TY - JOUR
T1 - CornerNet
T2 - Detecting Objects as Paired Keypoints
AU - Law, Hei
AU - Deng, Jia
N1 - Funding Information:
This work is partially supported by a Grant from Toyota Research Institute and a DARPA Grant FA8750-18-2-0019. This article solely reflects the opinions and conclusions of its authors.
Publisher Copyright:
© 2019, Springer Science+Business Media, LLC, part of Springer Nature.
PY - 2020/3/1
Y1 - 2020/3/1
N2 - We propose CornerNet, a new approach to object detection where we detect an object bounding box as a pair of keypoints, the top-left corner and the bottom-right corner, using a single convolution neural network. By detecting objects as paired keypoints, we eliminate the need for designing a set of anchor boxes commonly used in prior single-stage detectors. In addition to our novel formulation, we introduce corner pooling, a new type of pooling layer that helps the network better localize corners. Experiments show that CornerNet achieves a 42.2% AP on MS COCO, outperforming all existing one-stage detectors.
AB - We propose CornerNet, a new approach to object detection where we detect an object bounding box as a pair of keypoints, the top-left corner and the bottom-right corner, using a single convolution neural network. By detecting objects as paired keypoints, we eliminate the need for designing a set of anchor boxes commonly used in prior single-stage detectors. In addition to our novel formulation, we introduce corner pooling, a new type of pooling layer that helps the network better localize corners. Experiments show that CornerNet achieves a 42.2% AP on MS COCO, outperforming all existing one-stage detectors.
KW - Associative embedding
KW - Hourglass network
KW - Object detection
UR - http://www.scopus.com/inward/record.url?scp=85070320598&partnerID=8YFLogxK
UR - http://www.scopus.com/inward/citedby.url?scp=85070320598&partnerID=8YFLogxK
U2 - 10.1007/s11263-019-01204-1
DO - 10.1007/s11263-019-01204-1
M3 - Article
AN - SCOPUS:85070320598
SN - 0920-5691
VL - 128
SP - 642
EP - 656
JO - International Journal of Computer Vision
JF - International Journal of Computer Vision
IS - 3
ER -