Deep Learning Points

논문 읽기

• 210408: Mask R-CNN + R-CNN 정리 pdf
• 210415: MAML
• 210422: Everybody Dance Now (GAN) md
• 210429: StyleGAN md
• 210506: Variational Autoencoders for Collaborative Filtering
• 210506: Objectron: A Large Scale Dataset of Object-Centric Videos in the wild with pose annotations closed issue
• 210527: ViT: An Image is Worth 16x16 Words: Transformers for Image Recognition at Scale pdf
• 210603: DALL-E: Zero-Shot Text-to-Image Generation issue
• 210729: Multi-Source Domain Adaptation with Collaborative Learning for Semantic Segmentation 정리
• 210730: Incremental Few-Shot Instance Segmentation issue
• 210803: BOP Challenge 2020 on 6D Object Localization closed issue
• 210805: MLP-Mixer: An all-MLP Architecture for Vision closed issue
• 210811: Pix2Pose: Pixel-Wise Coordinate Regression of Objects for 6D Pose Estimation issue
• 210812: Multimodal Transformer for Unaligned Multimodal Language Sequences issue
• 210813: A Survey of Uncertainty in Deep Neural Networks issue
• 210815: Uncertainty in Computer Vision(Papers) issue
• 210823: Bayesian Neural Network issue
• 210826: MART: Memory-Augmented Recurrent Transformer for Coherent Video Paragraph Captioning (ACL2020) issue
• 210825-: Neural Process issue
• 211006: Generalization Through Hand-Eye Coordination: An Action Space for Learning Spatially-Invariant Visuomotor Control pdf
• 211012: SimSiam: Exploring Simple Siamese Representation Learning md
• 211021: Prototypical Cross-domain Self-supervised Learning for Few-shot Unsupervised Domain Adaptation md
• 211031: Zero-shot instance segmentation pdf
• 211102-211207: A Probabilistic U-Net for Segmentation of Ambiguous Images md
• 211103: 3D Attention-Driven Depth Acquisition for Object Identification md
• 211107-: Domain Randomization Issues
• 211111-: Integrated Task and Motion Planning md
• 211114: MQA: Answering the Question via Robotic Manipulation (RSS 2021) md
• 211124: SOLO: A Simple Framework for Instance Segmentation(NeurIPS 2020) md
• 211123: Efficient learning of goal-oriented push-grasping synergy in clutter(RA-L2021) md
• 211130: Reinforcement Learning of Active Vision for Manipulating Objects under Occlusions (CoRL2018) md
• 211201-: Set related papers: Deep sets, Set transformer etc. Issue
• 211203: Swin Transformer: Hierarchical Vision Transformer using Shifted Windows Issue
• 211211-: Transformer Issue
• 211219: Video Instance Segmentation Tracking with a Modified VAE Architecture md
• 220110: Cascade R-CNN: Delving into High Quality Object Detection md, [pdf]()
• 220111: Hybrid Task Cascade for Instance Segmentation paper
• 220115: SNP: Sequential Neural Processes, ASNP: Attentive Sequential Neural Processes, RMR: Robustifying Sequential Neural Processes md
• 2202-2203: Uncertainty estimation issue
• 220328-: Pushing in clutter issue
• 220414: Panoptic-DeepLab: A Simple, Strong, and Fast Baseline for Bottom-Up Panoptic Segmentation Issue

Theory Concept

• Real Basics of Whole Deep Learning Networks
• CNN: Object Detection/Semantic Segmentation/Instance Segmentation
  • AlexNet, ResNet, VGGNet, GoogLeNet, MobileNet
    • Skip Connection
  • YOLO, SSD
  • R-CNN, Fast R-CNN, Faster R-CNN, Mask R-CNN 전체 정리
    • FPN(Feature Pyramid Net)
    • Selective Search, RPN(Region Proposal Network)
    • RoI Pooling, RoI Align
  • Transformer & Vision
    • ViT: Transformer & Vision pdf
  • 3D Object Detection
    • Objectron closed issue
• NLP
  • [x] Attention
  • [ ] Transformer
  • [ ] BERT
    • [ ] VideoBERT issue
    • [ ] ALBERT: A Lite BERT for self-supervised Learning of Language Representation issue
  • [ ] GPT (내용 보충 필요) 정리
• Multi-Modal
  • [ ] DALL-E: Zero-Shot Text-to-Image Generation issue
• Self-Supervised Learning
• Generative Models (코드 구현 연습 필요)
  • [x] AutoEncoder
    • VAE
  • [x] GAN
    • StyleGAN
• [x] Reinforcement Learning
  • Markov Property, Markov Decision Process, Markov Reward Process
  • Dynamic Programming: Bellman Equation, Policy Iteration, Value Iteration
  • Model-Free Prediction: Monte Carlo, Temporal Difference
  • Model-Free Control: MC control, TD Control(SARSA, Q-Learning)
  • Value Function Approximation: Linear Value Function Approximation, Neural Network(DQN), Batch Reinforcement Learning
  • REINFORCE, Policy Gradient, Actor Critic
• Methodology
  • Neural Process: CNP, LNP, SNP, ASNP, RMR
• Meta Learning
• Skils
  • Transfer learning