Image Classification with Visual Relationship
- Resource Type
- Conference
- Authors
- Li, Yugang; Wang, Yongbin; Chen, Zhe
- Source
- 2019 IEEE/ACIS 18th International Conference on Computer and Information Science (ICIS) Computer and Information Science (ICIS), 2019 IEEE/ACIS 18th International Conference on. :214-219 Jun, 2019
- Subject
- Communication, Networking and Broadcast Technologies
Components, Circuits, Devices and Systems
Computing and Processing
Engineering Profession
Robotics and Control Systems
Visualization
Semantics
Image classification
Genomics
Bioinformatics
Task analysis
Logic gates
image classification
visual relationship
semantic gap
visual search
- Language
While traditional image classification has shown great success propelled by deep convolutional neural networks (CNNs), but computers perform poorly on cognitive tasks, such as connecting computer vision and natural language. In this paper, we propose a framework for image classification base on semantic labels which are generated by visual relationship. The task of visual relationship is to seek the most likely "relationship" between objects in a given image. The proposed method provides a means of extracting the rich meaning embedded in the images. We utilize Long Short-Term Memory (LSTM) to gather the visual relationships of an image to yield a semantic label. Experimental results on Visual Genome and Microsoft COCO demonstrate that the proposed architecture achieves a good performance.