Code for ``.
Install TensorFlow and h5py. The code has been tested with Python 3.6, TensorFlow 1.7.0, CUDA 9.1 and cuDNN 7.1 on ArchLinux.
To train or evaluate or visualize:
python run_cls.py
Log files and network parameters will be saved to log_model_dataset folder in default.
We can use TensorBoard to view the network architecture and monitor the training progress.
tensorboard --logdir ./log_model_dataset