Deep Learning for Precise and Efficient Object Detection

Aim and Scopes

Object detection is one of the most challenging and important tasks of computer vision and is widely used in applications such as autonomous vehicle, biometrics, video surveillance, and human-machine interactions. In the past five years, significant success has been achieved with the development of deep learning, especially deep convolutional neural networks. Typical categories of advanced object detection methods are one-stage, two-stage, and anchor-free methods. Nevertheless, the performance in accuracy and efficiency is far from satisfying. On the one hand, the average precision of state-of-the-art object detection methods is very low (e.g., merely about 40% on the COCO dataset). The performance is even worse for small and occluded objects. On the another hand, to obtain precision the detection speed is very low. It is challenging to get a satisfying trade-off between the detection precision and speed. Therefore, much efforts have to be engaged to remarkably improve the performance of object detection in both precision and efficiency.

This special issue will publish papers presenting state-of-the-art methods in dealing with the challenging problems of object detection within the framework of deep learning. We invite authors to submit manuscripts that are highly related to the topics of this special issue and which have not been published before. The topics of interest include, but are not limited to:

•  Anchor and Anchor-free object detection
•  Detecting small or occluded objects
•  Context and attention mechanism for object detection
•  Fast object detection algorithms
•  New backbone for object detection
•  Architecture search for object detection
•  3D object detection
•  Object detection in challenging conditions
•  Handling scale problems in object detection
•  Improving localization accuracy
•  Fusion of point cloud and images for object detection
•  Relationship between object detection and other computer vision tasks.
•  Large-scale datasets for object detection

Important Dates

Submission period: Dec. 1-31, 2020

First notification to authors: Mar. 1, 2021

Submission of revised papers: Apr. 15, 2021

Final notification to authors: June 15, 2021

Online publication: Jul. 1, 2021

Submission of Manuscripts

Prospective authors should write manuscripts according to the Guide for Authors of Pattern Recognition Letters available at the website https://ees.elsevier.com/prletters/. Please use article type name by: VSI:DL4PEOD.

Guest Editors

Dr. Yanwei Pang, Tianjin University, China, pyw@tju.edu.cn, MGE

Dr. Jungong Han, Warwick University, U.K., jungong.han@warwick.ac.uk

Dr. Xin Lu, Adobe Inc., U.S.A., xinl@adobe.com

Dr. Nicola Conci, University of Trento, Italy, nicola.conci@unitn.it