NOTE: 后续更新可见https://www.yuque.com/lart/htsg95,在语雀上写文档还是容易些。
由于几篇文章写的内容对于本人而言非常有用,但是为了防止原作“消失”的情况,这里将其进行了打印,保存成了pdf,存放在这里的GoodArticle
文件夹中,如果原作者觉得不妥,请联系我,我会删除。保护版权,人人有责!
- 迁移学习小册子-王晋东: https://zhuanlan.zhihu.com/p/35352154
- Bilateral Filtering: Theory and Applications: https://www.researchgate.net/publication/220427978_Bilateral_Filtering_Theory_and_Applications
- 半小时学会 PyTorch Hook - 尹相楠的文章 - 知乎:https://zhuanlan.zhihu.com/p/75054200
- 14 DESIGN PATTERNS TO IMPROVE YOUR CONVOLUTIONAL NEURAL NETWORKS:https://www.topbots.com/14-design-patterns-improve-convolutional-neural-network-cnn-architecture/
- 移动平均:你知道的与你不知道的 - 石川的文章 - 知乎:https://zhuanlan.zhihu.com/p/38276041
- A Comprehensive Introduction to Different Types of Convolutions in Deep Learning: https://towardsdatascience.com/a-comprehensive-introduction-to-different-types-of-convolutions-in-deep-learning-669281e58215
- einsum满足你一切需要:深度学习中的爱因斯坦求和约定 - 知乎: https://zhuanlan.zhihu.com/p/44954540
- 从SGD到NadaMax,十种优化算法原理及实现 - 知乎: https://zhuanlan.zhihu.com/p/81020717
- CS229:https://github.com/afshinea/stanford-cs-229-machine-learning
- transferlearning:https://github.com/jindongwang/transferlearning
- 数据挖掘:https://github.com/lyltj2010/DataMining
- 机器学习:https://github.com/allmachinelearning/MachineLearning
- 论文翻译: https://github.com/SnailTyan/deep-learning-papers-translation
- 综述论文: https://github.com/mlreview/machine-learning-surveys
- 深度学习500问:https://github.com/scutan90/DeepLearning-500-questions
- 深度学习面试宝典(含数学、机器学习、深度学习、计算机视觉、自然语言处理和SLAM等方向):https://github.com/amusi/Deep-Learning-Interview-Book
- zouxy09博客原创性博文导航:https://blog.csdn.net/zouxy09/article/details/14222605
- DeepLearning ON GPU:https://deeplearningongpu.readthedocs.io/en/latest/
- ML Records in 1110 Lab of BUPT:https://github.com/wmn7/ML_Practice
- 各种计算机领域资料的合集:https://github.com/ty4z2008/Qix
- 如何理解深度学习分布式训练中的large batch size与learning rate的关系? - 谭旭的回答 - 知乎:https://www.zhihu.com/question/64134994/answer/216895968
- 数据增强
- This is a list of awesome methods about data augmentation.: https://github.com/CrazyVertigo/awesome-data-augmentation
- A collection of awesome things about mixed sample data augmentation: https://github.com/JasonZhang156/awesome-mixed-sample-data-augmentation
- L1和L2范数:http://www.chioka.in/differences-between-l1-and-l2-as-loss-function-and-regularization/
- 想要算一算Wasserstein距离?这里有一份PyTorch实战:https://www.jiqizhixin.com/articles/19031102
- Gumbel-Softmax Trick(将不可数的argmax采样)和Gumbel分布:https://www.cnblogs.com/initial-h/p/9468974.html
Gumbel Trick 是一种从离散分布取样的方法,它的形式可以允许我们定义一种可微分的,离散分布的近似取样,这种取样方式不像「干脆以各类概率值的概率向量替代取样」这么粗糙,也不像直接取样一样不可导(因此没办法应对可能的 bp )。如何理解Gumbel-Max trick? - 曹恭泽的回答 - 知乎 https://www.zhihu.com/question/62631725/answer/201338234
- 让机器“一叶知秋”:弱监督视觉语义分割:https://zhuanlan.zhihu.com/p/37488849
- 介绍了CVPR19的一篇使用蒸馏处理语义分割的论文:https://segmentfault.com/a/1190000018493751
- 半监督深度学习小节:https://zhuanlan.zhihu.com/p/33196506
- 半监督深度学习又小结之Consistency Regularization:https://zhuanlan.zhihu.com/p/46893709
- Attention算法调研(弱监督语义分割):https://zhuanlan.zhihu.com/p/53218967
- 多任务学习Multi-task Learning(MTL)概述:https://blog.csdn.net/zaf0516/article/details/90380732
- 一片综述:https://arxiv.org/pdf/1706.05098.pdf
- 多任务学习(Multi-Task Learning, MTL):https://blog.csdn.net/laolu1573/article/details/78205180#
- 多任务深度学习的三个经验教训:https://www.leiphone.com/news/201902/SlFB1OlWd6U8ubJj.html
- Multi-Task Learning Using Uncertainty to Weigh Losses for Scene Geometry and Semantics:https://arxiv.org/pdf/1705.07115.pdf
- Awesome-Deblurring: https://github.com/subeeshvasu/Awesome-Deblurring
- awesome-gan-for-medical-imaging: https://github.com/xinario/awesome-gan-for-medical-imaging
- How to Train a GAN? Tips and tricks to make GANs work: https://github.com/soumith/ganhacks
- Awesome Database Learning: https://github.com/pingcap/awesome-database-learning
- ppwwyyxx/GroupNorm-reproduce: https://github.com/ppwwyyxx/GroupNorm-reproduce
- awesome-gan-for-medical-imaging: https://github.com/xinario/awesome-gan-for-medical-imaging
- 自己的学习历程,重点包括各种好玩的图像处理算法、运动捕捉、机器学习: https://github.com/1165048017/BlogLearning
- 较为简单直观的一份行人重识别的代码教程: https://github.com/layumi/Person_reID_baseline_pytorch
- 比较全面完整的一份代码:https://github.com/L1aoXingyu/reid_baseline
- 行人重识别(ReID)概述,将该任务的关键内容介绍了一下:https://blog.csdn.net/weixin_41427758/article/details/81188164
- 行人重识别重要数据集
DukeMTMC-reID
的一个简单介绍,了解一个任务的数据处理,看数据集的结构是比较直接的一个方法:https://blog.csdn.net/ctwy291314/article/details/83544142
- 综合
- 设计数据密集型应用 - 中文翻译: https://github.com/Vonng/ddia
- Computer Science from the Bottom Up: https://www.bottomupcs.com/index.xhtml
- Mining Social Media: http://socialdata.site/
- 重新审视《GOTO 语句被认为有害》: https://www.emon100.me/goto-translation/
- GitHub中文排行榜,帮助你发现高分优秀中文项目、更高效地吸收国人的优秀经验成果;榜单每周更新一次,敬请关注!Resources: https://github.com/kon9chunkit/GitHub-Chinese-Top-Charts
- Editor
- 笨方法学Vimscript: https://www.kancloud.cn/kancloud/learn-vimscript-the-hard-way
- VSCode: https://geek-docs.com/vscode
- Lisp
- build-your-own-lisp: https://ksco.gitbooks.io/build-your-own-lisp/content/
- Rust
- Rust 简明教程: https://geektutu.com/post/quick-rust.html
- Python
- Python 最佳实践指南 2018: https://learnku.com/docs/python-guide/2018
- Python最佳实践指南!: https://pythonguidecn.readthedocs.io/zh/latest/index.html
- What the f*ck Python! 🐍(中文翻译): https://github.com/leisurelicht/wtfpython-cn
- Python - 100天从新手到大师:https://github.com/jackfrued/Python-100-Days
- 🌠 来自一位 Pythonista 的编程经验分享,内容涵盖编码技巧、最佳实践与思维模式等方面:https://github.com/piglei/one-python-craftsman
- 🌠 Comprehensive Python Cheatsheet: https://github.com/gto76/python-cheatsheet
- Python深入06 Python的内存管理:https://www.cnblogs.com/vamei/p/3232088.html
- 10种检测Python程序运行时间、CPU和内存占用的方法:https://www.jb51.net/article/63244.htm
- Golang:
- Building a BitTorrent client from the ground up in Go: https://blog.jse.li/posts/torrent/
- Go 101: https://go101.org/article/101.html
- Go 高级教程: https://github.com/chai2010/advanced-go-programming-book
- 《Go语法树入门——开启自制编程语言和编译器之旅》(开源免费图书/Go语言进阶/掌握抽象语法树/Go语言AST/LLVM/LLIR/凹语言):https://github.com/chai2010/go-ast-book
- 7days-golang: https://github.com/geektutu/7days-golang
- Go 语言简明教程: https://geektutu.com/post/quick-golang.html
- Bash
- dylanaraps/pure-bash-bible:https://github.com/dylanaraps/pure-bash-bible
- Algorithm
- 手把手撕LeetCode题目,扒各种算法套路的裤子:https://github.com/labuladong/fucking-algorithm
- 小浩算法是我在疫情期间完成的一部图解算法题典! 目前共完成 140+ 道高频面试算法题目,总计 30w 字,全部采用漫画图解的方式,简单易懂,适合初中级读者:https://github.com/geekxh/hello-algorithm
- Matrix Analysis and Applied Linear Algebra:http://www.cse.zju.edu.cn/eclass/attachments/2015-10/01-1446085870-145420.pdf
- Introduction to Linear Algebra(4th):http://www.math.hcmus.edu.vn/~bxthang/Linear%20algebra%20and%20its%20applications.pdf
- https://github.com/golang101/golang101
- TOML中文文档:https://github.com/LongTengDao/TOML/
- 【FFmepg】学习音视频知识,整理资料,编写技术手册:https://github.com/feixiao/ffmpeg
- 将程序转化为命令行接口:https://github.com/google/python-fire
- 统计模型运算量和参数量的pytorch-OpCounter:https://github.com/Lyken17/pytorch-OpCounter
- 一些PyTorch的技巧,但是似乎有些旧了:https://github.com/kevinzakka/pytorch-goodies
- PyTorch Autograd:Understanding the heart of PyTorch’s magic:https://towardsdatascience.com/pytorch-autograd-understanding-the-heart-of-pytorchs-magic-2686cd94ec95
- CNN可视化:
- https://github.com/MisaOgura/flashtorch
- https://github.com/utkuozbulak/pytorch-cnn-visualizations
- Ways for Visualizing Convolutional Networks:https://buptldy.github.io/2016/09/25/2016-09-25-cnn_vis/
- CNN-Visualization:https://github.com/scutan90/CNN-Visualization
- 谷歌的新CNN特征可视化方法,构造出一个华丽繁复的新世界:https://www.leiphone.com/news/201711/aNw8ZjqMuqvygzlz.html
- How to visualize convolutional features in 40 lines of code:https://towardsdatascience.com/how-to-visualize-convolutional-features-in-40-lines-of-code-70b7d87b0030
- 【深度学习系列】CNN模型的可视化:https://www.cnblogs.com/charlotte77/p/8343700.html
- 高级封装:
- 训练提速:
- 简单两步加速PyTorch里的Dataloader - 巽二的文章 - 知乎:https://zhuanlan.zhihu.com/p/68191407
- 给训练踩踩油门——Pytorch加速数据读取 - MrTian的文章 - 知乎:https://zhuanlan.zhihu.com/p/80695364
- https://www.cnblogs.com/king-lps/p/10936374.html
- Image Test Time Augmentation with PyTorch!: https://github.com/qubvel/ttach
- SublimeText:本身没啥问题,而且随着版本的更新,原本在linux上的各种问题基本上也被解决,但是有个关键的问题,在使用GPU跑代码的时候,ST会特别卡顿,只好舍弃,转投VSCode了
- VSCode:微软的产品,还不错,性能问题逐渐得到优化,是肉眼可见的提升与改进,颜值可以很高
- Vim:终端修改文件必备,偶尔看看文件,或者远程通过ssh使用Vim写写代码,但是用的不是很痛快,终究还是被PyCharm这样的IDE惯坏了
- TeXmacs:一款对于数学编辑极其方便的工具,从FTP下载顺道可以下载对应的guile(a Scheme implementation):ftp://ftp.texmacs.org/TeXmacs/tmftp/Linux/
- PyCharm:写Python必备,前提是电脑配置不要太拖后腿,时常卡顿,但是功能强大又贴心,实在无法替代
- wingide:流畅是流畅,但是一来要激活,二来在我的Ubuntu上无法输入中文,而且创建项目的诡异流程让我怀疑这玩意是怎么想的:(,弃之
- ubuntu16.04安装NIVIDIA显卡驱动,cuda8.0,cuDNN6.0以及基于Anaconda安装Tensorflow-GPU:https://blog.csdn.net/pursuit_zhangyu/article/details/79362128
- 【解决】Ubuntu安装NVIDIA驱动(咨询NVIDIA工程师的解决方案):https://blog.csdn.net/u012759136/article/details/53355781#commentBox
- [专业亲测]Ubuntu16.04安装Nvidia显卡驱动(cuda)--解决你的所有困惑:https://blog.csdn.net/ghw15221836342/article/details/79571559
- 最全面解析 Ubuntu 16.04 安装nvidia驱动 以及各种错误:https://blog.csdn.net/u014561933/article/details/79958017
- ubuntu安装nvidia显卡驱动:https://blog.csdn.net/acelove40/article/details/69257574
- Ubuntu 15.10 使用 Xorg.conf 修改分辨率:https://segmentfault.com/a/1190000004510095
- windows的下载镜像链接藏得好深,速度可真慢,根本不给你下载的机会啊,隔壁
I Tell You
都是用ed2k
链接,好用的工具也只有迅雷,但迅雷(浏览器?!!)又不好好做下载,不让人喜欢,现在对于一般的下载感觉也限速了的样子,Windows重装的梦想看来难以实现了:https://www.microsoft.com/zh-cn/software-download/windows10ISO - 不要手动下载镜像,可以使用前面页面中提供的安装介质自动下载镜像,速度又快~:https://zhuanlan.zhihu.com/p/78326370
- 中英混排中的标点符号问题:https://thetype.com/2018/02/14211/
- 中英文兼顾的编程字体:https://github.com/be5invis/Sarasa-Gothic
- 非常棒的英文编程字体(要不是因为不支持中文,我就用这个了,这个字体偏宽一些):https://github.com/tonsky/FiraCode
- Placeholder.com – The Free Image Placeholder Service Favoured By Designers:https://placeholder.com/