Modern CNN-based object detectors rely on bounding box regression and
non-maximum suppression to localize objects. While the probabilities for class
labels naturally reflect classification confidence, localization confidence is
absent. This makes properly localized bounding boxes degenerate during
iterative regression or even suppressed during NMS. In the paper we propose
IoU-Net learning to predict the IoU between each detected bounding box and the
matched ground-truth. The network acquires this confidence of localization,
which improves the NMS procedure by preserving accurately localized bounding
boxes. Furthermore, an optimization-based bounding box refinement method is
proposed, where the predicted IoU is formulated as the objective. Extensive
experiments on the MS-COCO dataset show the effectiveness of IoU-Net, as well
as its compatibility with and adaptivity to several state-of-the-art object
detectors.
%0 Generic
%1 citeulike:14632721
%A xxx,
%D 2018
%K detection loss pooling rcnn
%T Acquisition of Localization Confidence for Accurate Object Detection
%U http://arxiv.org/abs/1807.11590
%X Modern CNN-based object detectors rely on bounding box regression and
non-maximum suppression to localize objects. While the probabilities for class
labels naturally reflect classification confidence, localization confidence is
absent. This makes properly localized bounding boxes degenerate during
iterative regression or even suppressed during NMS. In the paper we propose
IoU-Net learning to predict the IoU between each detected bounding box and the
matched ground-truth. The network acquires this confidence of localization,
which improves the NMS procedure by preserving accurately localized bounding
boxes. Furthermore, an optimization-based bounding box refinement method is
proposed, where the predicted IoU is formulated as the objective. Extensive
experiments on the MS-COCO dataset show the effectiveness of IoU-Net, as well
as its compatibility with and adaptivity to several state-of-the-art object
detectors.
@misc{citeulike:14632721,
abstract = {{Modern CNN-based object detectors rely on bounding box regression and
non-maximum suppression to localize objects. While the probabilities for class
labels naturally reflect classification confidence, localization confidence is
absent. This makes properly localized bounding boxes degenerate during
iterative regression or even suppressed during NMS. In the paper we propose
IoU-Net learning to predict the IoU between each detected bounding box and the
matched ground-truth. The network acquires this confidence of localization,
which improves the NMS procedure by preserving accurately localized bounding
boxes. Furthermore, an optimization-based bounding box refinement method is
proposed, where the predicted IoU is formulated as the objective. Extensive
experiments on the MS-COCO dataset show the effectiveness of IoU-Net, as well
as its compatibility with and adaptivity to several state-of-the-art object
detectors.}},
added-at = {2019-02-27T22:23:29.000+0100},
archiveprefix = {arXiv},
author = {xxx},
biburl = {https://www.bibsonomy.org/bibtex/27daf18fc3d04bfc4be2750ed3f90e4f2/nmatsuk},
citeulike-article-id = {14632721},
citeulike-linkout-0 = {http://arxiv.org/abs/1807.11590},
citeulike-linkout-1 = {http://arxiv.org/pdf/1807.11590},
day = 30,
eprint = {1807.11590},
interhash = {7d3b74df41ef9509e9a6d772b22f661e},
intrahash = {7daf18fc3d04bfc4be2750ed3f90e4f2},
keywords = {detection loss pooling rcnn},
month = jul,
posted-at = {2018-09-05 08:52:24},
priority = {4},
timestamp = {2019-02-27T22:23:29.000+0100},
title = {{Acquisition of Localization Confidence for Accurate Object Detection}},
url = {http://arxiv.org/abs/1807.11590},
year = 2018
}