Cluster Shape Detection Using Dual Attention Induced Convolutional-Clustering Algorithm

Cluster shape detection holds a crucial role in the realms of data analysis, image processing, and pattern recognition, acting as a linchpin for revealing concealed structures and relationships within datasets. This study introduces an innovative approach, the Dual Attention Induced Convolutional-Clustering (DAICC) algorithm, to address cluster shape detection. The uniqueness of this approach lies in its distinctive feature extraction methodology, consisting of two key modules. The first module employs an attention mechanism that accentuates features across both spatial and channel dimensions within the feature space. Subsequently, a convolution operation in the second module abstracts high-level features. Following feature extraction, a Fuzzy C-means clustering approach categorizes data points into distinct clusters. Subsequent to the clustering process, a novel algorithm for cluster shape detection is employed. This algorithm is characterized by its simplicity, computational efficiency, and effectiveness in identifying the shapes of clusters. Performance analysis underscores the superior efficacy of the proposed technique compared to state-of-the-art methods. This innovative approach holds promise as a valuable tool for iden-tifying cluster shapes in diverse applications, including image segmentation, medical fields (such as discerning the shapes of different microbes), and architectural contexts (e.g., identifying various aspects of urban structures).