createamind / busyplan

plan
11 stars 3 forks source link

videogan资料 #4

Open zdx3578 opened 6 years ago

zdx3578 commented 6 years ago

gan video 两周 第一周, 熟悉已有论文,选用一个算法复现。使用自己的数据集,调参。 第二周,分析隐变量的语义相关信息,自动驾驶的转向角度和z的关系 。

zdx3578 commented 6 years ago

ref: VAE: Adversarial Variational Bayes: Unifying VAE and GAN beta vae

GAN:

https://scholar.google.com/scholar?um=1&ie=UTF-8&lr&cites=7977179880967579431 https://scholar.google.com/scholar?um=1&ie=UTF-8&lr&cites=12629733064507558057

waxz commented 6 years ago

note 1.Temporal Generative Adversarial Nets with Singular Value Clipping

2.Generating Videos with Scene Dynamics

3.Learning to Generate Time-Lapse Videos Using Multi-Stage Dynamic Generative Adversarial Networks

  1. MoCoGAN: Decomposing Motion and Content for Video Generation
    • 引入动作和内容先验知识,将噪声Z分为Zc内容噪声和Zm动作噪声
    • 利用RNN从动作空间获得Zm分布Rm,随机变量e --> RNN --> Zm
    • 通过锁定Zm或Zc改变动作或环境
    • 整体模型含4个网络,分别为:
    • 获取动作分布 - GRU
    • 生成器Gi 、 判别器Di - DCGAN - 控制图像生成质量
    • 判别器Dv - spatio-temporal CNN 控制视频&动作生成质
    • Conditional MoCoGAN, 将label与E一同输入RNN
fzd9752 commented 6 years ago

各种GAN和VAE的代码:

Tensorflow https://github.com/hwalsuklee/tensorflow-generative-model-collections

PyTorch https://github.com/znxlwm/pytorch-generative-model-collections