Deep Image Clustering Based on Label Similarity and Maximizing Mutual Information across Views

Most existing deep image clustering methods use only class-level representations for clustering. However, the representation alone is not sufficient to describe differences between images belonging same cluster. This may lead high intra-class differences, which will harm performance. To address this problem, paper proposes a model named Deep Image Clustering based on Label Similarity and Maximizing Mutual Information Across Views (DCSM). DCSM consists of backbone network, instance-level mapping block. The block learns discriminative features by selecting similar (dissimilar) pairs samples. proposed extended mutual information maximize extracted from views that were obtained using data augmentation as constraint forces capture high-level affect multiple image, thus reducing differences. Four representative datasets are selected our experiments, results show superior current advanced models.