Multiple attentional pyramid networks for Chinese herbal recognition

Chinese herbs play a critical role in Traditional Medicine. Due to different recognition granularity, they can be recognized accurately only by professionals with much experience. It is expected that automatically using new techniques like machine learning. However, there no herbal image dataset available. Simultaneously, learning method which deal well. Therefore, this paper begins building standard Chinese-Herbs dataset. Subsequently, Attentional Pyramid Networks (APN) for proposed, where both novel competitive attention and spatial collaborative are proposed then applied. APN adaptively model images feature scales. Finally, framework as application of APN. Experiments conducted on our constructed validate the effectiveness methods.