2021/11/22 Updated by

Google Colab

Japanese

English

Deep Learning with Tensorflow 2 on Gooble Colab

AutoEncoder (AE) for MNIST dataset

AE_MNIST_Train: html ipynb
AE_MNIST_Analysis: html ipynb
outputs: AE01 AE02 AE03

VariationalAutoEncoder (VAE) for MNIST dataset

VAE_MNIST_Train: html ipynb
VAE_MNIST_Analysis: html ipynb
output: VAE01 VAE02 VAE03

VariationalAutoEncoder (VAE) for CelebA dataset

VAE_CelebA_Train: html ipynb
VAE_CelebA_Train2: html ipynb
VAE_CelebA_Analysis: html ipynb
outputs: VAE_CelebA01 VAE_CelebA02 VAE_CelebA03

Neural Style Transfer

NeuralStyleTransfer_paris: html ipynb

Generative Adversarial Network (GAN) for Camel dataset

GAN_Camel_Train: html ipynb
outputs: GAN_Camel01

Wasserstein Generative Adversarial Network (WGAN) for CIFER-10 dataset

WGAN_CIFER_Train: html ipynb
outputs: WGAN_CIFAR01

Wasserstein Generative Adversarial Network with Gradients Penalty (WGAN-GP) for CelebA dataset

WGAN_GP_CelebA_Train: html ipynb
outputs: WGAN_GP_CelebA01

CycleGAN for VidTIMIT (Faces) dataset

CycleGAN_VidTIMIT_Train: html ipynb (run locally on generative env: html ipynb )
CycleGAN_VidTIMIT_Train2: html ipynb (run locally on generative env: html ipynb )
CycleGAN_VidTIMIT_Analysis: html ipynb (run locally on generative env: html ipynb )
outputs: CycleGAN_VidTIMIT01

LSTM Music

LSTM_MusicCompose_Train: html ipynb (run locally on generative env: html ipynb run nw/LSTMMusic.py ) html (with model.summary)
LSTM_MusicCompose_Train2: html ipynb (run locally on generative env: html ipynb run )
LSTM_MusicCompose_Analysis: html ipynb (run locally on generative env: html ipynb run )

Google Colab 上でファイルを Google Drive に作成すると、Google Drive の 'マイドライブ/Colab Notebooks' にファイルができる。

Keras / TPU integration in Tensorflow 2.x
fit()を使うtraining例: html ipynb
GradientTape()を使うtraining例: html ipynb
MNIST: html ipynb

メモ