RepPoints: Point Set Representation for Object Detection

ICCV 2019  ·  Ze Yang, Shaohui Liu, Han Hu, Li-Wei Wang, Stephen Lin ·

Modern object detectors rely heavily on rectangular bounding boxes, such as anchors, proposals and the final predictions, to represent objects at various recognition stages. The bounding box is convenient to use but provides only a coarse localization of objects and leads to a correspondingly coarse extraction of object features. In this paper, we present \textbf{RepPoints} (representative points), a new finer representation of objects as a set of sample points useful for both localization and recognition. Given ground truth localization and recognition targets for training, RepPoints learn to automatically arrange themselves in a manner that bounds the spatial extent of an object and indicates semantically significant local areas. They furthermore do not require the use of anchors to sample a space of bounding boxes. We show that an anchor-free object detector based on RepPoints can be as effective as the state-of-the-art anchor-based detection methods, with 46.5 AP and 67.4 $AP_{50}$ on the COCO test-dev detection benchmark, using ResNet-101 model. Code is available at https://github.com/microsoft/RepPoints.

PDF Abstract ICCV 2019 PDF ICCV 2019 Abstract

Datasets


Results from the Paper


Task Dataset Model Metric Name Metric Value Global Rank Result Benchmark
Object Detection COCO minival RPDet (ResNet-101-DCN, multi-scale) box AP 46.4 # 97
Object Detection COCO minival RPDet (ResNet-50, multi-scale train) box AP 40.8 # 158
Object Detection COCO minival RPDet (ResNeXt-101-DCN) box AP 44.5 # 117
Object Detection COCO minival RPDet (ResNet-101-DCN, multi-scale train) box AP 44.8 # 112
Object Detection COCO minival RPDet (ResNeXt-101-DCN, multi-scale) box AP 46.8 # 95
Object Detection COCO minival RPDet (ResNet-50) box AP 38.6 # 179
Object Detection COCO minival RPDet (ResNet-101) box AP 40.3 # 164
Object Detection COCO test-dev RPDet (ResNet-101-DCN) box mAP 42.8 # 162
AP50 65.0 # 77
AP75 46.3 # 112
APS 24.9 # 95
APM 46.2 # 99
APL 54.7 # 104
Hardware Burden None # 1
Operations per network pass None # 1
Object Detection COCO test-dev RPDet (ResNet-101-DCN, multi-scale) box mAP 46.5 # 120
AP50 67.4 # 58
AP75 50.9 # 69
APS 30.3 # 46
APM 49.7 # 62
APL 57.1 # 79
Hardware Burden None # 1
Operations per network pass None # 1
Object Detection COCO test-dev RPDet (ResNet-101) box mAP 41 # 179
AP50 62.9 # 103
AP75 44.3 # 128
APS 23.6 # 110
APM 44.1 # 117
APL 51.7 # 124
Hardware Burden None # 1
Operations per network pass None # 1

Methods