A Deep Learning-Based Cluster Analysis Method for Large-Scale Multi-Label Images

Large-scale multi-label image classification requires determining the presence or absence of a target object in large number sample images. For highly specialized and complex sets, it is especially important to ensure accuracy classification. Traditional deep learning models usually don’t take into account image-label correlation constraints when classifying images, strategy images based only on their own features greatly limits model performance. In this context, paper focuses learning-based cluster analysis method for large-scale We constructed category recognition, which consists global feature extraction module, activation vector generation module an inter-label connection module. Using graph convolutional network (GCN), we aggregated information label nodes structure, while exploring between labels. A detailed description presented how introduce attention mechanism mentioned above recognition. Experimental results have validated effectiveness model.