BIT-DA / Transfer-Learning-Study

learning materials

迁移学习paper list与学习上手指南

推荐大家经常使用知乎和google搜索你看不懂的问题~自己主动学习十分重要！
Best代表马上阅读
Oral代表以后必须读

---

目录

• 迁移学习paper list与学习上手指南
  • 目录
  • 0. 热点方向
    
    • 3D point cloud
      
    • Continuous Learning
      
    • Few-shot/ semi-supervised/ long-tailed classification/segmentaiton/detection
      
    • Vision-Language/ prompt learning/ Multimodal
      
    • Diffusion model
      
    • OOD detection/generalization
    • 大模型/ pre-training/ self-supervised/ representation learning
    • BEV
    • edge computing transfer/ federate learning
    • transfer reinforcement learning
  • 1.迁移学习的背景介绍
    
  • 2.样本加权方法
  • 3.特征学习方法及其扩展
  • 4.纯深度学习网络结构研究
  • 5.生成对抗网络（GAN）
  • 6.深度迁移学习
  • 7.Partial Domain Adaptation （PDA）
  • 8.迁移学习在Semantic Segmentation中的应用
  • 9.迁移学习在Object Detection中的应用
  • 10. 迁移学习在Video Classification中的应用
  • 11.迁移学习在Person Re-identification中的应用
  • 12.迁移学习理论文章
  • 其他应该看论文
  • 顶会接收论文网站
  • 迁移学习资源汇总网站
  • 大牛主页

---

0. 热点方向

3D point cloud

paper	来源	Novelty	代码复现	简称

---

Continuous Learning

paper	来源	Novelty	代码复现	简称

---

Few-shot/ semi-supervised/ long-tailed classification/segmentaiton/detection

paper	来源	Novelty	代码复现	简称

---

Vision-Language/ prompt learning/ Multimodal

paper	来源	Novelty	代码复现	简称
Learning transferable visual models from natural language supervision	ICML 2021	Best	需要	CLIP
StyleCLIP: Text-Driven Manipulation of StyleGAN Imagery	ICCV 2021	Best		StyleCLIP
Learning to Prompt for Vision-Language Models	IJCV 2022	Oral		CoOp
Open-Vocabulary Object Detection via Vision and Language Knowledge Distillation	ICLR 2022	Oral		VILD
OPEN-VOCABULARY OBJECT DETECTION VIA VISION AND LANGUAGE KNOWLEDGE DISTILLATION	ICLR 2022	Oral		LSeg
PIX2SEQ: A LANGUAGE MODELING FRAMEWORK FOR OBJECT DETECTION	ICLR 2022	Oral		Pix2Seq
DenseCLIP: Language-Guided Dense Prediction with Context-Aware Prompting	CVPR 2022	Oral		DenseCLIP

---

Diffusion model

paper	来源	Novelty	代码复现	简称

---

OOD detection/generalization

paper	来源	Novelty	代码复现	简称

---

大模型/ pre-training/ self-supervised/ representation learning

paper	来源	Novelty	代码复现	简称

---

BEV

paper	来源	Novelty	代码复现	简称

---

edge computing transfer/ federate learning

paper	来源	Novelty	代码复现	简称

---

transfer reinforcement learning

paper	来源	Novelty	代码复现	简称

---

1.迁移学习的背景介绍

这些文章无需讲解，最先阅读，对迁移学习有个浅显了解

paper	来源	Novelty	代码复现	简称
迁移学习简明手册	这是中文版pdf	Best
A Survey on Transfer Learning	IEEE TKDE	Best
Deep Visual Domain Adaptation A Survey		Oral
Transfer Adaptation Learning：A Decade Survey		Oral

---

2.样本加权方法

这类方法已是很早期的方法了，但是需要了解

paper	来源	Novelty	代码复现	简称
Prediction Reweighting for Domain Adaptation	IEEE TNNLS	Best	PRDA
Correcting Sample Selection Bias by Unlabeled Data	NIPS	Oral	KMM
Unsupervised Domain Adaptation with Distribution Matching Machines	AAAI 2018	Oral

---

3.特征学习方法及其扩展

主要是浅层优化算法

paper	来源	Novelty	代码复现	简称
Domain Adaptation via Transfer Component Analysis	IEEE TNNLS	Best		TCA
Transfer Feature Learning with Joint Distribution Adaptation	CVPR	Best	需要	JDA
Domain Invariant and Class Discriminative Feature Learning for Visual Domain Adapation	IEEE TIP	Best	需要	DICD
Transfer Joint Matching for Unsupervised Domain Adaptation	CVPR	Oral		TJM
Unsupervised Domain Adaptation With Label and Structural Consistency	IEEE TIP	Oral
Joint Geometrical and Statistical Alignment for Visual Domain Adaptation	CVPR 2018	Oral		JGSA
Adaptation Regularization: A General Framework for Transfer Learning	IEEE TKDE	Oral

---

4.纯深度学习网络结构研究

基础中的基础！推荐先看Stanford CS231n课程，百度搜索即可，可以快速了解深度学习

paper	来源	Novelty	代码复现	简称

| |ImageNet Classification with Deep Convolutional Neural Networks |NIPS|Best|需要|AlexNet| |Very Deep Convolutional Networks for Large-Scale Image Recognition ||Best|需要|VGG| |Deep Residual Learning for Image Recognition |CVPR 2016 Best paper|Best|需要|ResNet| |Densely Connected Convolutional Networks |CVPR 2017 Best paper|Best|需要|DenseNet| |Squeeze-and-Excitation Networks |CVPR|Best||SENet| |Batch Normalization Accelerating Deep Network Training by Reducing Internal Covariate Shift |ICML|Best ||BN| |Deep Networks with Stochastic Depth |ECCV 2016 spotlight|Oral||| |Going deeper with convolutions |NIPS|Oral||InceptionV1/GoogLeNet| |Rethinking the Inception Architecture for Computer Vision |CVPR|Oral||InceptionV2/V3| |Inception-v4, Inception-ResNet and the Impact of Residual Connections on Learning |CVPR|Oral||InceptionV4/Inception-ResNet| |Aggregated Residual Transformations for Deep Neural Networks |CVPR|Oral||ResNext|

---

5.生成对抗网络（GAN）

paper	来源	Novelty	代码复现	简称
Generative Adversarial Nets	NIPS	Best	需要	GAN
Wasserstein GAN		Best		WGAN
Conditional Generative Adversarial Nets		Oral		CGAN
Least Squares Generative Adversarial Networks	ICCV	Oral		LSGAN

---

6.深度迁移学习

Deep Domain Adaptation，针对分类问题，研究重点！！

paper	来源	Novelty	代码复现	简称
How transferable are features in deep neural networks	NIPS 2014	Best
Learning Transferable Features with Deep Adaptation Networks	ICML 2015	Best	需要	DAN
Unsupervised Domain Adaptation by Backpropagation	ICML 2015	Best	需要	DANN/RevGrad
Maximum Classifier Discrepancy for Unsupervised Domain Adaptation	CVPR 2018	Best	需要	MCD
Unsupervised Domain Adaptation with Residual Transfer Networks	NIPS 2015	Oral		RTN
CyCADA: Cycle-Consistent Adversarial Domain Adaptation	ICML	Oral		CyCADA
Multi-Adversarial Domain Adaptation	AAAI 2018	Oral		MADA
Sliced Wasserstein Discrepancy for Unsupervised Domain Adaptation	CVPR 2019	Oral		SWD
Moment Matching for Multi-Source Domain Adaptation	ICML 2019	Oral
Bridging Theory and Algorithm for Domain Adaptation	ICML 2019	Oral		MDD
Conditional Adversarial Domain Adaptation	NIPS 2019	Best		CDAN
Contrastive Adaptation Network for Unsupervised Domain Adaptation	CVPR 2018	Oral		CAN

---

7.Partial Domain Adaptation （PDA）

paper	来源	Novelty	代码复现	简称
Partial Transfer Learning with Selective Adversarial Networks	CVPR 2018	Best	需要	SAN
Importance Weighted Adversarial Nets for Partial Domain Adaptation	CVPR 2018	Best	需要	IWAN
Partial Adversarial Domain Adaptation	ECCV 2018	Oral		PADA
Learning to Transfer Examples for Partial Domain Adaptation	CVPR 2019	Oral		ETN
Universal Domain Adaptation	CVPR 2019	Oral

---

8.迁移学习在Semantic Segmentation中的应用

paper	来源	Novelty	代码复现	简称
Fully Convolutional Networks for Semantic Segmentation	CVPR 2015	Best		FCN
SegFormer: Simple and Efficient Design for Semantic Segmentation with Transformers	NeurIPS 2021	必读 1	需要	SegFormer
Learning to Adapt Structured Output Space for Semantic Segmentation	CVPR 2018	必读 1	需要	AdaptSegNet
DAFormer: Improving Network Architectures and Training Strategies for Domain-Adaptive Semantic Segmentation	CVPR 2022	必读 1	需要	DAFormer
CyCADA - Cycle-Consistent Adversarial Domain Adaptation	ICML	Best		CyCADA
Taking A Closer Look at Domain Shift: Category-level Adversaries for Semantics Consistent Domain Adaptation	CVPR 2019	Oral		CLAN
ADVENT: Adversarial Entropy Minimization for Domain Adaptation in Semantic Segmentation	CVPR 2019	Oral		ADVENT
FDA: Fourier Domain Adaptation for Semantic Segmentation	CVPR 2020	Oral		FDA
Confidence Regularized Self-Training	ICCV 2019	Oral		CRST
Prototypical Pseudo Label Denoising and Target Structure Learning for Domain Adaptive Semantic Segmentation	CVPR 2021	Oral		ProDA

---

9.迁移学习在Object Detection中的应用

paper	来源	Novelty	代码复现	简称
Domain Adaptive Faster R-CNN for Object Detection in the Wild	CVPR 2018	Oral
Multi-adversarial Faster-RCNN for Unrestricted Object Detection	ICCV 2019	Oral

---

10. 迁移学习在Video Classification中的应用

paper	来源	Novelty	代码复现	简称
Temporal Attentive Alignment for Large-Scale Video Domain Adaptation	ICCV 2019	Best

---

11.迁移学习在Person Re-identification中的应用

paper	来源	Novelty	代码复现	简称
Image-Image Domain Adaptation with Preserved Self-Similarity and Domain-Dissimilarity for Person Re-identification	CVPR 2018	Best
Self-similarity Grouping: A Simple Unsupervised Cross Domain Adaptation Approach for Person Re-identification	ICCV 2019	Best

---

12.迁移学习理论文章

paper	来源	Novelty	代码复现	简称
Analysis of Representations for Domain Adaptation	NIPS	Best
A Theory of Learning from Different Domains		Best
A Kernel Method for the Two Sample Problem	NIPS	Best		MMD的全面分析

---

其他应该看论文

paper	来源	Novelty	代码复现	简称
Distance Metric Learning for Large Margin Nearest Neighbor Classification	NIPS	Best		LMNN
FaceNet - A Unified Embedding for Face Recognition and Clustering	Best		Triplet Loss

---

顶会接收论文网站

• CVPR/ICCV/ECCV
• ICML 2019
• AAAI
• ICLR 2019
• NIPS 2018

迁移学习资源汇总网站

• awsome DA 貌似只有论文
• 迁移学习

大牛主页

• 清华 黄高老师
• 清华 龙明盛老师
• 日本 原田研究所（MCD作者等）