Multi-scale volumes for deep object detection and localization. (January 2017)

• Record Type: Journal Article
• Authors: Ohn-Bar, Eshed; Trivedi, Mohan Manubhai
• Is Part Of: Pattern recognition. Volume 61(2017:Jan.)
• Page Start: 557
• Page End: 572
• Publication Date: 2017-01
• Subjects: Multi-scale reasoning -- Context modeling -- Efficient detection with deep features -- Scale variation handling -- Structured prediction; Pattern perception -- Periodicals; Perception des structures -- Périodiques; Patroonherkenning; 006.4
• DOI: 10.1016/j.patcog.2016.06.002
• Languages: English
• ISSNs: 0031-3203

Abstract

Abstract: This study aims to analyze the benefits of improved multi-scale reasoning for object detection and localization with deep convolutional neural networks. To that end, an efficient and general object detection framework which operates on scale volumes of a deep feature pyramid is proposed. In contrast to the proposed approach, most current state-of-the-art object detectors operate on a single-scale in training, while testing involves independent evaluation across scales. One benefit of the proposed approach is in better capturing of multi-scale contextual information, resulting in significant gains in both detection performance and localization quality of objects on the PASCAL VOC dataset and a multi-view highway vehicles dataset. The joint detection and localization scale-specific models are shown to especially benefit detection of challenging object categories which exhibit large scale variation as well as detection of small objects. Abstract : Highlights: Multi-scale feature reasoning for deep object detection in images is analyzed. A multi-scale contextual reasoning approach is proposed using multi-scale volumes. Scale-specific, joint detection and localization models increase robustness. The approach efficiently handles challenging cases of large variation in scale.