You always look again: Learning to detect the unseen objects

Author(s)

• Khanh-Duy Nguyen, Vietnam National UniversityFollow
• Khang Nguyen, Vietnam National University
• Duy-Dinh Le, Vietnam National University
• Duc Anh Duong, Vietnam National University
• Tam V. Nguyen, University of DaytonFollow

Document Type

Article

Publication Date

2-21-2019

Publication Source

Journal of Visual Communication and Image Representation

Abstract

Object detection has always attracted a lot of attention in computer vision due to its practical applications, i.e., robotics engineering, autonomous vehicles, and surveillance systems. Recently deep learning approaches have successfully improved the performance of object detection by a significant amount. However, there exist many challenging objects in the images that state-of-the-art approaches still fail to detect. In this paper, we propose an efficient approach that intentionally learns to detect the unseen (missing) objects. In particular, we utilize a dual-level of deep networks to efficiently detect difficult objects in images. The extensive experiments on three benchmarking datasets, PASCAL VOC, KITTI, and MS-COCO, show the superiority of our approach over the state-of-the-art methods.

Inclusive pages

206-216

ISBN/ISSN

1047-3203

Publisher

Elsevier

Volume

60

Keywords

Deep learning, Dual-level deep networks, Object detection

eCommons Citation

Nguyen, Khanh-Duy; Nguyen, Khang; Le, Duy-Dinh; Duong, Duc Anh; and Nguyen, Tam V., "You always look again: Learning to detect the unseen objects" (2019). Computer Science Faculty Publications. 229.
https://ecommons.udayton.edu/cps_fac_pub/229