Cross-modal multi-label image classification modeling and recognition based on nonlinear

Abstract Recently, it has become a popular strategy in multi-label image recognition to predict those labels that co-occur picture. Previous work concentrated on capturing label correlation but neglected correctly fuse picture features and embeddings, which substantial influence the model’s convergence efficiency restricts future accuracy improvement. In order better classify labeled training samples of corresponding categories field classification, cross-modal classification modeling method based nonlinear is proposed. Multi-label models deep convolutional neural networks are constructed respectively. The visual model uses natural images simple biomedical with single achieve heterogeneous transfer learning homogeneous learning, general proprietary field, while text description learning. experimental results show combining two modes can obtain hamming loss similar best performance evaluation task, macro average F 1 value increases from 0.20 0.488, about 52.5% higher. algorithm alleviate problem overfitting most classes retrieval performance. addition, effectiveness rationality mapping techniques verified.