Image Classification Using Federated Learning

Multiple users can jointly train a universal model using federated learning, a revolutionary AI approach, without having to reveal their personal data. With this strategy, anonymity is maintained while a secure learning environment is guaranteed. In contrast to the Workplace 31 dataset, this study presents a brand-new dataset made up of 23,326 photos obtained from eight different corporate sources and painstakingly organised into 31 classifications. It is noteworthy that this dataset is the first image classification dataset created specifically for federated learning. Along with the dataset, we also suggest the revolutionary federated learning algorithms Handled Cyclic and Handled Star. A cycle progression is established via Handled cycle, whereby a user gets weights from a previous user, adjusts them through localised training, and then transmits them to the following user. The essential job of image categorization within the context of computer vision is considerably improved by this novel approach, which highlights the development of federated learning. Traditional centralised deep learning model training requires centralised access to enormous datasets, which raises questions about data security and privacy. Federated learning addresses these issues by facilitating dispersed device collaboration for model training while restricting raw data exposure. This paper highlights the crucial role that federated learning has had in furthering AI research by giving an overview of its use in the context of image categorization.