main.py

# BSD 3-Clause License
#
# Copyright (c) 2021 BBC
# All rights reserved.
#
# Redistribution and use in source and binary forms, with or without
# modification, are permitted provided that the following conditions are met:
#
#  1. Redistributions of source code must retain the above copyright notice,
#  this list of conditions and the following disclaimer.
#
#  2. Redistributions in binary form must reproduce the above copyright notice,
#  this list of conditions and the following disclaimer in the documentation
#  and/or other materials provided with the distribution.
#
#  3. Neither the name of the copyright holder nor the names of its contributors may
#  be used to endorse or promote products derived from this software without
#  specific prior written permission.
#
# THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS "AS IS"
# AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
# IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
# ARE DISCLAIMED. IN NO EVENT SHALL THE COPYRIGHT HOLDER OR CONTRIBUTORS
# BE LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR
# CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF
# SUBSTITUTE GOODS OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS
# CONTRACT, STRICT LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE)
# ARISING IN ANY WAY OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF
# THE POSSIBILITY OF SUCH DAMAGE.

from model_scratch import ScratchCNN, ScratchActCNN, ScratchBiasCNN, ScratchAllCNN, ScratchOneCNN, SRCNN
from model_shared import SharedCNN
from model_competition import CompetitionCNN
from utils import import_path, frac_positions
from itertools import product
import tensorflow as tf
import os
import argparse


if __name__ == '__main__':
    parser = argparse.ArgumentParser(description='CNN Subpixel Motion Interpolation')
    parser.add_argument('-c', '--config', type=str, default=None, required=True, help='Config filepath')
    parser.add_argument('-a', '--action', type=str, default=None, required=True, help='train or test')
    parser.add_argument('-g', '--gpu', type=int, default=0, help='CUDA device number')
    parser.add_argument('-gf', '--gpu_fraction', type=float, default=1.0, help='Fraction of GPU usage')
    args = parser.parse_args()

    os.environ["CUDA_VISIBLE_DEVICES"] = str(args.gpu)

    # import config file details
    model_cfg = import_path(args.config)

    # create directories where necessary
    os.makedirs(model_cfg.checkpoint_dir, exist_ok=True)
    os.makedirs(model_cfg.results_dir, exist_ok=True)
    os.makedirs(model_cfg.graphs_dir, exist_ok=True)

    # loop through all fractional shifts and QPs the network has to be trained for
    combined = 1 if model_cfg.model_name == "sharedcnn" or model_cfg.model_name == "competitioncnn" else 0
    frac_xy = ["all"] if combined else frac_positions()
    qp_list = [27] if combined else [22, 27, 32, 37]

    # prepare dictionaries that will store the corresponding filters
    for frac, qp in product(frac_xy, qp_list):
        model_cfg.fractional_pixel = frac
        model_cfg.qp = qp

        tf.reset_default_graph()
        gpu_options = tf.GPUOptions(per_process_gpu_memory_fraction=args.gpu_fraction)
        with tf.Session(config=tf.ConfigProto(gpu_options=gpu_options)) as sess:

            # initialise the specified cnn model
            if model_cfg.model_name == "scratchcnn":
                cnn_model = ScratchCNN(sess, model_cfg)
            elif model_cfg.model_name == "scratchcnn_activation":
                cnn_model = ScratchActCNN(sess, model_cfg)
            elif model_cfg.model_name == "scratchcnn_bias":
                cnn_model = ScratchBiasCNN(sess, model_cfg)
            elif model_cfg.model_name == "scratchcnn_all":
                cnn_model = ScratchAllCNN(sess, model_cfg)
            elif model_cfg.model_name == "scratchcnn_onelayer":
                cnn_model = ScratchOneCNN(sess, model_cfg)
            elif model_cfg.model_name == "srcnn":
                cnn_model = SRCNN(sess, model_cfg)
            elif model_cfg.model_name == "sharedcnn":
                cnn_model = SharedCNN(sess, model_cfg)
            elif model_cfg.model_name == "competitioncnn":
                cnn_model = CompetitionCNN(sess, model_cfg)
            else:
                raise ValueError("Invalid model name specified!")

            # train or test the model
            if args.action == "train":
                if not os.path.exists(model_cfg.dataset_dir):
                    raise ValueError("Invalid training dataset directory specified!")
                cnn_model.train()
            elif args.action == "test":
                if not os.path.exists(model_cfg.test_dataset_dir):
                    raise ValueError("Invalid testing dataset directory specified!")
                cnn_model.test()
            else:
                raise ValueError("Invalid action specified!")