def deblur1(path):
args = {
'model': 'color',
'datalist':
'C:/Users/smita/Desktop/face-group-recognition-facerecognition-api-updated/datalist_gopro.txt',
'batch_size': 16,
'epoch': 4000,
'lr': 1e-4,
'gpu': '0',
'phase': 'test'
}
print(args['model'])
# args = parse_args()
# set gpu/cpu mode
# if int(args.gpu_id) >= 0:
# os.environ['CUDA_VISIBLE_DEVICES'] = args.gpu_id
# else:
os.environ['CUDA_VISIBLE_DEVICES'] = ''
# set up deblur models
deblur = model.DEBLUR(args)
if args['phase'] == 'test':
out = "hello.png"
res = deblur.test(720, 1280, path, out)
path = 'static/Quality/' + str(time.time()) + ".png"
cv2.imwrite(path, res)
# print(x)
return path
elif args['phase'] == 'train':
deblur.train()
else:
print('phase should be set to either test or train')
def main(_):
args = parse_args()
# set gpu/cpu mode
if int(args.gpu_id) >= 0:
os.environ['CUDA_VISIBLE_DEVICES'] = args.gpu_id
device = '/gpu:{}'.format(args.gpu_id)
else:
os.environ['CUDA_VISIBLE_DEVICES'] = ''
device = '/cpu:0'
config = tf.ConfigProto(allow_soft_placement=True)
config.gpu_options.allow_growth = True
# set up deblur models
with tf.device(device):
with tf.Session(config=config) as sess:
deblur = model.DEBLUR(args)
if args.phase == 'test':
deblur.test(sess, args.height, args.width, args.input_path, args.output_path)
elif args.phase == 'train':
deblur.train(sess)
else:
print('phase should be set to either test or train')
def main(_):
args = parse_args()
os.environ['TF_ENABLE_AUTO_MIXED_PRECISION'] = '1'
os.environ['TF_CPP_VMODULE'] = "auto_mixed_precision=2"
# set gpu/cpu mode
if int(args.gpu_id) >= 0:
os.environ['CUDA_VISIBLE_DEVICES'] = args.gpu_id
else:
os.environ['CUDA_VISIBLE_DEVICES'] = ''
# set up deblur models
deblur = model.DEBLUR(args)
if args.phase == 'test':
deblur.test(args.height, args.width, args.input_path, args.output_path)
elif args.phase == 'train':
deblur.train()
elif args.phase == 'check':
deblur.check(args.height, args.width)
elif args.phase == 'convert':
deblur.convert_tflite(args.height, args.width)
elif args.phase == 'eval':
deblur.eval(file_dir=args.eval_path)
else:
print('phase should be set to either test or train')
def main(_):
args = parse_args()
# set gpu/cpu mode
if int(args.gpu_id) >= 0:
os.environ['CUDA_VISIBLE_DEVICES'] = args.gpu_id
else:
os.environ['CUDA_VISIBLE_DEVICES'] = ''
# set up deblur models
M = model.DEBLUR(args)
if not os.path.exists(args.output_path):
os.makedirs(args.output_path)
with tf.Graph().as_default():
before = tf.placeholder(tf.float32,
shape=[1, args.height, args.width, 3])
current = tf.placeholder(tf.float32,
shape=[1, args.height, args.width, 3])
after = tf.placeholder(tf.float32,
shape=[1, args.height, args.width, 3])
predict = M.generator(before, current, after, istrain=False)
saver = tf.train.Saver()
with tf.Session() as sess:
checkpoint_dir = os.path.join('./checkpoints/', args.name)
load_checkpoint(sess, checkpoint_dir, saver)
imgsName = sorted(os.listdir(args.input_path))
time_start = time.time()
for imgName in imgsName:
print('Processing: %s' % imgName)
img = cv2.imread(os.path.join(args.input_path, imgName))
inputs = split_input(img)
inputs = [np.expand_dims(img, 0) for img in inputs]
result = sess.run(predict,
feed_dict={
before: inputs[0],
current: inputs[1],
after: inputs[2]
})
result = result[1][0, :, :, :]
result = im2uint8(result)
cv2.imwrite(os.path.join(args.output_path, imgName), result)
time_end = time.time()
process_time = (time_end - time_start)
print('avage process time = %.4f' % (process_time / len(imgsName)))
def main(_):
args = parse_args()
# set gpu/cpu mode
if int(args.gpu_id) >= 0:
os.environ['CUDA_VISIBLE_DEVICES'] = args.gpu_id
else:
os.environ['CUDA_VISIBLE_DEVICES'] = ''
# set up deblur models
deblur = model.DEBLUR(args)
if args.phase == 'test':
deblur.test(args.height, args.width, args.input_path, args.output_path)
elif args.phase == 'train':
deblur.train()
else:
print('phase should be set to either test or train')
def main(_):
args = parse_args()
# Set gpu/cpu mode.
if int(args.gpu_id) >= 0:
os.environ['CUDA_VISIBLE_DEVICES'] = args.gpu_id
else:
os.environ['CUDA_VISIBLE_DEVICES'] = '-1'
# Set up deblur models.
deblur = model.DEBLUR(args)
if args.phase == 'test':
deblur.validate()
elif args.phase == 'train':
deblur.train()
else:
print('Runtime error: args.phase can be only [train, test]')
exit(1)
def main(_):
args = parse_args()
# set gpu/cpu mode
if int(args.gpu_id) >= 0:
os.environ['CUDA_VISIBLE_DEVICES'] = args.gpu_id
else:
os.environ['CUDA_VISIBLE_DEVICES'] = ''
checkpoint_dir = os.path.join('./checkpoints/', args.name)
logger.set_logger_dir(checkpoint_dir)
# set up deblur models
M = model.DEBLUR(args)
ds_train = get_data(args.dataroot,
phase='train',
crop_size=args.cropSize,
batch_size=args.batchSize)
ds_val = get_data(args.dataroot,
phase='val',
crop_size=args.cropSize,
batch_size=args.batchSize)
trainer = SeparateGANTrainer(ds_train, M, g_period=6)
trainer.train_with_defaults(
callbacks=[
ModelSaver(max_to_keep=5, checkpoint_dir=checkpoint_dir),
ScheduledHyperParamSetter('learning_rate',
[(300, args.learning_rate),
(args.max_epoch, 0)],
interp='linear'),
InferenceRunner(ds_val, [
ScalarStats('PSNR_BASE'),
ScalarStats('PSNR_2'),
ScalarStats('PSNR_IMPRO2'),
ScalarStats('pixel_loss2'),
ScalarStats('feature_loss2')
])
],
session_init=SaverRestore(checkpoint_dir +
'/model-431249.data-00000-of-00001')
if args.continue_train else None,
starting_epoch=1,
steps_per_epoch=args.steps_per_epoch,
max_epoch=args.max_epoch)
def main(_):
args = parse_args()
deblur = model.DEBLUR(args)
deblur.test_one(args.height, args.width, args.input_path, args.output_path)
global deblur
global train_dir
global graph
global sess
global ALLOWED_EXTENSIONS
ALLOWED_EXTENSIONS = set(['png', 'jpg', 'jpeg'])
upload_directory = '/src/upload/'
create_directory(upload_directory)
checkpoint_dir = "/src/checkpoints/"
create_directory(checkpoint_dir)
url_prefix = 'http://pretrained-models.auth-18b62333a540498882ff446ab602528b.storage.gra.cloud.ovh.net/image/SRN-Deblur/'
model_zip = "srndeblur_models.zip"
get_model_bin(url_prefix + model_zip, checkpoint_dir + model_zip)
os.system("cd " + checkpoint_dir + " && unzip " + model_zip)
checkpoint_dir = os.path.join(checkpoint_dir, args.model)
deblur = model.DEBLUR(args)
port = 5000
host = '0.0.0.0'
app.run(host=host, port=port, threaded=True)