def main():
args = parser.parse_args()
with open(args.constants, 'r') as f:
constants = bunchify(yaml.load(f))
logger.info("Loading image %s ..." % args.input)
img = cv2.imread(args.input, flags=cv2.IMREAD_COLOR)
# image scaled in 0-1 range
img = img / 255.0
logger.info("Generating dark channel prior ...")
dark_channel = steps.generate_dark_channel(img, constants)
logger.info("Estimating airlight ...")
airlight = steps.estimate_airlight(img, dark_channel, constants)
logger.info("Estimated airlight is %s", str(airlight))
logger.info("Estimating transmission map ...")
tmap = steps.estimate_tmap(dark_channel, constants)
logger.info("Smooth transmission map ...")
tmap = steps.smooth_tmap(img, tmap, constants)
logger.info("Dehazing image ...")
dehazed = steps.dehaze(img, airlight, tmap, constants)
tools.show_img([img, dehazed])
def main():
args = parser.parse_args()
modify_arguments(args)
# Resetting the graph and setting seeds
tf.reset_default_graph()
tf.set_random_seed(args.seed)
np.random.seed(args.seed)
with open(args.config_file, 'r') as stream:
args.config = bunchify(yaml.load(stream))
logger.info(args)
if args.mode == 'train':
train(args)
else:
test(args)
def main():
"""The main method of script."""
args = parser.parse_args()
with open(args.config_file, 'r') as stream:
args.config = bunchify(yaml.load(stream))
args.train_dir = os.path.join(args.train_dir, args.job_id)
args.best_dir = os.path.join(args.best_dir, args.job_id)
if not os.path.exists(args.train_dir):
os.makedirs(args.train_dir)
if not os.path.exists(args.best_dir):
os.makedirs(args.best_dir)
logger.info(args)
if args.mode == 'train':
train(args)
elif args.mode == 'test' or args.mode == 'valid':
test(args)
elif args.mode == 'generate':
generate(args)
elif args.mode == 'beam':
diverse_beam_search(args)
def main():
args = parser.parse_args()
modify_arguments(args)
# Resetting the graph and setting seeds
tf.reset_default_graph()
tf.set_random_seed(args.seed)
np.random.seed(args.seed)
random.seed(args.seed)
with open(args.config_file, 'r') as stream:
config = yaml.load(stream)
args.config = Munch(modify_config(args, config))
logger.info(args)
if args.mode == 'train':
train(args)
elif args.mode == 'test':
test(args)
elif args.mode == 'analysis':
analysis(args)
def main():
args = parser.parse_args()
with open(args.constants, 'r') as f:
constants = bunchify(yaml.load(f))
logger.info("Loading image %s ..." % args.input)
img = cv2.imread(args.input, flags=cv2.IMREAD_COLOR)
# image scaled in 0-1 range
img = img / 255.0
# Scale array must be in decreasing order
scaled_imgs = steps.scale(
img,
[1, 300.0 / 384, 200.0 / 384, 150.0 / 384, 120.0 / 384, 100.0 / 384])
if not args.no_cache:
patches, pairs = load(args.input)
else:
patches, pairs = None, None
if patches is None and pairs is None:
logger.info("Extracting alternate patches ...")
patches = steps.generate_patches(scaled_imgs, constants, False)
print("\nNumber of patches extracted per scaled image")
print(map(len, patches))
logger.info("Smoothening std deviations of patches ...")
steps.smoothen(scaled_imgs, patches, constants)
logger.info("Putting patches in buckets ...")
steps.set_patch_buckets(patches, constants)
logger.info("Generating pairs of patches ...")
pairs = steps.generate_pairs(patches, constants)
print("\nNumber of pairs generated using generate_pairs")
print(len(pairs))
# logger.info("Saving patches and pairs ...")
# save(args.input, patches, pairs)
else:
logger.info("Using saved patches and pairs ...")
logger.info("Filtering pairs for checking normalized correlation ...")
pairs = steps.filter_pairs(patches, pairs, constants, all_pairs=False)
print("\nNumber of pairs retained after filtering")
print(len(pairs))
logger.info("Removing outliers ...")
pairs = steps.remove_outliers(pairs, constants)
print("\nNumber of pairs retained after removing outliers")
print(len(pairs))
logger.info("Estimating global airlight ...")
airlight = steps.estimate_airlight(pairs)
logger.info("Estimated airlight is ...%s", str(airlight))
# T-map estimation code begins
del patches
gc.collect()
logger.info("Extracting ALL patches ...")
patches = steps.generate_patches([img], constants, True)
logger.info("Estimating t-map ...")
dehazed = tmap_steps.estimate_tmap(img, patches, pairs, airlight,
constants)
logger.info("Displaying dehazed output image ...")
h, w = len(img), len(img[0])
img = img[0:h - 7, 0:w - 7, :]
tools.save_img([img, dehazed])
def main():
args = parser.parse_args()
with open(args.constants, 'r') as f:
constants = bunchify(yaml.load(f))
logger.info("Loading image %s ..." % args.input)
img = cv2.imread(args.input, flags=cv2.IMREAD_COLOR)
# image scaled in 0-1 range
img = img / 255.0
# Scale array must be in decreasing order
# scaled_imgs = steps.scale(img, [1, 0.75, 0.5, 0.375, 0.3, 0.25])
scaled_imgs = steps.scale(
img,
[1, 300.0 / 384, 200.0 / 384, 150.0 / 384, 120.0 / 384, 100.0 / 384])
if not args.no_cache:
patches, pairs = load(args.input)
else:
patches, pairs = None, None
if patches is None and pairs is None:
logger.info("Extracting patches ...")
patches = steps.generate_patches(scaled_imgs, constants)
logger.info("Generating pairs of patches ...")
pairs = steps.generate_pairs(patches, constants)
logger.info("Removing duplicates ...")
pairs = steps.remove_duplicates(pairs)
logger.info("Saving patches and pairs ...")
save(args.input, patches, pairs)
else:
logger.info("Using saved patches and pairs ...")
logger.info("Filtering pairs of patches and estimating local airlight ...")
pairs = steps.filter_pairs(patches, pairs, constants)
pairs2 = steps.remove_overlaps(pairs, constants)
# k = 547
# p1 = pairs2[k]
# stats = []
# for i, p2 in enumerate(pairs2):
# if i == k:
# continue
# if p2.first == p1.first:
# stats.append(p2.second)
# elif p2.second == p1.first:
# stats.append(p2.first)
# sum1 = np.zeros((len(pairs), 3))
# for i, pair in enumerate(pairs2):
# sum1[i] = pair.airlight
# print np.mean(sum1, axis=0)
logger.info("Removing outliers ...")
pairs2 = steps.remove_outliers(pairs2, constants)
logger.info("Estimating global airlight ...")
airlight = steps.estimate_airlight(pairs)
logger.info("Estimatied airlight is ...")