papers AI Learner
The Github is limit! Click to go to the new site.

MetaAnchor: Learning to Detect Objects with Customized Anchors

2018-11-06
Tong Yang, Xiangyu Zhang, Zeming Li, Wenqiang Zhang, Jian Sun

Abstract

We propose a novel and flexible anchor mechanism named MetaAnchor for object detection frameworks. Unlike many previous detectors model anchors via a predefined manner, in MetaAnchor anchor functions could be dynamically generated from the arbitrary customized prior boxes. Taking advantage of weight prediction, MetaAnchor is able to work with most of the anchor-based object detection systems such as RetinaNet. Compared with the predefined anchor scheme, we empirically find that MetaAnchor is more robust to anchor settings and bounding box distributions; in addition, it also shows the potential on transfer tasks. Our experiment on COCO detection task shows that MetaAnchor consistently outperforms the counterparts in various scenarios.

Abstract (translated by Google)
URL

https://arxiv.org/abs/1807.00980

PDF

https://arxiv.org/pdf/1807.00980


Similar Posts

Comments