CornerNet: Detecting Objects as Paired Keypoints

ECCV 2018  ·  Hei Law, Jia Deng ·

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. read more

PDF Abstract ECCV 2018 PDF ECCV 2018 Abstract

Datasets


Results from the Paper


Task Dataset Model Metric Name Metric Value Global Rank Result Benchmark
Object Detection COCO minival CornerNet511 (Hourglass-104) box AP 38.4 # 117
AP50 53.8 # 87
AP75 40.9 # 74
APS 18.6 # 72
APM 40.5 # 68
APL 51.8 # 59
Object Detection COCO test-dev CornerNet511 (Hourglass-52, single-scale) box AP 37.8 # 161
AP50 53.7 # 143
AP75 40.1 # 136
APS 17.0 # 135
APM 39.0 # 136
APL 50.5 # 132
Object Detection COCO test-dev CornerNet511 (Hourglass-104, multi-scale) box AP 42.1 # 126
AP50 57.8 # 137
AP75 45.3 # 107
APS 20.8 # 127
APM 44.8 # 108
APL 56.7 # 78

Methods