CornerNet: Detecting Objects as Paired Keypoints

Hei Law, Jia Deng
8/3/2018
cs.CV

Official code: princeton-vl/CornerNet (2,371 stars, BSD-3-Clause, last commit 2020-09-18).

Abstract

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.

View on arXivView PDF

Code Implementations(1)

princeton-vl/CornerNetOfficial100%
BSD-3-Clause
2,371463C++, PythonAug 3, 20186 years agoBSD-3-Clause

Similar Papers

CenterNet: Keypoint Triplets for Object Detection

86% match1 repo1.9k2019

FCOS: Fully Convolutional One-Stage Object Detection

86% match2 repos3.3k2019

Anchor-Intermediate Detector: Decoupling and Coupling Bounding Boxes for Accurate Object Detection

86% match1 repo122023

SSD: Single Shot MultiBox Detector

85% match1 repo4.8k2015

CenterNet++ for Object Detection

84% match1 repo2102022

Faster R-CNN: Towards Real-Time Object Detection with Region Proposal Networks

84% match2 repos8.3k2015

Cite this paper

@article{law2018cornernet,
  title  = {CornerNet: Detecting Objects as Paired Keypoints},
  author = {Hei Law and Jia Deng},
  year   = {2018},
  eprint = {1808.01244},
  archivePrefix = {arXiv},
  url    = {https://arxiv.org/abs/1808.01244}
}

Add a badge to your README

Shows whether official code is linked for this paper.

Code of Paper badge
[![Code of Paper](https://codeofpaper.com/badge/1808.01244.svg)](https://codeofpaper.com/paper/1808.01244)

Discussion