Libra R-CNN: Towards Balanced Learning for Object Detection

CVPR 2019 Jiangmiao PangKai ChenJianping ShiHuajun FengWanli OuyangDahua Lin

Compared with model architectures, the training process, which is also crucial to the success of detectors, has received relatively less attention in object detection. In this work, we carefully revisit the standard training practice of detectors, and find that the detection performance is often limited by the imbalance during the training process, which generally consists in three levels - sample level, feature level, and objective level... (read more)

PDF Abstract

Evaluation Results from the Paper


TASK DATASET MODEL METRIC NAME METRIC VALUE GLOBAL RANK COMPARE
Object Detection COCO minival Libra R-CNN (ResNet-50 FPN) box AP 38.5 # 42
Object Detection COCO minival Libra R-CNN (ResNet-50 FPN) AP50 59.3 # 21
Object Detection COCO minival Libra R-CNN (ResNet-50 FPN) AP75 42.0 # 23
Object Detection COCO minival Libra R-CNN (ResNet-50 FPN) APS 22.9 # 23
Object Detection COCO minival Libra R-CNN (ResNet-50 FPN) APM 42.1 # 27
Object Detection COCO minival Libra R-CNN (ResNet-50 FPN) APL 50.5 # 26
Object Detection COCO test-dev Libra R-CNN (ResNeXt-101-FPN) box AP 43.0 # 30
Object Detection COCO test-dev Libra R-CNN (ResNeXt-101-FPN) AP50 64 # 21
Object Detection COCO test-dev Libra R-CNN (ResNeXt-101-FPN) AP75 47 # 21
Object Detection COCO test-dev Libra R-CNN (ResNeXt-101-FPN) APS 25.3 # 25
Object Detection COCO test-dev Libra R-CNN (ResNeXt-101-FPN) APM 45.6 # 26
Object Detection COCO test-dev Libra R-CNN (ResNeXt-101-FPN) APL 54.6 # 26