DeepLearning Tutorial
DeepLearning Tutorial
一**.** 入门资料
完备的 AI 学习路线,最详细的中英文资源整理 (opens new window) ⭐
AiLearning: 机器学习 - MachineLearning - ML、深度学习 - DeepLearning - DL、自然语言处理 NL (opens new window)
Machine-Learning (opens new window)
数学基础
机器学习基础
快速入门
- 机器学习算法地图 (opens new window)
- [机器学习 吴恩达 Coursera个人笔记](https://github.com/Mikoto10032/DeepLearning/blob/master/books/[ML-Coursera][2014][Andrew Ng]/[2014]机器学习个人笔记完整版v5.1.pdf) && 视频(含官方笔记) (opens new window)
- CS229 课程讲义中文翻译 (opens new window) && [机器学习 吴恩达 cs229个人笔记](https://github.com/Mikoto10032/DeepLearning/blob/master/books/[ML-CS229][2011][Andrew NG]/[2011]斯坦福大学机器学习课程个人笔.pdf) && 官网(笔记) (opens new window) && 视频(中文字幕) (opens new window)
- 百页机器学习 (opens new window)
深入理解
- 《统计学习方法》李航 (opens new window) && 《统计学习方法》各章节笔记 (opens new window) && 《统计学习方法》各章节笔记 (opens new window) && 推荐答案:statistical-learning-method-solutions-manual (opens new window) 《统计学习方法》各章节笔记 (opens new window) && 《统计学习方法》各章节代码实现与课后习题参考解答 (opens new window)
- 《模式识别与机器学习》 Christopher Bishop (opens new window)
- 《机器学习》 周志华 (opens new window) && 南瓜书:pumpkin-book (opens new window)
- [《机器学习实战》 PelerHarrington](https://github.com/Mikoto10032/DeepLearning/blob/master/books/机器学习实战 中文双页版.pdf)
- 机器学习与深度学习书单 (opens new window)
深度学习基础
快速入门
- 深度学习思维导图 (opens new window) && 深度学习算法地图 (opens new window)
- 《斯坦福大学深度学习基础教程》 Andrew Ng(吴恩达) (opens new window)
- 深度学习 吴恩达 个人笔记 (opens new window) && 视频 (opens new window)
- MIT深度学习基础-2019视频课程 (opens new window)
- 台湾大学(NTU)李宏毅教授课程 (opens new window) && [leeml-notes (opens new window)
- 图解深度学习_Grokking-Deep-Learning (opens new window)
- [《神经网络与深度学习》 Michael Nielsen](https://github.com/Mikoto10032/DeepLearning/blob/master/books/神经网络和深度学习neural networks and deep-learning-中文_ALL.pdf)
- CS321-Hinton (opens new window)
- CS230: Deep Learning (opens new window)
- CS294-112 (opens new window)
计算机视觉
自然语言处理
- CS224n: Natural Language Processing with Deep Learning (opens new window)
- NLP上手教程 (opens new window)
- NLP入门推荐书目(2019版) (opens new window)
深度强化学习
深入理解
- [《深度学习》 Yoshua Bengio.Ian GoodFellow](https://github.com/Mikoto10032/DeepLearning/blob/master/books/深度学习.Yoshua Bengio%2BIan GoodFellow.pdf)⭐
- [《自然语言处理》Jacob Eisenstein](https://github.com/Mikoto10032/DeepLearning/blob/master/books/自然语言处理.Jacob Eisenstein.pdf)
- [《强化学习》](https://github.com/Mikoto10032/DeepLearning/blob/master/books/Reinforcement Learning.Sutton.pdf) && 第二版 (opens new window)
- hangdong的深度学习博客,论文推荐 (opens new window)
- Practical Deep Learning for Coders, v3 (opens new window)
- [《Tensorflow实战Google深度学习框架》 郑泽宇 顾思宇](https://github.com/Mikoto10032/DeepLearning/blob/master/books/Tensorflow 实战Google深度学习框架.pdf)
一些书单
工程能力
如何系统地学习算法? (opens new window) && LeetCode (opens new window) && leetcode题解 (opens new window) && 《算法导论》中算法的C++实现 (opens new window)
如何成为一名算法工程师 (opens new window) && 从小白到入门算法,我的经验分享给你~ (opens new window) && 我的研究生这三年 (opens new window) ⭐
计算机相关技术面试必备 (opens new window) && CS-WiKi (opens new window) && 计算机基础面试问题全面总结 (opens new window) && TeachYourselfCS-CN (opens new window) && 面试算法笔记-中文 (opens new window)
常用算法:
- Feature Engineering:continue variable && categorical variable
- Classic machine learning algorithm:LR, KNN, SVM, Random Forest, GBDT(XGBoost&&LightGBM), Factorization Machine, Field-aware Factorization Machine, Neural Network
- Cross validation, model selection:grid search, random search, hyper-opt
- Ensemble learning
Kaggle入门系列:(一)机器学习环境搭建 (opens new window) && Kaggle入门系列:(二)Kaggle简介 (opens new window) && Kaggle入门系列(三)Titanic初试身手 (opens new window)
一个框架解决几乎所有机器学习问题 (opens new window) && Approaching (Almost) Any Machine Learning Problem | Abhishek Thakur (opens new window)
Kaggle 首战 Top 2%, APTOS 2019 复盘总结 + 机器学习竞赛通用流程归纳 (opens new window)
**二、 **神经网络模型概览
- 1. 一文看懂25个神经网络模型 (opens new window)
- 2. DNN概述论文:详解前馈、卷积和循环神经网络技术 (opens new window)
- 3. colah's blog (opens new window)
- 4. Model Zoom (opens new window)
- 5. DNN概述 (opens new window)
- GitHub上的机器学习/深度学习综述项目合集 (opens new window)
- AlphaTree-graphic-deep-neural-network (opens new window)
CNN
发展史
图像分类
- 从LeNet-5到DenseNet (opens new window)
- 深度学习笔记(十一)网络 Inception, Xception, MobileNet, ShuffeNet, ResNeXt, SqueezeNet, EfficientNet, MixConv (opens new window)
- CNN网络结构的发展 (opens new window)
- Awesome - Image Classification:论文&&代码大全 (opens new window)
- pytorch-image-models (opens new window)
目标检测
From RCNN to YOLOv3 (opens new window):上 (opens new window),下 (opens new window)
后 R-CNN时代, Faster R-CNN、SSD、YOLO 各类变体统治下的目标检测综述:Faster R-CNN系列胜了吗? (opens new window)
一文看尽21篇目标检测最新论文(腾讯/Google/商汤/旷视/清华/浙大/CMU/华科/中科院等 (opens new window)
Anchor-Free目标检测算法 (opens new window): 第一篇:arxiv2015_baidu_DenseBox (opens new window), 如何评价最新的anchor-free目标检测模型FoveaBox? (opens new window), FCOS: 最新的one-stage逐像素目标检测算法 (opens new window) && 最新的Anchor-Free目标检测模型FCOS,现已开源! (opens new window) && 中科院牛津华为诺亚提出CenterNet,one-stage detector可达47AP,已开源! (opens new window) && AnchorFreeDetection (opens new window)
聊聊Anchor的"前世今生"(上) (opens new window)&&聊聊Anchor的"前世今生"(下) (opens new window)
目标检测算法综述之FPN优化篇 (opens new window) && 一文看尽物体检测中的各种FPN (opens new window)
图像分割(语义分割、实例分割、全景分割)
- 图像语义分割(Semantic segmentation) Survey (opens new window)
- 干货 | 一文概览主要语义分割网络 (opens new window)
- 语义分割 发展综述 (opens new window)
- 9102年了,语义分割的入坑指南和最新进展都是什么样的 (opens new window)
- 实例分割最新最全面综述:从Mask R-CNN到BlendMask (opens new window)
- 语义分割综述:深度学习背景下的语义分割的发展状况【推荐】 (opens new window)
- Awesome Semantic Segmentation:论文&&代码 (opens new window)
- 一篇看完就懂的最新语义分割综述 (opens new window)
- 基于深度学习的语义分割综述 (opens new window)
轻量化卷积神经网络
人脸相关
- 人脸检测算法综述 (opens new window)
- 人脸检测背景介绍和发展现状 (opens new window)
- 人脸识别算法演化史 (opens new window)
- CascadeCNN (opens new window)
- MTCNN (opens new window)
- awesome-Face_Recognition (opens new window)
- 异质人脸识别研究综述 (opens new window)
- 老板来了:人脸识别+手机推送,老板来了你立刻知道。 (opens new window)&& 手把手教你用Python实现人脸识别 (opens new window) && 人脸识别项目,网络模型,损失函数,数据集相关总结 (opens new window)
- 基于深度学习的人脸识别技术综述 (opens new window) && 如何走近深度学习人脸识别?你需要这篇超长综述 (opens new window) && 人脸识别损失函数综述(附开源实现) (opens new window) && Face Recognition Loss on Mnist with Pytorch (opens new window) && 人脸识别的LOSS(上) (opens new window) && 人脸识别的LOSS(下) (opens new window)
- 【每周CV论文推荐】 初学深度学习人脸关键点检测必读文章 (opens new window)
- 从传统方法到深度学习,人脸关键点检测方法综述 (opens new window)
- 人脸关键点检测综述 (opens new window)
- 人脸专集4 | 遮挡、光照等因素的人脸关键点检测 (opens new window)
- 【Face key point detection】人脸关键点检测实现 (opens new window)
- OpenCV实战:人脸关键点检测(FaceMark) (opens new window)
- CenterFace+TensorRT部署人脸和关键点检测400fps (opens new window)
图像超分辨率
行人重识别
- 【CVPR2019正式公布】行人重识别论文 (opens new window)
- 【CVPR2019正式公布】行人重识别论文 (opens new window),2019 行人再识别年度进展回顾 (opens new window)
图像着色
边检测
**OCR&&**文本检测
点云
细粒度图像分类
图像检索
人群计数
教程
前馈神经网络
激活函数
反向传播算法
优化问题
- 神经网络训练中的梯度消失与梯度爆炸 (opens new window)
- 梯度消失和梯度爆炸问题详解 (opens new window)
- 详解深度学习中的梯度消失、爆炸原因及其解决方法 (opens new window) && 神经网络梯度消失和梯度爆炸及解决办法 (opens new window)
卷积层
- A Comprehensive Introduction to Different Types of Convolutions in Deep Learning (opens new window) && 翻译:上 (opens new window)、下 (opens new window)
- 卷积有多少种?一文读懂深度学习中的各种卷积 (opens new window)
- 各种卷积 (opens new window)
- Convolution Network及其变种(反卷积、扩展卷积、因果卷积、图卷积) (opens new window)
- 深度学习基础--卷积类型 (opens new window)
- 变形卷积核、可分离卷积 (opens new window)
- 对深度可分离卷积、分组卷积、扩张卷积、转置卷积(反卷积)的理解 (opens new window)
- 反卷积 (opens new window)
- Dilated/Atrous conv 空洞卷积/多孔卷积 (opens new window)
- 卷积层输出大小尺寸计算及 “SAME” 和 “VALID” (opens new window) && 卷积的三种模式full, same, valid以及padding的same, valid (opens new window)
- 正常卷积与空洞卷积输出特征图与感受野大小的计算 (opens new window)
- 【Tensorflow】tf.nn.depthwise_conv2d如何实现深度卷积? (opens new window)
- 【Tensorflow】tf.nn.atrous_conv2d如何实现空洞卷积? (opens new window)
- 【Tensorflow】tf.nn.separable_conv2d如何实现深度可分卷积? (opens new window)
- 【TensorFlow】tf.nn.conv2d_transpose是怎样实现反卷积的? (opens new window)
池化层
卷积神经网络
- 卷积神经网络工作原理 (opens new window)
- 「七夕的礼物」: 一日搞懂卷积神经网络 (opens new window)
- 一文读懂卷积神经网络中的1x1卷积核 (opens new window)
- 如何理解神经网络中通过add和concate的方式融合特征? (opens new window) && 神经网络中对需要concat的特征进行线性变换然后相加是否好于直接concat? (opens new window)
- CNN 模型所需的计算力(flops)和参数(parameters)数量是怎么计算的? (opens new window) && 深度学习中卷积的参数量和计算量 (opens new window)
图像分类网络详解
CBAM:卷积块注意力模块 (opens new window) && CBAM: Convolutional Block Attention Module (opens new window)
ResNeSt 之语义分割 (opens new window) && 关于ResNeSt的点滴疑惑 (opens new window) && ResNeSt在刷榜之后被ECCV2020 strong reject (opens new window)
目标检测网络详解
- 目标检测的性能评价指标 (opens new window) && NMS和计算mAP时的置信度阈值和IoU阈值 (opens new window)&& 白话mAP (opens new window) && 目标检测模型的评估指标mAP详解(附代码) (opens new window)
- 深度学习中IU、IoU(Intersection over Union) (opens new window)
- Selective Search for Object Detection (opens new window)(译文) (opens new window)
- Region Proposal Network(RPN) (opens new window)
- 边框回归(Bounding Box Regression)详解 (opens new window)
- NMS——非极大值抑制 (opens new window) && 非极大值抑制NMS的python实现 (opens new window)
- 一文打尽目标检测NMS——精度提升篇 (opens new window) && 一文打尽目标检测NMS——效率提升篇 (opens new window)
- 目标检测回归损失函数简介:SmoothL1/IoU/GIoU/DIoU/CIoU Loss (opens new window)
- 将CNN引入目标检测的开山之作:R-CNN (opens new window)
- R-CNN论文详解 (opens new window)
- 深度学习(十八)基于R-CNN的物体检测 (opens new window)
- Fast R-CNN (opens new window)
- 深度学习(六十四)Faster R-CNN物体检测 (opens new window) && 你真的学会RoI Pooling了吗? (opens new window)
- 目标检测论文阅读:Feature Pyramid Networks for Object Detection (opens new window)
- SSD (opens new window)
- 实例分割--Mask RCNN详解(ROI Align / Loss Function) (opens new window) && 令人拍案称奇的Mask RCNN (opens new window)
- 何恺明大神的「Focal Loss」,如何更好地理解? (opens new window) && FocalLoss 对样本不平衡的权重调节和减低损失值 (opens new window) && focal_loss 多类别和二分类 Pytorch代码实现 (opens new window) && 多分类focal loss及其tensorflow实现 (opens new window)
- 堪比Focal Loss!解决目标检测中样本不平衡的无采样方法 (opens new window)
- 目标检测正负样本区分策略和平衡策略总结(一) (opens new window) && 目标检测正负样本区分策略和平衡策略总结(二) (opens new window) && 目标检测正负样本区分策略和平衡策略总结(三) (opens new window)
- YOLO (opens new window) && 目标检测|YOLO原理与实现 (opens new window) && 图解YOLO (opens new window) && 【论文解读】Yolo三部曲解读——Yolov1 (opens new window)
- 目标检测|YOLOv2原理与实现(附YOLOv3) (opens new window) && YOLO2 (opens new window) && 【论文解读】Yolo三部曲解读——Yolov2 (opens new window)
- <机器爱学习>YOLO v3深入理解 (opens new window) && 【论文解读】Yolo三部曲解读——Yolov3 (opens new window)
- YOLOv4 (opens new window)
- 目标检测之CornerNet (opens new window), 1 (opens new window), 2 (opens new window), 3 (opens new window)
- 目标检测小tricks--样本不均衡处理 (opens new window)
图像分割网络详解
- 超像素、语义分割、实例分割、全景分割 傻傻分不清 (opens new window)&& 语义分割、实例分割和全景分割的区别 (opens new window)
- 语义分割卷积神经网络快速入门 (opens new window)
- 图像语义分割入门+FCN/U-Net网络解析 (opens new window) && 深入理解深度学习分割网络Unet (opens new window)
- Unet神经网络为什么会在医学图像分割表现好? (opens new window)
- 图像语义分割的工作原理和CNN架构变迁 (opens new window)
- 语义分割中的Attention和低秩重建 (opens new window)
- 打通多个视觉任务的全能Backbone:HRNet (opens new window)
注意力机制
- 深度学习中的注意力模型(2017版) (opens new window)
- Attention Model(mechanism) 的 套路 (opens new window)
- 计算机视觉中的注意力机制(推荐) (opens new window)
- More About Attention(推荐) (opens new window)
- NLP中的Attention Mechanism (opens new window)
- Transformer中的Attention (opens new window)
- 综述:图像处理中的注意力机制 (opens new window)
Action
PyTorch官方实现ResNet (opens new window) && pytorch_resnet_cifar10 (opens new window)
mxnet如何查看参数数量 (opens new window) && mxnet查看FLOPS (opens new window)
GAN
发展史
- 千奇百怪的GAN变体 (opens new window)
- 苏剑林博客,讲解得淋漓尽致 (opens new window)
- The GAN Landscape:Losses, Architectures, Regularization, and Normalization (opens new window)
- 深度学习新星:GAN的基本原理、应用和走向 (opens new window)
- GAN生成图像综述 (opens new window)
- 2017年GAN 计算机视觉相关paper汇总 (opens new window)
- 必读的10篇关于GAN的论文 (opens new window)
教程
- [Basic](https://github.com/Mikoto10032/DeepLearning/blob/master/books/GAN-Basic Idea (2017.04.21).pdf)
- [Improving](https://github.com/Mikoto10032/DeepLearning/blob/master/books/GAN-Improving GAN (2017.05.05).pdf)
Action
- GAN学习指南:从原理入门到制作生成Demo (opens new window)
- 机器之心GitHub项目:GAN完整理论推导与实现 (opens new window)
- 在Keras上实现GAN:构建消除图片模糊的应用 (opens new window)
RNN
发展史
教程
word2vec
原理
训练词向量
Action
- 推荐:nlp-tutorial (opens new window)
- nlp-tutorial (opens new window)
- tensorflow中RNNcell源码分析以及自定义RNNCell的方法 (opens new window)
- TensorFlow中RNN实现的正确打开方式 (opens new window)
- TensorFlow RNN 代码 (opens new window)
- Tensorflow实现的深度NLP模型集锦 (opens new window)
- 用tensorflow LSTM如何预测股票价格 (opens new window)
- TensorFlow的多层LSTM实践 (opens new window)
- 《安娜卡列尼娜》文本生成——利用TensorFlow构建LSTM模型 (opens new window)
GNN
发展史
- Graph Neural Network(GNN)综述 (opens new window)
- 深度学习时代的图模型,清华发文综述图网络 (opens new window)
- 清华大学图神经网络综述:模型与应用 (opens new window)
- 图神经网络概述第三弹:来自IEEE Fellow的GNN综述 (opens new window)
- GNN最全文献资料整理 (opens new window) && Awesome-Graph-Neural-Networks (opens new window)
教程
- 如何理解 Graph Convolutional Network(GCN) (opens new window)
- 图卷积网络(GCN)新手村完全指南 (opens new window)
- 何时能懂你的心——图卷积神经网络(GCN) (opens new window)
- 图卷积网络GCN的理解与介绍 (opens new window)
- 一文读懂图卷积GCN (opens new window)
- 2020 年 GNN 开卷有益与再谈图卷积 (opens new window)
- 【GCN】万字长文带你入门 GCN (opens new window)
- 如何解决图神经网络(GNN)训练中过度平滑的问题? (opens new window)
- 全连接的图卷积网络(GCN)和self-attention这些机制有什么区别联系 (opens new window) && CNN与GCN的区别、联系及融合 (opens new window)
Action
三**.** 深度模型的优化与正则化
权重衰减(weight decay)与学习率衰减(learning rate decay) (opens new window) && pytorch必须掌握的的4种学习率衰减策略 (opens new window)
8. 通俗讲解查全率和查准率 (opens new window) && 全面梳理:准确率,精确率,召回率,查准率,查全率,假阳性,真阳性,PRC,ROC,AUC,F1 (opens new window) && 机器学习之类别不平衡问题 (1) —— 各种评估指标 (opens new window) && 机器学习之类别不平衡问题 (2) —— ROC和PR曲线 (opens new window) && AUC详解与python实现 (opens new window) && 微平均和宏平均 (opens new window) && 机器学习中的性能度量 (opens new window)
激活函数一览 (opens new window) && 深度学习中几种常见的激活函数理解与总结 (opens new window)
机器学习中常见的损失函数及其应用场景 (opens new window) && PyTorch的十八个损失函数 (opens new window)
10. Coursera吴恩达《优化深度神经网络》课程笔记(3)-- 超参数调试、Batch正则化和编程框架 (opens new window)
13. 机器学习里的黑色艺术:normalization, standardization, regularization (opens new window) && 数据标准化/归一化normalization (opens new window) && 特征工程中的「归一化」有什么作用? (opens new window)
18. Batch Normalization(BN) (opens new window):1 (opens new window),2 (opens new window),3 (opens new window),4 (opens new window), 5 (opens new window), 6 (opens new window), 7 (opens new window)
19. 详解深度学习中的Normalization,不只是BN (opens new window) && 如何区分并记住常见的几种 Normalization 算法 (opens new window)
21. 详解深度学习中的梯度消失、爆炸原因及其解决方法 (opens new window) && 神经网络梯度消失和梯度爆炸及解决办法 (opens new window)
22. Dropout (opens new window), 1 (opens new window), 2 (opens new window), 3 (opens new window),系列解读Dropout (opens new window)
23.谱归一化(Spectral Normalization)的理解 (opens new window),常见向量范数和矩阵范数 (opens new window),谱范数正则(Spectral Norm Regularization)的理解 (opens new window)
24.L1正则化与L2正则化 (opens new window) && 深入理解L1、L2正则化 (opens new window) && L2正则=Weight Decay?并不是这样 (opens new window) && 都9102年了,别再用Adam + L2 regularization (opens new window)
25.为什么选用交叉熵而不是MSE (opens new window) &&为什么使用交叉熵作为损失函数 (opens new window) &&二元分类为什么不能用MSE做为损失函数? (opens new window)
交叉熵代价函数(作用及公式推导) (opens new window) && 交叉熵损失的来源、说明、求导与pytorch实现 (opens new window) && Softmax函数与交叉熵 (opens new window) && 极大似然估计与最小化交叉熵损失或者KL散度为什么等价 (opens new window)
梯度下降优化算法纵览 (opens new window), 1 (opens new window), 2 (opens new window), 几种优化算法的比较(BGD、SGD、Adam、RMSPROP) (opens new window)
Softmax:详解softmax函数以及相关求导过程 (opens new window) && softmax的log似然代价函数(公式求导) (opens new window) && 从最优化的角度看待Softmax损失函数 (opens new window) && 【技术综述】一文道尽softmax loss及其变种 (opens new window)
权重初始化
四**.** 炼丹术士那些事
调参经验
神经网络训练loss不下降原因集合 (opens new window) && loss不下降的解决方法 (opens new window)
深度学习:欠拟合问题的几种解决方案 (opens new window) &&过拟合和欠拟合问题 (opens new window)
不平衡数据集处理方法 (opens new window): 其一 (opens new window), 其二 (opens new window), 其三 (opens new window) && Awesome Imbalanced Learning (opens new window) && Class-balanced-loss-pytorch (opens new window)
- 凭什么相信你,我的CNN模型?(篇一:CAM和Grad-CAM) (opens new window) && pytorch-grad-cam (opens new window) && Grad-CAM-tensorflow (opens new window) && grad-cam.tensorflow (opens new window) && cnn_visualization (opens new window)
- 凭什么相信你,我的CNN模型?(篇二:万金油LIME) (opens new window)
- 论文笔记:Grad-CAM: Visual Explanations from Deep Networks via Gradient-based Localization (opens new window)
- CV:基于Keras利用训练好的hdf5模型进行目标检测实现输出模型中的表情或性别的gradcam(可视化) (opens new window)
大卷积核还是小卷积核? (opens new window) 1 (opens new window), 2 (opens new window)
刷排行榜的小技巧
图像分类
炼丹笔记三:数据增强 (opens new window) && 数据增强(Data Augmentation) (opens new window)
【技术综述】 深度学习中的数据增强(上) (opens new window) && 【技术综述】深度学习中的数据增强(下) (opens new window)
《Bag of Tricks for Image Classification with CNN》 (opens new window)&& pdf (opens new window)
深度神经网络模型训练中的最新tricks总结【原理与代码汇总】 (opens new window) && 神经网络训练trick (opens new window)
- 从0上手Kaggle图像分类挑战:冠军解决方案详解 (opens new window)
- Kaggle 冰山图像分类大赛近日落幕,看冠军团队方案有何亮点 (opens new window)
- 【Kaggle冠军分享】图像识别和分类竞赛,数据增强及优化算法 (opens new window)
- 识别座头鲸,Kaggle竞赛第一名解决方案解读 (opens new window)
- kaggle 首战拿金牌总结 (opens new window)
- 16岁高中生夺冠Kaggle地标检索挑战赛!而且竟然是Kaggle老兵 (opens new window)
- 6次Kaggle计算机视觉类比赛赛后感 (opens new window)
- Kaggle首战斩获第三-卫星图像识别 (opens new window)
目标检测
- ensemble
- deformable
- sync bn
- ms train/test
- 目标检测任务的优化策略tricks (opens new window)
- 目标检测小tricks--样本不均衡处理 (opens new window)
- 汇总|目标检测中的数据增强、backbone、head、neck、损失函数 (opens new window)
- 目标检测算法中的常见trick (opens new window)
- Bag of Freebies —— 提升目标检测模型性能的免费tricks (opens new window)
- 目标检测比赛中的tricks(已更新更多代码解析) (opens new window)
- Kaggle:肺癌自动诊断系统3D Deep Leaky Noisy-or Network 论文阅读 (opens new window)
- 干货|大神教你如何参加kaggle比赛——根据CT扫描图预测肺癌 (opens new window)
五**.** 年度总结
六**.** 科研相关
深度学习框架
Python3.x(先修)
- The Python Tutorial (opens new window)
- 廖雪峰Python教程 (opens new window)
- 菜鸟教程 (opens new window)
- 给深度学习入门者的Python快速教程 - 基础篇 (opens new window)
- Python - 100天从新手到大师 (opens new window)
- Python中读取,显示,保存图片的方法 (opens new window) && Python的图像打开保存显示的几种方式 (opens new window)
Numpy(先修)
- Quickstart tutorial (opens new window)
- Numpy快速入门(Numpy 1.14 官方文档中文翻译) (opens new window)
- Numpy中文文档 (opens new window)
- 给深度学习入门者的Python快速教程 - numpy和Matplotlib篇 (opens new window)
Opencv-python
- OpenCV-Python Tutorials (opens new window)
- OpenCV官方教程中文版(For Python) (opens new window)
- 数字图像处理系列 (opens new window)
- python+OpenCV图像处理 (opens new window)
- 给深度学习入门者的Python快速教程 - 番外篇之Python-OpenCV (opens new window)
Pandas
Tensorflow
- 如何高效地学习 TensorFlow 代码 (opens new window)
- 中文教程 (opens new window)
- TensorFlow官方文档 (opens new window)
- CS20:Tensorflow for DeepLearning Research (opens new window)
- 吴恩达TensorFlow专项课程 (opens new window)
- 【干货】史上最全的Tensorflow学习资源汇总 (opens new window)
- 《21个项目玩转深度学习———基于TensorFlow的实践详解》 (opens new window)
- 最全Tensorflow2.0 入门教程持续更新 (opens new window)
- Github优秀开源教程 (opens new window)
MXNet
PyTorch
- Pytorch版动手学深度学习 (opens new window)
- PyTorch中文文档 (opens new window)
- WELCOME TO PYTORCH TUTORIALS (opens new window)
- 史上最全的PyTorch学习资源汇总 (opens new window)
- 【干货】史上最全的PyTorch学习资源汇总 (opens new window)
- Hands-on tour to deep learning with PyTorch (opens new window)
- pytorch学习(五)—图像的加载/读取方式 (opens new window) && PyTorch—ImageFolder/自定义类 读取图片数据 (opens new window)
深度学习常用命令
- [command_for_deeplearning](https://github.com/Stephenfang51/command_for_deeplearning/blob/master/command for deeplearning.md)
Python****可视化
- Top 50 matplotlib Visualizations – The Master Plots (with full python code) (opens new window)
- Python之MatPlotLib使用教程 (opens new window)
- 十分钟上手matplotlib,开启你的python可视化 (opens new window)
- 给深度学习入门者的Python快速教程 - numpy和Matplotlib篇 (opens new window)
标注工具
目标检测标注工具
语义分割标注工具
数据集
- 1. 25个深度学习相关公开数据集 (opens new window)
- 2. 自然语言处理(NLP)数据集 (opens new window)
- 3.全唐诗(43030首) (opens new window)
- 4. 伯克利大学公开数据集 (opens new window)
- 5. ACL 2018资源:100+ 预训练的中文词向量 (opens new window)
- 6. 预训练中文词向量 (opens new window)
- 7. 公开数据集种子库 (opens new window)
- 8. 计算机视觉,深度学习,数据挖掘数据集整理 (opens new window)
- 9. 计算机视觉著名数据集CV Datasets (opens new window)
- 10. 计算机视觉相关数据集和比赛 (opens new window)
- 11. 这是一份非常全面的开源数据集,你,真的不想要吗? (opens new window)
- 12. 人群密度估计现有主要数据集特点及其比较 (opens new window)
- 13. DANBOORU2017: A LARGE-SCALE CROWDSOURCED AND TAGGED ANIME ILLUSTRATION DATASET (opens new window)
- 14. 行人重识别数据集 (opens new window)
- 15. 自然语言处理常见数据集、论文最全整理分享 (opens new window)
- 16. paper, code, sota (opens new window)
- 17. 旷视RPC大型商品数据集发布! (opens new window)
- 18. CVPR 2019「准满分」论文:英伟达推出首个跨摄像头汽车跟踪数据集(汽车Re-ID) (opens new window)
- 19.【OCR技术】大批量生成文字训练集 (opens new window)
- 20. 语义分析数据集-MSRA (opens new window)
- IEEE DataPort (opens new window)
- 数据集市 (opens new window)
- 医疗/医学图像数据集 (opens new window):Medical Data for Machine Learning (opens new window) && 医疗领域图像挑战赛数据集 (opens new window) && 【医学影像系列:一】数据集合集 最新最全 (opens new window) && medical-imaging-datasets (opens new window) && 【数据集】一文道尽医学图像数据集与竞赛 (opens new window) && 医学图像数据集汇总 (opens new window)
记笔记工具
- Markdown编辑器:Typora介绍 (opens new window)
- Markdown语法介绍(常用) (opens new window)
- Markdown 语法手册 (完整整理版) (opens new window)
- Markdown中Latex 数学公式基本语法 (opens new window)
会议期刊列表
- 国际会议日期表 (opens new window)
- ai-deadlines (opens new window)
- Keep Up With New Trends (opens new window)
- 计算机会议排名等级 (opens new window)
- 中国计算机学会(CCF)推荐国际学术刊物和会议 (opens new window)
论文写作工具
- Windows: Texlive+Texstudio (opens new window)
- Ubuntu: Texlive+Texmaker (opens new window)
- Latex:基本用法、表格、公式、算法 (opens new window)
- LaTeX 各种命令,符号 (opens new window)
论文画图工具
论文写作教程
- 刘知远_如何写一篇合格的NLP论文 (opens new window)
- 刘洋_如何写论文_V7 (opens new window)
- 如何端到端地写科研论文-邱锡鹏 (opens new window)
- 论文Introduction写作其一 (opens new window), 论文Introduction写作其二 (opens new window), 论文Introduction写作其三 (opens new window)
- 毕业论文怎么写 (opens new window)
- 浅谈学术论文rebuttal (opens new window)
- 研之成理写作实验室 (opens new window)
- 智源论坛·论文写作专题报告会 (opens new window):《论文写作小白的成长之路》 (opens new window) && 《谈如何写一篇合格的国际学术论文》 (opens new window) && 《计算机视觉会议论文从投稿到接收》 (opens new window)
ResearchGos
- ResearchGo:研究生活第一帖——文献检索与管理 (opens new window)
- ResearchGo:研究生活第二贴——文献阅读 (opens new window)
- ResearchGo:研究生活第三帖——阅读辅助 (opens new window)
- ResearchGo:研究生活第四帖——文献调研 (opens new window)
- ResearchGo:研究生活第五帖——文献综述 (opens new window)
- ResearchGo:研究生活第六帖——如何讲论文 (opens new window)
- ResearchGo:研究生活第七帖——专利检索与申请 (opens new window)
- ResearchGo:研究生活第八帖——写论文、做PPT、写文档必备工具集锦 (opens new window)
毕业论文排版
信号处理
傅里叶变换
- 傅里叶分析之掐死教程(完整版)更新于2014.06.06 (opens new window)
- 如何简明的总结傅里叶变换? (opens new window)
- 从连续时间傅里叶级数到快速傅里叶变换 (opens new window)
- 十分简明易懂的FFT(快速傅里叶变换) (opens new window)
- 傅里叶级数推导过程 (opens new window)
小波变换
- 形象易懂讲解算法I——小波变换 (opens new window)
- 小波变换完美通俗讲解系列之 (一) (opens new window) && 小波变换完美通俗讲解系列之 (二) (opens new window)
实战
- MWCNN中使用的haar小波变换 pytorch (opens new window)
- 【小波变换】小波变换入门----haar小波 (opens new window)
- (3)小波变换原理及应用 (opens new window)
- 图像处理-小波变换 (opens new window)
机器学习理论与实战
ID3、C4.5、CART、随机森林、bagging、boosting、Adaboost、GBDT、xgboost算法总结 (opens new window)
数据挖掘十大算法简要说明 (opens new window),机器学习十大经典算法入门 (opens new window) && 【算法模型】轻松看懂机器学习十大常用算法 (opens new window)
机器学习理论篇之经典算法
信息论
多层感知机**(MLP)**
- 多层感知机(MLP)学习与总结博客 (opens new window)
- 多层感知机:Multi-Layer Perceptron (opens new window)
- 神经网络基础-多层感知器(MLP) (opens new window)
k近邻(KNN)
k均值(K-means)
朴素贝叶斯**(Naive Bayesian)**
- 一个例子搞清楚(先验分布/后验分布/似然估计) (opens new window)
- 朴素贝叶斯分类器(Naive Bayesian Classifier) (opens new window)
- 朴素贝叶斯分类器 详细解析 (opens new window)
决策树**(Decision Tree)**
- Python3《机器学习实战》学习笔记(二):决策树基础篇之让我们从相亲说起 (opens new window)
- Python3《机器学习实战》学习笔记(三):决策树实战篇之为自己配个隐形眼镜 (opens new window)
- 机器学习实战教程(十三):树回归基础篇之CART算法与树剪枝 (opens new window)
- 《机器学习实战》基于信息论的三种决策树算法(ID3,C4.5,CART) (opens new window)
- 说说决策树剪枝算法 (opens new window)
- 机器学习实战 第九章 树回归 (opens new window)
- 决策树值ID3、C4.5实现 (opens new window)
- 决策树之CART实现 (opens new window)
随机森林**(Random Forest)**
线性回归(Linear Regression)
- 线性回归最小二乘法和最大似然估计 (opens new window)
- 【从入门到放弃】线性回归 (opens new window)
- 线性回归(频率学派-最大似然估计)与岭回归(贝叶斯角度-最大后验估计)的概率解释 (opens new window)
- 机器学习笔记四:线性回归回顾与logistic回归 (opens new window)
逻辑回归**(Logistic Regression)**
- 【机器学习面试总结】—— LR(逻辑回归) (opens new window)
- 【机器学习面试题】逻辑回归篇 (opens new window)
- 极大似然概率和最小损失函数,以及正则化简介 (opens new window)
- GLM(广义线性模型) 与 LR(逻辑回归) 详解 (opens new window)
支持向量机**(SVM)**
- 【机器学习面试总结】—— SVM (opens new window)
- SVM系列-从基础到掌握 (opens new window)
- SVM通俗导论 July (opens new window)
- 核函数 (opens new window): 机器学习有很多关于核函数的说法,核函数的定义和作用是什么? (opens new window) && SVM中,高斯核为什么会把原始维度映射到无穷多维? (opens new window) && svm核函数的理解和选择 (opens new window) && 核函数和径向基核函数 (Radial Basis Function)--RBF (opens new window) && SVM核函数 (opens new window)
提升方法**(Adaboost)**
梯度提升决策树**(GBDT)**
- LightGBM大战XGBoost (opens new window)
- 概述XGBoost、Light GBM和CatBoost的同与不同 (opens new window) && XGBoost、LightGBM、Catboost总结 (opens new window) && XGBoost、Light GBM和CatBoost的参数及性能比较 (opens new window)
- 梯度提升决策树 (opens new window)
- GBDT原理及应用 (opens new window)
- XGBOOST原理篇 (opens new window)
- xgboost入门与实战(原理篇) (opens new window) && xgboost入门与实战(实战调参篇) (opens new window)
- 【干货合集】通俗理解kaggle比赛大杀器xgboost (opens new window)
- GBDT分类的原理及Python实现 (opens new window)
- GBDT原理及利用GBDT构造新的特征-Python实现 (opens new window)
- Python+GBDT算法实战——预测实现100%准确率 (opens new window)
EM(期望最大化)
高斯混合模型**(GMM)**
马尔科夫决策过程**(MDP)**
- 马尔科夫决策过程之Markov Processes(马尔科夫过程) (opens new window)
- 马尔科夫决策过程之Markov Reward Process(马尔科夫奖励过程) (opens new window)
- 马尔科夫决策过程之Bellman Equation(贝尔曼方程) (opens new window)
- 马尔科夫决策过程之Markov Decision Process(马尔科夫决策过程) (opens new window)
- 马尔科夫决策过程之最优价值函数与最优策略 (opens new window)
条件随机场**(CRF,** 判别式模型**)**
- 如何轻松愉快地理解条件随机场 (opens new window)
- 如何用简单易懂的例子解释条件随机场(CRF)模型?它和HMM有什么区别? (opens new window)
- HMM ,MHMM,CRF 优缺点与区别 (opens new window)
降维算法
主成分分析**(PCA)**
奇异值分解**(SVD)**
- 强大的矩阵奇异值分解(SVD)及其应用 (opens new window)
- 奇异值分解(SVD) (opens new window)
- 奇异值分解(SVD)原理详解及推导 (opens new window)
- SVD在推荐系统中的应用详解以及算法推导 (opens new window)
线性判别分析**(LDA)**
标签传播算法**(Label Propagation Algorithm)**
- [参考资料](https://github.com/Mikoto10032/DeepLearning/blob/master/books/Semi-Supervised Learning with Graphs.pdf)
蒙塔卡罗树搜索**(MCTS)**
集成**(Ensemble)**
- 集成学习之bagging,stacking,boosting概念理解 (opens new window)
- 集成学习法之bagging方法和boosting方法 (opens new window)
- Bagging,Boosting,Stacking (opens new window) && 常用的模型集成方法介绍:bagging、boosting 、stacking (opens new window)
t分布随机邻居嵌入(TSNE)
- 流形学习-高维数据的降维与可视化 (opens new window)
- tSNE (opens new window)
- 使用t-SNE可视化图像embedding (opens new window)
谱聚类**(Spectral Clustering)**
异常点检测
- 数据挖掘中常见的「异常检测」算法有哪些? (opens new window)
- 异常点检测算法综述 (opens new window)
- 异常检测的N种方法,其中有一个你一定想不到 (opens new window)
- 异常检测资源汇总:anomaly-detection-resources (opens new window)
机器学习实战篇
- 15分钟带你入门sklearn与机器学习——分类算法篇 (opens new window) && 如何为你的回归问题选择最合适的机器学习方法? (opens new window)
- 十分钟上手sklearn:安装,获取数据,数据预处理 (opens new window) && 十分钟上手sklearn:特征提取,常用模型,交叉验证 (opens new window)
- MachineLearning_Python (opens new window)
- Machine Learning Course with Python (opens new window)
- Statistical-Learning-Method_Code (opens new window)
- Python3机器学习 (opens new window)
- 含大牛总结的分类模型一般需要调节的参数 (opens new window)
机器学习、深度学习的一些研究方向
多任务学习**(Multi-Task Learning)**
- 模型汇总-14 多任务学习-Multitask Learning概述 (opens new window)
- (译)深度神经网络的多任务学习概览(An Overview of Multi-task Learning in Deep Neural Networks) (opens new window)
- Multi-task Learning and Beyond: 过去,现在与未来 (opens new window);
零次学习**(Zero Shot Learning)**
小样本学习**(Few-Shot Learning)**
- few-shot learning是什么 (opens new window)
- 零次学习(Zero-Shot Learning)入门 (opens new window)
- 小样本学习(Few-shot Learning)综述 (opens new window)
- Few-Shot Learning in CVPR 2019 (opens new window)
- 当小样本遇上机器学习 fewshot learning (opens new window)
多视觉学习**(Multi-View Learning)**
嵌入**(Embedding)**
- 万物皆Embedding,从经典的word2vec到深度学习基本操作item2vec (opens new window)
- YJango的Word Embedding--介绍 (opens new window)
迁移学习**(Transfer Learning)**
- 1. 迁移学习:经典算法解析 (opens new window)
- 2. 什么是迁移学习 (Transfer Learning)?这个领域历史发展前景如何? (opens new window)
- 3. 迁移学习个人笔记 (opens new window)
- 迁移学习总结(One Shot Learning, Zero Shot Learning) (opens new window)
域自适应**(Domain Adaptation)**
- Domain Adaptation视频教程(附PPT)及经典论文分享 (opens new window)
- 模型汇总15 领域适应性Domain Adaptation、One-shot/zero-shot Learning概述 (opens new window)
- 【深度学习】论文导读:无监督域适应(Deep Transfer Network: Unsupervised Domain Adaptation) (opens new window)
- 【论文阅读笔记】基于反向传播的无监督域自适应研究 (opens new window)
- 【Valse大会首发】领域自适应及其在人脸识别中的应用 (opens new window)
- CVPR 2018:基于域适应弱监督学习的目标检测 (opens new window)
元学习**(Meta Learning)**
强化学习**(Reinforcement Learning)**
推荐系统**(Recommendation System)**
论文列表
- Embedding从入门到专家必读的十篇论文 (opens new window)
- Reco-papers (opens new window)
- Ad-papers (opens new window)
- deep-recommender-system (opens new window)
- CTR预估系列入门手册 (opens new window)
教程
- 推荐系统从入门到接着入门 (opens new window)
- 深度学习推荐系统笔记 (opens new window)
- 推荐系统干货总结 (opens new window)
- 入门推荐系统,你不应该错过的知识清单 (opens new window)
- 推荐系统技术演进趋势:从召回到排序再到重排 (opens new window)
- 深入理解推荐系统:召回 (opens new window) && 深入理解推荐系统:排序 (opens new window)
- 《深度学习推荐系统》总结系列一 (opens new window) && 《深度学习推荐系统》总结系列二 (opens new window)
- 推荐系统--完整的架构设计和算法(协同过滤、隐语义) (opens new window) && 从0到1打造推荐系统-架构篇 (opens new window)
- 协同过滤和基于内容推荐有什么区别? (opens new window)
实战