Deep Cross-Media Neural Model for Generating Image Descriptions

  in Special Issue   Posted on June 3, 2020

Information for the Special Issue

Submission Deadline: Thu 10 Sep 2020
Journal Impact Factor : 2.671
Journal Name : Image and Vision Computing
Journal Publisher:
Website for the Special Issue:
Journal & Submission Website:

Special Issue Call for Papers:

Summary and Scope:

Understanding and generating image descriptions (UGID) are hot topics that combines the computer vision (CV) and natural language processing (NLP). UGID has broad application prospects in many fields of AI. Different from coarse-grained image understanding of independent labeling, the image description task needs to learn the natural language descriptions of images. This requires not only the model to recognize the objects in the image, but also other visual elements (e.g., actions and attributes of objects), but also understand the interrelationships between objects and generate human-readable description sentences, which is challenging. The real image understanding is to describe image with natural language and let the machine emulate humans for better human-computer interaction. With the fast development of deep learning in the fields of CV and NLP, the encoder-decoder based deep neural models have obtained breakthrough results in generating image descriptions in cross-media domains. As such, the image understanding may become a reality in future. However, current models can only provide a simple description about image, i.e., the number of descriptive words is usually limited and even the sentences are logically wrong.

In this special issue, we invite the original contributions from diverse research fields, developing new deep cross-media neural model for understanding and generating image descriptions, which aims to reduce the gap between image understanding and natural language descriptions.

The topics of interest include, but are not limited to:

  • Attention guided UGID
  • Visual relationship in UGID
  • Compositional architectures for UGID
  • Multimodal learning for UGID
  • Describing novel objects in UGID
  • Natural language processing model
  • New datasets for UGID
  • Novel encoder-decoder based architecture
  • Deep cross-media neural model with applications of UGID, e.g., early childhood education, medical image analysis, assisted blinding and news automation, etc.

Important Dates:

Paper submission due: Sep 10, 2020

First notification: Nov 10, 2020

Final decision made on all manuscripts: Mar 30, 2021

Managing Guest Editor:

Prof. Zhao Zhang, Hefei University of Technology, China

Other Guest Editors:

  1. Dr. Sheng Li, University of Georgia, USA
  2. Prof. Meng Wang, Hefei University of Technology, China
  3. Prof. Shuicheng Yan, National University of Singapore, Singapore

Other Special Issues on this journal

Closed Special Issues


Reliable Automatic Facial Coding

Image and Vision Computing
Mon 15 May 2017