Method Details

Details for method 'LevelSet R-CNN [COCO]'

name	LevelSet R-CNN [COCO]
challenge	instance-level semantic labeling
details	Obtaining precise instance segmentation masks is of high importance in many modern applications such as robotic manipulation and autonomous driving. Currently, many state of the art models are based on the Mask R-CNN framework which, while very powerful, outputs masks at low resolutions which could result in imprecise boundaries. On the other hand, classic variational methods for segmentation impose desirable global and local data and geometry constraints on the masks by optimizing an energy functional. While mathematically elegant, their direct dependence on good initialization, non-robust image cues and manual setting of hyperparameters renders them unsuitable for modern applications. We propose LevelSet R-CNN, which combines the best of both worlds by obtaining powerful feature representations that are combined in an end-to-end manner with a variational segmentation framework. We demonstrate the effectiveness of our approach on COCO and Cityscapes datasets.
publication	LevelSet R-CNN: A Deep Variational Method for Instance Segmentation Namdar Homayounfar, Yuwen Xiong, Justin Liang*, Wei-Chiu Ma, Raquel Urtasun ECCV 2020 https://arxiv.org/abs/2007.15629
project page / code
used Cityscapes data	fine annotations
used external data	ImageNet, COCO
runtime	n/a
subsampling	no
submission date	March, 2020
previous submissions

Class	AP	AP50%	AP100m	AP50m
person	43.3661	76.2846	60.7964	61.0507
rider	33.8624	72.5066	47.7805	48.7313
car	58.9969	83.1619	77.6188	80.3708
truck	37.6435	49.3553	51.4051	60.1713
bus	49.3671	66.752	68.6348	76.5683
train	39.4466	58.7626	53.2084	60.9755
motorcycle	32.4831	61.7653	41.0268	41.8584
bicycle	24.9368	57.0586	35.4237	35.3439