Skip to content

Latest commit

 

History

History

ann

Folders and files

NameName
Last commit message
Last commit date

parent directory

..
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 

ANN

Asymmetric Non-local Neural Networks for Semantic Segmentation

Introduction

Official Repo

Code Snippet

Abstract

The non-local module works as a particularly useful technique for semantic segmentation while criticized for its prohibitive computation and GPU memory occupation. In this paper, we present Asymmetric Non-local Neural Network to semantic segmentation, which has two prominent components: Asymmetric Pyramid Non-local Block (APNB) and Asymmetric Fusion Non-local Block (AFNB). APNB leverages a pyramid sampling module into the non-local block to largely reduce the computation and memory consumption without sacrificing the performance. AFNB is adapted from APNB to fuse the features of different levels under a sufficient consideration of long range dependencies and thus considerably improves the performance. Extensive experiments on semantic segmentation benchmarks demonstrate the effectiveness and efficiency of our work. In particular, we report the state-of-the-art performance of 81.3 mIoU on the Cityscapes test set. For a 256x128 input, APNB is around 6 times faster than a non-local block on GPU while 28 times smaller in GPU running memory occupation. Code is available at: this https URL.

Citation

@inproceedings{zhu2019asymmetric,
  title={Asymmetric non-local neural networks for semantic segmentation},
  author={Zhu, Zhen and Xu, Mengde and Bai, Song and Huang, Tengteng and Bai, Xiang},
  booktitle={Proceedings of the IEEE/CVF International Conference on Computer Vision},
  pages={593--602},
  year={2019}
}

Results and models

Cityscapes

Method Backbone Crop Size Lr schd Mem (GB) Inf time (fps) mIoU mIoU(ms+flip) config download
ANN R-50-D8 512x1024 40000 6 3.71 77.40 78.57 config model | log
ANN R-101-D8 512x1024 40000 9.5 2.55 76.55 78.85 config model | log
ANN R-50-D8 769x769 40000 6.8 1.70 78.89 80.46 config model | log
ANN R-101-D8 769x769 40000 10.7 1.15 79.32 80.94 config model | log
ANN R-50-D8 512x1024 80000 - - 77.34 78.65 config model | log
ANN R-101-D8 512x1024 80000 - - 77.14 78.81 config model | log
ANN R-50-D8 769x769 80000 - - 78.88 80.57 config model | log
ANN R-101-D8 769x769 80000 - - 78.80 80.34 config model | log

ADE20K

Method Backbone Crop Size Lr schd Mem (GB) Inf time (fps) mIoU mIoU(ms+flip) config download
ANN R-50-D8 512x512 80000 9.1 21.01 41.01 42.30 config model | log
ANN R-101-D8 512x512 80000 12.5 14.12 42.94 44.18 config model | log
ANN R-50-D8 512x512 160000 - - 41.74 42.62 config model | log
ANN R-101-D8 512x512 160000 - - 42.94 44.06 config model | log

Pascal VOC 2012 + Aug

Method Backbone Crop Size Lr schd Mem (GB) Inf time (fps) mIoU mIoU(ms+flip) config download
ANN R-50-D8 512x512 20000 6 20.92 74.86 76.13 config model | log
ANN R-101-D8 512x512 20000 9.5 13.94 77.47 78.70 config model | log
ANN R-50-D8 512x512 40000 - - 76.56 77.51 config model | log
ANN R-101-D8 512x512 40000 - - 76.70 78.06 config model | log