Integrating multi-level deep learning and concept ontology for large-scale visual recognition
作者：
Highlights：
• A multi-level deep learning algorithm is developed for large-scale visual.
• Multiple deep networks are learned simultaneously to achieve more discriminative.
• The concept ontology is constructed for organizing large numbers of object classes hierarchically.
• An end-to-end process is developed to learn the deep networks and the tree classifier jointly.
• Our experiments are performed on multiple image sets for algorithm evaluation.
摘要
•A multi-level deep learning algorithm is developed for large-scale visual.•Multiple deep networks are learned simultaneously to achieve more discriminative.•The concept ontology is constructed for organizing large numbers of object classes hierarchically.•An end-to-end process is developed to learn the deep networks and the tree classifier jointly.•Our experiments are performed on multiple image sets for algorithm evaluation.
论文关键词：Large-scale visual recognition,Multi-level deep learning,Multiple deep networks,Concept ontology,Multi-task learning,Tree classifier
论文评审过程：Received 26 January 2017, Revised 18 January 2018, Accepted 24 January 2018, Available online 31 January 2018, Version of Record 5 February 2018.
论文官网地址：https://doi.org/10.1016/j.patcog.2018.01.027
