知识体系结构
- 1. 深度学习框架
- 2. 机器学习
- 3. 深度学习
- 4. 个人常用网络
- 5. 推荐系统
- 6. 语音技术
- 7. 模型融合
- 8. 数据集不均衡问题
- 9. 图神经网络
- 10. 轻量化网络
- 11. 强化学习
- 12. AI-调参工具及策略模型
- 13. 回归问题
- 14.图像基础
- 15. 医学图像诊断模型
- 16. 深度学习CV基础
- 17. 深度学习NLP基础
- 18. python极简入门
- 19. 数学基础
- 20. 图像检索
- 21. 数据标注
- 22.图
- 23.人脸识别系列
- 24.神经网络的可解释性
- 25.深度学习的回归任务
- 26.深度学习之多模态
- 27.自监督、无监督学习
- 28.图像生成
- 29.视频生成
- 30.超分辨率
主要的学习方式:还是先用,后学,再思考
无论是学分类、检测、分割、检索、NLP还是推荐,搞学术,最好能构建一套适合自己的baseline,其自定义的性质最好要高,方便自己深入研究
目前深度学习流行的特征提取器:
有三大类:
CNN 、 LSTM 、 Transformer
用投票模型融合方法进行最后结果的融合,看融合效果好不好,可以一定程度上看出,两个模型的提取能力是否一致等等
Pytorch的基本使用
Tenorflow和Keras的基本使用
工程应用
一 深度学习框架
二 机器学习
三 深度学习
四 个人常用网络
五 推荐系统
六 语音技术
七 模型融合
八 数据集不均衡问题
九 图神经网络
十 轻量化网络
十一 强化学习
十二. AI-调参工具及策略模型
十三 回归问题
十四. 图像基础
十五. 医学图像诊断模型
十六 深度学习CV基础
十七 深度学习NLP基础
十八 python极简入门
十九 数学基础
二十 图像检索
二十一 数据标注
二十二 图的相关操作
二十三 人脸识别系列
二十四 深度学习的回归任务
二十六 深度学习之多模态
二十七 自监督、无监督学习
AI各方向
28.图像生成
Progressive Growing of GANs for Improved Quality, Stability, and Variation
A Style-Based Generator Architecture for Generative Adversarial Networks
Analyzing and Improving the Image Quality of StyleGAN
Alias-Free Generative Adversarial Networks
Very Deep VAEs Generalize Autoregressive Models and Can Outperform Them on Images
A Contrastive Learning Approach for Training Variational Autoencoder Priors
StyleGAN-XL: Scaling StyleGAN to Large Diverse Datasets
Diffusion-GAN: Training GANs with Diffusion
Improved Training of Wasserstein GANs
Self-Attention Generative Adversarial Networks
Large Scale GAN Training for High Fidelity Natural Image Synthesis
CSGAN: Cyclic-Synthesized Generative Adversarial Networks for Image-to-Image Transformation
LOGAN: Latent Optimisation for Generative Adversarial Networks
A U-Net Based Discriminator for Generative Adversarial Networks
Instance-Conditioned GAN
Conditional GANs with Auxiliary Discriminative Classifier
29.视频生成
Temporal Generative Adversarial Nets with Singular Value Clipping
Generating Videos with Scene Dynamics
MoCoGAN: Decomposing Motion and Content for Video Generation
Stochastic Video Generation with a Learned Prior
Video-to-Video Synthesis
Probabilistic Video Generation using Holistic Attribute Control
ADVERSARIAL VIDEO GENERATION ON COMPLEX DATASETS
Sliced Wasserstein Generative Models
Train Sparsely, Generate Densely: Memory-efficient Unsupervised Training of High-resolution Temporal GAN
Latent Neural Differential Equations for Video Generation
VideoGPT: Video Generation using VQ-VAE and Transformers
Diverse Video Generation using a Gaussian Process Trigger
NÜWA: Visual Synthesis Pre-training for Neural visUal World creAtion
StyleGAN-V: A Continuous Video Generator with the Price, Image Quality and Perks of StyleGAN2
Video Diffusion Models
30.超分辨率
Image Super-Resolution Using Deep Convolutional Networks
Deeply-Recursive Convolutional Network for Image Super-Resolution
Accelerating the Super-Resolution Convolutional Neural Network
Real-Time Single Image and Video Super-Resolution Using an Efficient Sub-Pixel Convolutional Neural Network
Photo-Realistic Single Image Super-Resolution Using a Generative Adversarial Network
Image Restoration Using Convolutional Auto-encoders with Symmetric Skip Connections
Accurate Image Super-Resolution Using Very Deep Convolutional Networks
Image super-resolution via deep recursive residual network
Deep Laplacian Pyramid Networks for Fast and Accurate Super-Resolution
Image Super-Resolution Using Very Deep Residual Channel Attention Networks
Image Super-Resolution via Dual-State Recurrent Networks
Recovering Realistic Texture in Image Super-resolution by Deep Spatial Feature Transform
Cascade Convolutional Neural Network for Image Super-Resolution
Image Super-Resolution with Cross-Scale Non-Local Attention and Exhaustive Self-Exemplars Mining
Single Image Super-Resolution via a Holistic Attention Network
One-to-many Approach for Improving Super-Resolution