Learning Compact Representations for Scalable Visual Recognition and Retrieval

Motivations and Topics:

With the explosive growth of visual data, traditional hand-crafted features or learning-based representations will induce inapplicable computational complexity and large memory costs, due to exhausting computations in large-scale and high-dimensional feature space. Therefore, these conventional methods are lack of scalability for large-scale visual applications, e.g. image/video recognition and retrieval. It is highly desirable to learn much more compact feature representations to reduce computational loads for massive visual data and make big data analysis more feasible. In recent years, compact feature learning algorithms have been intensively exploited and attracted broad interests in visual analysis. For instance, benefiting from the hashing technique, we can obtain compact binary representations, based on which efficient XOR computations in the Hamming space can operate in constant time. The above compact feature learning approaches have been proved to achieve promising performance in various large-scale visual applications, such as content-based image/video retrieval and recognition. In addition, these techniques will be essential for the applications on portable/wearable devices.

The special issue will focus on the most recent progress on compact representation learning for a variety of large-scale visual data analysis, such as content-based image/video retrieval, image/video recognition, annotation and segmentation, object detection and recognition, visual processing and affective computing. This special issue will also target on related feature selection, subspace learning and deep learning techniques, which can well handle large-scale visual tasks. The primary objective of this special issue fosters focused attention on the latest research progress in this interesting area.

The special issue seeks for original contributions of work, which addresses the challenges from the compact representation learning and the related efficient representation learning algorithms for large-scale visual data. The list of possible topics includes, but not limited to:

• Novel compact representation learning algorithms
• Large-scale visual (image, video) indexing algorithms
• Learning-based or data-dependent binary coding/hashing methods
• Novel vector quantization algorithms
• Visual recognition (e.g., detection, categorization, segmentation) with discriminative representation learning techniques
  • Compact feature learning for object classification/detection/retrieval/tracking
  • Novel applications of compact representation learning
  • Deep learning techniques for compact representation learning
• Efficient feature extraction methods for visual data analysis
• Efficient learning algorithms for visual data representation

Submission Guideline

Authors should prepare their manuscript according to the Guide for Authors available from the online submission page of the Pattern Recognition Letters journal at https://www.elsevier.com/journals/pattern-recognition-letters/0167-8655/guide-for-authors/. All the papers will be peer-reviewed following the Pattern Recognition Letters reviewing procedures. When submitting their papers through the online system, Authors should select the acronym “SI:LCR4SVRR” to make it clear that they are submitting to this SI.

The proposed schedule is given below:

• Paper submission due: Jan. 31, 2018
• First review notification: March. 31, 2018
• Revision submission : May, 15, 2018
• Second review notification: June, 30, 2018
• Acceptance notification: Aug. 15, 2018

Managing Guest Editor:

• Dr. Li Liu, University of East Anglia, UK (liuli1213@gmail.com)

Guest Editors:

• Dr. Fumin Shen, University of Electronic Science and Technology of China, China (fumin.shen@gmail.com)
• Prof. Guodong Guo, West Virginia University, USA (guodong.guo@mail.wvu.edu)