Giter Club home page Giter Club logo

modules-on-cnn's Introduction

Skip to content Search or jump to… Pull requests Issues Codespaces Marketplace Explore

@hggzjx hggzjx / modules-on-cnn Public Cannot fork because you own this repository and are not a member of any organizations. Code Issues Pull requests Actions Projects Wiki Security Insights Settings modules-on-cnn/readme.txt @hggzjx hggzjx Add files via upload Latest commit 71a2598 now History 1 contributor 42 lines (23 sloc) 2.31 KB

这几天看了几篇CV语义分割的论文,写了一些围绕着CNN的提升模型性能的插件。在基于预训练微调的大环境下这些插件可以很容易地嵌入模型之中,提高网络提取特征的能力或者信息交互能力(感觉在孪生网络中有很大的发挥空间)。这里写的插件,在很多SOTA模型中会看到它们的影子,甚至就是这些插件的拼接组合。相较于很多基本都在为了play而去plug,结果加入后还不如baseline的Attention,这些插件的效果应该经得起推敲。不过归根结底这些都是trick,只能提升模型的上限,真正有意义的学术还是提升模型的下限。 我是调包侠,用的Keras接口写的,这些插件只要input的shape是(batch_size,height,width,channels)就行,还有一堆超参,调这些超参应该比网络的实现麻烦很多。

module:adaptively_spatial_feature_fusion(ASFF) paper:Adaptively Spatial Feature Fusion Learning Spatial Fusion for Single-Shot Object Detection https://link.zhihu.com/?target=https%3A//arxiv.org/abs/1911.09516v1

module:atrous_spatial_pyramid_pooling(ASPP) paper:DeepLab: Semantic Image Segmentation with Deep Convolutional Nets, Atrous Conv https://link.zhihu.com/?target=https%3A//arxiv.org/pdf/1606.00915.pdf

module:non_local_block paper:Non-local Neural Networks https://link.zhihu.com/?target=https%3A//arxiv.org/abs/1711.07971

module:squeeze_excite_block (SE) paper:Squeeze-and-Excitation Networks https://link.zhihu.com/?target=https%3A//arxiv.org/pdf/1709.01507.pdf

module:cbam_block paper:CBAM: Convolutional Block Attention Module https://link.zhihu.com/?target=https%3A//openaccess.thecvf.com/content_ECCV_2018/papers/Sanghyun_Woo_Convolutional_Block_Attention_ECCV_2018_paper.pdf

module:ghost_block paper:GhostNet: More Features from Cheap Operations https://link.zhihu.com/?target=https%3A//arxiv.org/pdf/1911.11907.pdf

module:receptive_field_block paper:Receptive Field Block Net for Accurate and Fast Object Detection https://link.zhihu.com/?target=https%3A//arxiv.org/abs/1711.07767

Footer © 2023 GitHub, Inc. Footer navigation Terms Privacy Security Status Docs Contact GitHub Pricing API Training Blog About modules-on-cnn/readme.txt at main · hggzjx/modules-on-cnn

modules-on-cnn's People

Contributors

hggzjx avatar

Watchers

 avatar

Recommend Projects

  • React photo React

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

  • Vue.js photo Vue.js

    🖖 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. 📊📈🎉

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 ❤️ Open Source for everyone.

  • D3 photo D3

    Data-Driven Documents codes.