knet

K-Net

K-Net: Towards Unified Image Segmentation

Introduction

Official Repo

Code Snippet

Abstract

Semantic, instance, and panoptic segmentations have been addressed using different and specialized frameworks despite their underlying connections. This paper presents a unified, simple, and effective framework for these essentially similar tasks. The framework, named K-Net, segments both instances and semantic categories consistently by a group of learnable kernels, where each kernel is responsible for generating a mask for either a potential instance or a stuff class. To remedy the difficulties of distinguishing various instances, we propose a kernel update strategy that enables each kernel dynamic and conditional on its meaningful group in the input image. K-Net can be trained in an end-to-end manner with bipartite matching, and its training and inference are naturally NMS-free and box-free. Without bells and whistles, K-Net surpasses all previous published state-of-the-art single-model results of panoptic segmentation on MS COCO test-dev split and semantic segmentation on ADE20K val split with 55.2% PQ and 54.3% mIoU, respectively. Its instance segmentation performance is also on par with Cascade Mask R-CNN on MS COCO with 60%-90% faster inference speeds. Code and models will be released at this https URL.

Results and models

ADE20K

Method	Backbone	Crop Size	Lr schd	Mem (GB)	Inf time (fps)	Device	mIoU	mIoU(ms+flip)	config	download
KNet + FCN	R-50-D8	512x512	80000	7.01	19.24	V100	43.60	45.12	config	model | log
KNet + PSPNet	R-50-D8	512x512	80000	6.98	20.04	V100	44.18	45.58	config	model | log
KNet + DeepLabV3	R-50-D8	512x512	80000	7.42	12.10	V100	45.06	46.11	config	model | log
KNet + UperNet	R-50-D8	512x512	80000	7.34	17.11	V100	43.45	44.07	config	model | log
KNet + UperNet	Swin-T	512x512	80000	7.57	15.56	V100	45.84	46.27	config	model | log
KNet + UperNet	Swin-L	512x512	80000	13.5	8.29	V100	52.05	53.24	config	model | log
KNet + UperNet	Swin-L	640x640	80000	13.54	8.29	V100	52.21	53.34	config	model | log

Note:

• All experiments of K-Net are implemented with 8 V100 (32G) GPUs with 2 samplers per GPU.

Citation

@inproceedings{zhang2021knet,
    title={{K-Net: Towards} Unified Image Segmentation},
    author={Wenwei Zhang and Jiangmiao Pang and Kai Chen and Chen Change Loy},
    year={2021},
    booktitle={NeurIPS},
}