GeneGAN: Learning Object Transfiguration and Attribute Subspace from Unpaired Data
2017-11-27 09:05
369 查看
本文的工作属于图像到图像间的翻译。类似于 DiscoGAN、CycleGAN 和 DualGAN,作者通过对偶学习,加上弱监督信息(weak 0/1 label),成功实现图像属性的迁移。G 采用 autoencoder 结构,encoder 将图像编码成图像主体信息(如:图像背景和人脸 ID 等)和属性信息(如:微笑、戴眼镜、发型等),decoder 则类似于 CGAN,将图像主体信息和属性信息翻译成图像。通过修改输入
decoder 的属性信息,实现属性的迁移。这种迁移成功的关键在于对偶学习机制(L1 重构误差),以及平行四边形 loss 和 nulling loss。值得一提的是,跟 CycleGAN 等相比,GeneGAN 只需要 generator 和 discriminator 各一个。文章在 CelebA 和 Multi PIE 数据集上进行实验,得到了不错的人脸属性迁移效果。 亮点推荐:推荐学习文章对弱监督信息的处理技巧。
代码地址:https://github.com/Prinsphield/GeneGAN
论文链接:https://arxiv.org/abs/1705.04932
推荐人:洪佳鹏,北京大学(PaperWeekly
arXiv组志愿者)
decoder 的属性信息,实现属性的迁移。这种迁移成功的关键在于对偶学习机制(L1 重构误差),以及平行四边形 loss 和 nulling loss。值得一提的是,跟 CycleGAN 等相比,GeneGAN 只需要 generator 和 discriminator 各一个。文章在 CelebA 和 Multi PIE 数据集上进行实验,得到了不错的人脸属性迁移效果。 亮点推荐:推荐学习文章对弱监督信息的处理技巧。
代码地址:https://github.com/Prinsphield/GeneGAN
论文链接:https://arxiv.org/abs/1705.04932
推荐人:洪佳鹏,北京大学(PaperWeekly
arXiv组志愿者)
相关文章推荐
- #Paper Reading# Topic Modeling using Topics from Many Domains, Lifelong Learning and Big Data
- 【每周一文】Learning Classifiers from Only Positive and Unlabeled Data(2008)
- (论文分析) Machine Learning -- Learning from labeled and unlabeled data
- 不平衡学习的方法 Learning from Imbalanced Data
- How-to: make spark streaming collect data from Kafka topics and store data into hdfs
- Retrieve data from xml and show in listField
- [NLP论文阅读] Supervised Learning of Universal Sentence Representations from NLI data
- Recognizing and Learning Object Categories --- 连接放送
- How to: Send and Receive Large Amounts of Data to and from a Web Service
- AttributeError: 'Bunch' object has no attribute 'data'
- Reading Note: DSOD: Learning Deeply Supervised Object Detectors from Scratch
- 行人属性“Weakly-supervised Learning of Mid-level Features for Pedestrian Attribute Recognition and Loca”
- Read data from excel into the object of DataSet.
- Self-taught Learning: Transfer Learning from Unlabeled Data
- Machine Learning and Data Mining——1. 机器学习和数据挖掘领域的领军人物
- Some 3D Graphics (rgl) for Classification with Splines and Logistic Regression (from The Elements of Statistical Learning)(转)
- 'DataFrame' object has no attribute 'sort'
- The method update(String, Object...) from the type JdbcTemplate refers to the missing type DataAcces
- Note for video Machine Learning and Data Mining——error and noise
- Using Presto to combine data from Hive and MySQL in one SQL-like query