Deep neural network project from M.sc in software engineering.
A different approach to text to image synthesis that show in this paper here
- python >=3.6
- tensorflow-gpu >=1.11
we define a simple approach that trained on fasion_mnist dataset.
A simple GAN model .
A simple CNN
-
to run the main Algorithm:
runpython ./fashion_GAN/main.py --label <one of the labels name below>
- ['t_shirt', 'trouser', 'pullover', 'dress', 'coat', 'sandal', 'shirt', 'sneaker', 'bag', 'ankle_boots']
-
to train the cnn model:
run thefashion_cnn_tpu.ipynb
notebook on colab -
to train the gan model (V2 is reccomended):
run thefashion_mnist_gan_v1/v2.ipynb
notebook on colab -
to see the CNN model graph on tensorboard:
runtensorboard --logdir ./fashion_GAN/fashion_cnn/logs
-
to see the GAN model graph on tensorboard:
runtensorboard --logdir ./fashion_GAN/fashion_gan/logs