])
video_transforms = functools.partial(video_transform,
image_transform=image_transforms)
video_length = int(args['--video_length'])
image_batch = int(args['--image_batch'])
video_batch = int(args['--video_batch'])
dim_z_content = int(args['--dim_z_content'])
dim_z_motion = int(args['--dim_z_motion'])
dim_z_category = int(args['--dim_z_category'])
# dataset = data.VideoFolderDataset(args['<dataset>'], cache=os.path.join(args['<dataset>'], 'local.db'))
dataset = data.VideoFolderDataset(args['<dataset>'], cache=None)
image_dataset = data.ImageDataset(dataset, image_transforms)
image_loader = DataLoader(image_dataset,
batch_size=image_batch,
drop_last=True,
num_workers=2,
shuffle=True)
video_dataset = data.VideoDataset(dataset, 16, 2, video_transforms)
video_loader = DataLoader(video_dataset,
batch_size=video_batch,
drop_last=True,
num_workers=2,
shuffle=True)
generator = models.VideoGenerator(n_channels, dim_z_content,
dim_z_category, dim_z_motion,
else:
logging.info('Saving model parameters to %s...' % params_path)
with open(params_path, 'w') as f:
json.dump(params, f)
logging.info('Loading training dataset...')
train_set = data.PatchDataset(params['train_partitions'], params['batch_size'],
params['temporal_patch_size'],
params['spatial_patch_size'],
params['spatial_stride'])
logging.info('Loading validation dataset...')
validation_set = data.ImageDataset(params['validation_partitions'],
params['temporal_patch_size'])
logging.info('Loading test dataset...')
test_set = data.ImageDataset(params['test_partitions'],
params['temporal_patch_size'])
inputs = tf.placeholder(tf.float32)
ground_truth = tf.placeholder(tf.float32)
global_step = tf.Variable(0, trainable=False, name='global_step')
network = get_network(inputs, params)
base_loss = tf.losses.mean_squared_error(network.outputs, ground_truth)
weight_loss = params['weight_decay'] * tf.reduce_sum(
tf.stack([tf.nn.l2_loss(weight) for weight in network.weights]))
loss = base_loss + weight_loss
lambda x: x[:n_channels, ::],
transforms.Normalize((0.5, 0.5, .5), (0.5, 0.5, 0.5)),
])
video_transforms = functools.partial(video_transform, image_transform=image_transforms)
video_length = int(args['--video_length'])
image_batch = int(args['--image_batch'])
video_batch = int(args['--video_batch'])
dim_z_content = int(args['--dim_z_content'])
dim_z_motion = int(args['--dim_z_motion'])
dim_z_category = int(args['--dim_z_category'])
dataset = data.VideoFolderDataset(args['<dataset>'], cache=os.path.join(args['<dataset>'], 'local.db'))
image_dataset = data.ImageDataset(dataset,audio_dir, image_transforms)
image_loader = DataLoader(image_dataset, batch_size=image_batch, drop_last=True, num_workers=6, shuffle=True)
print('args[<dataset>',args['<dataset>'])
print('args[<dataset_test>',args['<dataset_test>'])
dataset_test = data_test.VideoFolderDataset(args['<dataset_test>'], cache=os.path.join(args['<dataset_test>'], 'local_test.db'))
image_dataset_test = data_test.ImageDataset(dataset_test,audio_dir_test, image_transforms)
image_loader_test = DataLoader(image_dataset_test, batch_size=image_batch, drop_last=True, num_workers=6, shuffle=False)
ImageModel = models.ImageConvNet().cuda()
audio_encoder = WaveGANDiscriminator512(model_size=model_size, ngpus=ngpus)
# video_dataset = data.VideoDataset(dataset, 16, 2, video_transforms)
# video_loader = DataLoader(video_dataset, batch_size=video_batch, drop_last=True, num_workers=6, shuffle=True)