CornerNet: Detecting Objects as Paired Keypoints

ECCV 2018 Hei LawJia 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... (read more)

PDF Abstract ECCV 2018 PDF ECCV 2018 Abstract

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 # 51
AP50 53.8 # 44
AP75 40.9 # 37
APS 18.6 # 38
APM 40.5 # 37
APL 51.8 # 32
Object Detection COCO test-dev CornerNet511 (Hourglass-52, single-scale) box AP 37.8 # 72
AP50 53.7 # 76
AP75 40.1 # 73
APS 17.0 # 78
APM 39.0 # 71
APL 50.5 # 67
Object Detection COCO test-dev CornerNet511 (Hourglass-104, multi-scale) box AP 42.1 # 47
AP50 57.8 # 71
AP75 45.3 # 52
APS 20.8 # 70
APM 44.8 # 52
APL 56.7 # 35

Methods used in the Paper