- tf training stackoverflow
- document with tf code
- sample code
- stargan
- cramer gan
- nvidia gan
- EFFICIENT GAN-BASED ANOMALY DETECTION
- ADVERSARIAL FEATURE LEARNING
- variable scope
- import images in tf.train.batch
- graphs for training/testing dataset
- freeze partial variables
- save/restore with dataset
- ema, exponential moving average
- name_scope vs variable_scope
- saved model
- use training/validation dataset in the training phase
- Frozen graph with variables:
python -m tensorflow.python.tools.freeze_graph \
--input_checkpoint ./model.ckpt \
--input_graph ./nasnet_inf_graph.pb \
--input_binary True \
--output_graph ./frozen_graph.pb