Object detection methods fall into two categories, i.e., two-stage and
single-stage detectors. The former is characterized by high detection accuracy
while the latter usually has considerable inference speed. Hence, it is
imperative to fuse their metrics for a better accuracy vs. speed trade-off. To
this end, we propose a dual refinement network (Dual-RefineDet) to boost the
performance of the single-stage detector. Inheriting from advantages of the
two-stage approach (i.e., two-step regression and accurate features for
detection), anchor refinement and feature offset refinement are conducted in
anchor-offset detection, where the detection head is comprised of deformable
convolutions. Moreover, to leverage contextual information for describing
objects, we design a multi-deformable head, in which multiple detection paths
with different respective field sizes devote themselves to detecting objects.
Extensive experiments on PASCAL VOC datasets are conducted, and we achieve the
state-of-the-art results and a better accuracy vs. speed trade-off, i.e.,
\$81.3\%\$ mAP vs. \$42.3\$ FPS with \$320320\$ input image on VOC2007
dataset. Codes will be made publicly available.
%0 Generic
%1 citeulike:14628428
%A xxx,
%D 2018
%K arch deformable detection pooling ssd
%T Dual Refinement Network for Single-Shot Object Detection
%U http://arxiv.org/abs/1807.08638
%X Object detection methods fall into two categories, i.e., two-stage and
single-stage detectors. The former is characterized by high detection accuracy
while the latter usually has considerable inference speed. Hence, it is
imperative to fuse their metrics for a better accuracy vs. speed trade-off. To
this end, we propose a dual refinement network (Dual-RefineDet) to boost the
performance of the single-stage detector. Inheriting from advantages of the
two-stage approach (i.e., two-step regression and accurate features for
detection), anchor refinement and feature offset refinement are conducted in
anchor-offset detection, where the detection head is comprised of deformable
convolutions. Moreover, to leverage contextual information for describing
objects, we design a multi-deformable head, in which multiple detection paths
with different respective field sizes devote themselves to detecting objects.
Extensive experiments on PASCAL VOC datasets are conducted, and we achieve the
state-of-the-art results and a better accuracy vs. speed trade-off, i.e.,
\$81.3\%\$ mAP vs. \$42.3\$ FPS with \$320320\$ input image on VOC2007
dataset. Codes will be made publicly available.
@misc{citeulike:14628428,
abstract = {{Object detection methods fall into two categories, i.e., two-stage and
single-stage detectors. The former is characterized by high detection accuracy
while the latter usually has considerable inference speed. Hence, it is
imperative to fuse their metrics for a better accuracy vs. speed trade-off. To
this end, we propose a dual refinement network (Dual-RefineDet) to boost the
performance of the single-stage detector. Inheriting from advantages of the
two-stage approach (i.e., two-step regression and accurate features for
detection), anchor refinement and feature offset refinement are conducted in
anchor-offset detection, where the detection head is comprised of deformable
convolutions. Moreover, to leverage contextual information for describing
objects, we design a multi-deformable head, in which multiple detection paths
with different respective field sizes devote themselves to detecting objects.
Extensive experiments on PASCAL VOC datasets are conducted, and we achieve the
state-of-the-art results and a better accuracy vs. speed trade-off, i.e.,
\$81.3\%\$ mAP vs. \$42.3\$ FPS with \$320\times 320\$ input image on VOC2007
dataset. Codes will be made publicly available.}},
added-at = {2019-02-27T22:23:29.000+0100},
archiveprefix = {arXiv},
author = {xxx},
biburl = {https://www.bibsonomy.org/bibtex/254147df96edafa66b548c45bcee105e0/nmatsuk},
citeulike-article-id = {14628428},
citeulike-linkout-0 = {http://arxiv.org/abs/1807.08638},
citeulike-linkout-1 = {http://arxiv.org/pdf/1807.08638},
day = 23,
eprint = {1807.08638},
interhash = {d3e2a422efa548b1145e7109f25b4361},
intrahash = {54147df96edafa66b548c45bcee105e0},
keywords = {arch deformable detection pooling ssd},
month = jul,
posted-at = {2018-08-23 13:10:02},
priority = {4},
timestamp = {2019-02-27T22:23:29.000+0100},
title = {{Dual Refinement Network for Single-Shot Object Detection}},
url = {http://arxiv.org/abs/1807.08638},
year = 2018
}