先前NVIDIA Research在CVPR 2018提出了pix2pixHD的方法，將Image to image translation的畫質提升到了另一個境界之後，其原班人馬最近又上傳了一篇效果令人驚豔的論文vid2vid：利用已有的影片語意分割（video semantic maps) 當做輸入，去操作產生維持原本語意（semantic）的新影片。效果上不但維持了原本的high resolution結果，也在每一個video frame之間，保持了很好的temporal smoothness。在結果的轉換影片中可以看到，NVIDIA為了在道路上生成更多的街景資料下了許多功夫，藉由調整instance level的generator，可以生成許多種不同的街景style，除了建築的部分有一些明顯可見的artifact之外，其餘物體間的轉換與流暢度的效果真的很好。有興趣瞭解更多細節的讀者可以參考以下資源：
Project: https://tcwang0509.github.io/vid2vid/
Code: https://github.com/NVIDIA/vid2vid
Tutorial Video: https://www.youtube.com/watch?v=S1OwOd-war8
NVIDIA提出最新影像操作合成技術Vid2Vid
近期文章
近期迴響
- Deep Learning Resources – 難得糊塗 在 多目標深度學習框架之概念與應用：Multi-Task Learning in Deep Neural Networks
- 益达 在 在3D Point Cloud Data上有效地使用深度學習取特徵 PointNet: Deep Learning on Point Sets for 3D Classification and Segmentation
- Chien-Yi Wang 在 在3D Point Cloud Data上有效地使用深度學習取特徵 PointNet: Deep Learning on Point Sets for 3D Classification and Segmentation
- 益达 在 在3D Point Cloud Data上有效地使用深度學習取特徵 PointNet: Deep Learning on Point Sets for 3D Classification and Segmentation
- json formatter online 在 網路爬蟲系列 (Crawler) (1) : 使用Chrome「檢查元素」找出 PChome購物網站 資料位置
彙整
分類
- 20150704 蟲友小聚
- 3D Computer Vision
- a crawler a day
- About 木刻思
- AC-GAN
- Actor-Critic
- Adversarial Autoencoder (AAE)
- AI Gaming
- AI Simulator
- Alibaba
- AndrewNg
- Applications
- archlinux 各種雷
- Attention
- Autoencoder
- Backpropagation
- Baidu
- Bayesian
- BEGAN
- Ben 專欄
- Berkeley
- Berkeley
- Business Intelligence
- Caffe
- Capsules
- Catastrophic Forgetting
- cGAN
- Chatbot
- Chrome Extensions
- Classics
- Clustering
- CMU
- CNN
- CNN in NLP
- Codes & Projects
- Computer Vision
- Continuous Learning
- Course recommend
- courses
- Crawler 102
- crawler skills
- cross-lingual embeddings
- Cross-Modality
- css selector
- Data Science
- DataCamp
- Deconvolutional
- Deep Learning
- Deep Structured Semantic Model
- DeepMind
- DeepTransfer
- Dialogue Systems
- Distributed Training
- DNC
- Domain Confusion
- DyNet
- EBGAN
- ECommerce
- EndToEnd
- Events & Study Groups
- Events & Study Groups & Hackathons
- FB Graph API
- Few-shot Learning
- Finance
- Gambling
- Gaming Agent Envi
- GAN
- generative ladder network
- GET
- Git101
- Google Brain
- GoogleBrain
- Gradient Descent
- Highlighted Research Paper
- homepage
- Image generation
- Image style transfer
- Image-to-Image Translation
- IMAGENET
- Instagram API
- Instance Segmentation
- Intrinsic Reward
- json parser
- Keras
- KL-divergence
- Language Model
- linux problems
- LSGAN
- LSTM
- Machine Translation
- Mask RCNN
- Meta Learning
- Microsoft
- Microsoft R Server / Revolution Analytics
- MILA
- mLSTM
- Model-based
- Multi-agent
- Multi-Loss
- Multi-task Learning
- N-gram
- Natural Language Generation
- Natural Language Interface
- Newsletters
- NLP
- Nvidia
- Object Detection
- OpenAI
- Optimization Technique
- Oxford
- Partial Occlusion
- PatchMatch
- PhotoStyleTransfer
- Planning
- Policy Gradient
- Pose Estimation
- Posts & Articles
- Power BI
- Pre-train
- PyCrawler 101
- PyCrawler 102
- PyData
- PyMC
- Python
- Python
- Python
- PyTM 101
- PyTorch
- Quantitative Trading
- quantmod
- quantstrat
- R
- R Crawler 102
- RCNN
- RCrawler101
- Recommender Systems
- REGEL
- Reinforcement Learning
- Relational Reasoning
- Representation Learning
- ResNet
- RNN
- Robustness
- RStudio
- Salesforce
- Saliency Maps
- Self-DrivingCar
- Semantic Parsing
- Semantic Segmentation
- Semi-supervised Learning
- Sentiment Analysis
- Seq2Seq
- sharing
- sharing links
- Socher
- Social Media
- SpeechSynthesis
- Sport
- Standford
- Strategy Mining
- StyleTransfer
- table parser
- TAFB
- Tags
- Tensorflow
- Text Classification
- TextSummarization
- TextToSpeech
- Tools
- Trading
- Transfer Learning
- Trevor Darrell
- Tutorials & Learning Materials
- Twitter API
- U-Net
- Uncategorized
- Unsupervised Learning
- Useful Tools
- VAE
- Value Iteration
- Variational Autoencoder (VAE)
- Variational Inference
- Video Understaning
- Visualization
- Web Crawler
- Web Crawlers in Python
- Web Crawlers in R
- WGAN
- Windows 各種雷
- Word Embedding
- Word Embeddings
- xpath selector
- Zero-shot Learning
- 木刻思專欄
- 蟲友小聚
其它
Written by:Chien-Yi Wang
Be First to Comment