seungwookim / vae.tensorflow.slim

Variational Auto-Encoder with TensorFlow & slim

Requirements

• TensorFlow 1.1.0 or greater (?)
• opencv (for generate.py)
• numpy

Training

editing train.sh

• Set the "TRAIN_DIR" (where to save checkpoint) to path you want
• Set the hyper-parameters

train.sh

# Working directory
WORKING_DIR=$HOME/projects

# Where the training (fine-tuned) checkpoint and logs will be saved to.
TRAIN_DIR=$WORKING_DIR/vae.tensorflow.slim/exp1

CUDA_VISIBLE_DEVICES=0 \
python train.py \
    --train_dir=${TRAIN_DIR} \
    --batch_size=64 \
    --max_steps=100000 \
    --save_steps=5000 \

run ./train.sh

$ ./train.sh

• You can use tensorboard for monitoring loss and generated images

$ tensorboard --logdir=exp1

Generate images

generate.sh

# Working directory
WORKING_DIR=$HOME/projects

# Where the training (fine-tuned) checkpoint and logs will be saved to.
TRAIN_DIR=$WORKING_DIR/vae.tensorflow.slim/exp1

batch=$1

CUDA_VISIBLE_DEVICES=0 \
python image_translate.py \
    --checkpoint_path=${TRAIN_DIR} \
    --checkpoint_step=-1 \
    --batch_size=$batch \
    --seed=12345 \
    --save_step=1000 \

convert -delay 30 -loop 0 *.jpg generated_images.gif

run ./generate.sh

$ ./generate.sh batch_size (the number of images you want)

Results

MNIST datasets

Epoch 1

Epoch 100

Gif

References

Papers

• Diederik P. Kingma, Max Welling. Auto-Encoding Variational Bayes. arXiv: 1312.6114
• Carl Doersch. Tutorial on Variational Autoencoders. arXiv: 1606.05908

Blogs

• Tutorial - What is a variational autoencoder? (https://jaan.io/what-is-variational-autoencoder-vae-tutorial)
• Variational Autoencoder: Intuition and Implementation (http://wiseodd.github.io/techblog/2016/12/10/variational-autoencoder)

My presentation file (DeepLAB in ModuLabs)

• http://www.modulabs.co.kr/DeepLAB_library/13347

Author

Il Gu Yi