Skip to content

Latest commit

 

History

History
42 lines (31 loc) · 3.21 KB

README.md

File metadata and controls

42 lines (31 loc) · 3.21 KB
Raw

SOLO: Segmenting Objects by Locations

Introduction

@inproceedings{wang2020solo,
  title     =  {{SOLO}: Segmenting Objects by Locations},
  author    =  {Wang, Xinlong and Kong, Tao and Shen, Chunhua and Jiang, Yuning and Li, Lei},
  booktitle =  {Proc. Eur. Conf. Computer Vision (ECCV)},
  year      =  {2020}
}

Results and Models

SOLO

Backbone Style MS train Lr schd Mem (GB) Inf time (fps) mask AP Download
R-50 pytorch N 1x 8.0 14.0 33.1 model | log
R-50 pytorch Y 3x 7.4 14.0 35.9 model | log

Decoupled SOLO

Backbone Style MS train Lr schd Mem (GB) Inf time (fps) mask AP Download
R-50 pytorch N 1x 7.8 12.5 33.9 model | log
R-50 pytorch Y 3x 7.9 12.5 36.7 model | log
  • Decoupled SOLO has a decoupled head which is different from SOLO head. Decoupled SOLO serves as an efficient and equivalent variant in accuracy of SOLO. Please refer to the corresponding config files for details.

Decoupled Light SOLO

Backbone Style MS train Lr schd Mem (GB) Inf time (fps) mask AP Download
R-50 pytorch Y 3x 2.2 31.2 32.9 model | log
  • Decoupled Light SOLO using decoupled structure similar to Decoupled SOLO head, with light-weight head and smaller input size, Please refer to the corresponding config files for details.