Code Monkey home page Code Monkey logo

openseg.pytorch's Introduction

openseg.pytorch

PWC

PWC

PWC

PWC

PWC

News

  • 2020/01/13 The source code for reproduced HRNet+OCR has been made public.

  • 2020/01/09 "HRNet + OCR + SegFix" achieves Rank#1 on Cityscapes leaderboard with mIoU as 84.5%.

  • 2020/01/07 "HRNet+OCR[Mapillary+Coarse]" currently achieves 84.26% on Cityscapes test with better Mapillary pretraining, where we pretrain the HRNet+OCR model on the original Mapillary training set and achieve 50.8% on Mapillary val set. We can expect higher performance with various improvements, e.g., ASP-OCR, larger batch size/crop size (as in Panoptic-DeepLab) and our novel post-processing mechanism.

  • 2020/01/03 "HRNet+OCR" will be made open-source in the code-base HRNet-Semantic-Segmentation very soon, thanks for your patience.

  • 2020/01/02 Please email us ([email protected]) if you need the code for our OCR module and we would like to share it with you ASAP. We also hope you could try our method in your own code base and share the results with us.

  • 2019/11/19 We have updated the paper OCR. Our approach achieves 83.7% and we can further achieve 84.0% on Cityscapes test set with a novel yet simple model-agnostic post-processing scheme. Our model-agnostic post-processing scheme is a new work under progress, which can be applied to improve the results of any existing approaches without any re-training or fine-tuning.

  • 2019/09/25 We have released the paper OCR, which is method of our Rank#2 entry to the leaderboard of Cityscapes.

  • 2019/07/31 We have released the paper ISA, which is very easy to use and implement while being much more efficient than OCNet or DANet based on conventional self-attention.

  • 2019/07/23 We (HRNet + OCR w/ ASP) achieve Rank#1 on the leaderboard of Cityscapes (with a single model) on 3 of 4 metrics.

  • 2019/06/19 We achieve 83.3%+ on the leaderboard of Cityscapes test set based on single model HRNetV2 + OCR. Cityscapes leaderboard We achieve 56.02% on the leaderboard of ADE20K test set based on single model ResNet101 + OCR without any bells or whistles. ADE20K leaderboard

  • 2019/05/27 We achieve SOTA on 6 different semantic segmentation benchmarks including: Cityscapes, ADE20K, LIP, Pascal-Context, Pascal-VOC, COCO-Stuff. We provide the source code for our approach on all the six benchmarks.

Citation

Please consider citing our work if you find it helps you,

@article{yuan2018ocnet,
  title={Ocnet: Object context network for scene parsing},
  author={Yuan Yuhui and Wang Jingdong},
  journal={arXiv preprint arXiv:1809.00916},
  year={2018}
}

@article{huang2019isa,
  title={Interlaced Sparse Self-Attention for Semantic Segmentation},
  author={Huang Lang and Yuan Yuhui and Guo Jianyuan and Zhang Chao and Chen Xilin and Wang Jingdong},
  journal={arXiv preprint arXiv:1907.12273},
  year={2019}
}

@article{yuan2019ocr,
  title={Object-Contextual Representations for Semantic Segmentation},
  author={Yuan Yuhui and Chen Xilin and Wang Jingdong},
  journal={arXiv preprint arXiv:1909.11065},
  year={2019}
}

Performances with openseg.pytorch

  • Cityscapes (testing with single scale whole image)
Methods Backbone Train. mIOU Val. mIOU Test. mIOU BS Iters
FCN MobileNetV2 - - - - -
FCN 3x3-ResNet101 - - - 8 4W
FCN Wide-ResNet38 - - - 8 4W
FCN HRNetV2-48 - - - 8 10W
OCNet MobileNetV2 - - - - -
OCNet 3x3-ResNet101 - - - 8 4W
OCNet Wide-ResNet38 - - - 16 2W
OCNet HRNetV2-48 - - - 8 10W
ISA MobileNetV2 - - - - -
ISA 3x3-ResNet101 - - - 8 4W
ISA Wide-ResNet38 - - - 16 2W
ISA HRNetV2-48 - - - 8 10W
OCR MobileNetV2 - - - - -
OCR 3x3-ResNet101 - - - 8 4W
OCR Wide-ResNet38 - - - 16 2W
OCR HRNetV2-48 - - - 8 10W
  • ADE20K (testing with single scale whole image)
Methods Backbone Val. mIOU PixelACC BS Iters
FCN 3x3-ResNet101 - - 16 15W
FCN Wide-ResNet38 - - 16 15W
FCN HRNetV2-48 - - 16 15W
OCNet 3x3-ResNet101 - - 16 15W
OCNet Wide-ResNet38 - - 16 15W
OCNet HRNetV2-48 - - 16 15W
ISA 3x3-ResNet101 - - 16 15W
ISA Wide-ResNet38 - - 16 15W
ISA HRNetV2-48 - - 16 15W
OCR 3x3-ResNet101 - - 16 15W
OCR Wide-ResNet38 - - 16 15W
OCR HRNetV2-48 - - 16 15W
  • LIP (testing with single scale whole image + left-right flip)
Methods Backbone Val. mIOU PixelACC BS Iters
FCN 3x3-ResNet101 - - 32 10W
FCN Wide-ResNet38 - - 32 10W
FCN HRNetV2-48 - - 32 10W
OCNet 3x3-ResNet101 - - 32 10W
OCNet Wide-ResNet38 - - 32 10W
OCNet HRNetV2-48 - - 32 10W
ISA 3x3-ResNet101 - - 32 10W
ISA Wide-ResNet38 - - 32 10W
ISA HRNetV2-48 - - 32 10W
OCR 3x3-ResNet101 - - 32 10W
OCR Wide-ResNet38 - - 32 10W
OCR HRNetV2-48 - - 32 10W
  • Pascal-VOC (testing with single scale whole image)
Methods Backbone Val. mIOU PixelACC BS Iters
FCN 3x3-ResNet101 - - 16 6W
FCN Wide-ResNet38 - - 16 6W
FCN HRNetV2-48 - - 16 6W
OCNet 3x3-ResNet101 - - 16 6W
OCNet Wide-ResNet38 - - 16 6W
OCNet HRNetV2-48 - - 16 6W
ISA 3x3-ResNet101 - - 16 6W
ISA Wide-ResNet38 - - 16 6W
ISA HRNetV2-48 - - 16 6W
OCR 3x3-ResNet101 - - 16 6W
OCR Wide-ResNet38 - - 16 6W
OCR HRNetV2-48 - - 16 6W
  • Pascal-Context (testing with single scale whole image)
Methods Backbone Val. mIOU PixelACC BS Iters
FCN 3x3-ResNet101 - - 16 3W
FCN Wide-ResNet38 - - 16 3W
FCN HRNetV2-48 - - 16 3W
OCNet 3x3-ResNet101 - - 16 3W
OCNet Wide-ResNet38 - - 16 3W
OCNet HRNetV2-48 - - 16 3W
ISA 3x3-ResNet101 - - 16 3W
ISA Wide-ResNet38 - - 16 3W
ISA HRNetV2-48 - - 16 3W
OCR 3x3-ResNet101 - - 16 3W
OCR Wide-ResNet38 - - 16 3W
OCR HRNetV2-48 - - 16 3W
  • COCO-Stuff-10K (testing with single scale whole image)
Methods Backbone Val. mIOU PixelACC BS Iters
FCN 3x3-ResNet101 - - 16 6W
FCN Wide-ResNet38 - - 16 6W
FCN HRNetV2-48 - - 16 6W
OCNet 3x3-ResNet101 - - 16 6W
OCNet Wide-ResNet38 - - 16 6W
OCNet HRNetV2-48 - - 16 6W
ISA 3x3-ResNet101 - - 16 6W
ISA Wide-ResNet38 - - 16 6W
ISA HRNetV2-48 - - 16 6W
OCR 3x3-ResNet101 - - 16 6W
OCR Wide-ResNet38 - - 16 6W
OCR HRNetV2-48 - - 16 6W

Acknowledgment

This project is developed based on the segbox.pytorch and the author of segbox.pytorch donnyyou retains all the copyright of the reproduced Deeplabv3, PSPNet related code.

openseg.pytorch's People

Contributors

pkurainbow avatar

Recommend Projects

  • React photo React

    A declarative, efficient, and flexible JavaScript library for building user interfaces.

  • Vue.js photo Vue.js

    ๐Ÿ–– Vue.js is a progressive, incrementally-adoptable JavaScript framework for building UI on the web.

  • Typescript photo Typescript

    TypeScript is a superset of JavaScript that compiles to clean JavaScript output.

  • TensorFlow photo TensorFlow

    An Open Source Machine Learning Framework for Everyone

  • Django photo Django

    The Web framework for perfectionists with deadlines.

  • D3 photo D3

    Bring data to life with SVG, Canvas and HTML. ๐Ÿ“Š๐Ÿ“ˆ๐ŸŽ‰

Recommend Topics

  • javascript

    JavaScript (JS) is a lightweight interpreted programming language with first-class functions.

  • web

    Some thing interesting about web. New door for the world.

  • server

    A server is a program made to process requests and deliver data to clients.

  • Machine learning

    Machine learning is a way of modeling and interpreting data that allows a piece of software to respond intelligently.

  • Game

    Some thing interesting about game, make everyone happy.

Recommend Org

  • Facebook photo Facebook

    We are working to build community through open source technology. NB: members must have two-factor auth.

  • Microsoft photo Microsoft

    Open source projects and samples from Microsoft.

  • Google photo Google

    Google โค๏ธ Open Source for everyone.

  • D3 photo D3

    Data-Driven Documents codes.