met expectations, given the good performance of identification.