def prereq_load_and_compute( SIFT=False):
# print("Loading feature paths")
# if SIFT==True:
# print("SIFT")
# paths = hf.load_sift_paths('test')
# else:
# print("BRISK")
# paths = hf.load_brisk_paths('test')
# print("loading features...")
# features = hf.load_features(paths)
# print(str(len(features)) + " items loaded.")
# print("Normalizing features")
# modified_feature_arr = hf.normalize_array(features, mode = 'test')
# No_Of_Test_Items = len(modified_feature_arr)
l_channel_paths = hf.load_luminance_paths('test')
print("Loading Luma")
l_channel_chromas = hf.load_luminance(l_channel_paths)
print(str(len(l_channel_chromas)) + " items loaded.")
modified_feature_arr = l_channel_chromas
No_Of_Test_Items = len(modified_feature_arr)
print("modifying the shape of input and output")
test_x = np.array(modified_feature_arr).reshape([No_Of_Test_Items, modified_feature_arr[0].shape[0], modified_feature_arr[0].shape[1], 1])
print("test_x shape: ",test_x.shape)
return test_x
def predict_and_dump(test_x, mode):
luminance_paths = hf.load_luminance_paths('test')
print("loading luminance...")
luminance = hf.load_luminance(luminance_paths)
if mode == 'a':
b_channel_paths = hf.load_b_channel_chroma_paths('test')
print("loading b channel chroma...")
b_channel_chromas = hf.load_b_channel_chroma(b_channel_paths)
# try:
model = load_a_model()
# except:
# print("Error loading model")
if mode == 'b':
a_channel_paths = hf.load_a_channel_chroma_paths('test')
print("loading a channel chroma...")
a_channel_chromas = hf.load_a_channel_chroma(a_channel_paths)
# try:
model = load_b_model()
# except:
# print("Error loading model")
predictions = model.predict(test_x)
print("Dumping predictions")
if (mode == 'a'):
hf.save_blob(predictions, 'predicted_a_chroma')
# for i in range(len(predictions)):
# a_channel_chroma = hf.scale_image(predictions[i])
# hf.reconstruct(luminance[i], a_channel_chroma,b_channel_chromas[i], i, 'A')
if (mode == 'b'):
hf.save_blob(predictions, 'predicted_b_chroma')
def prereq_load_and_compute(mode, SIFT=False):
# if SIFT==True:
# print("SIFT")
# paths = hf.load_sift_paths('train')
# else:
# print("BRISK")
# paths = hf.load_brisk_paths('train')
# print("loading features...")
# features = hf.load_features(paths)
# print(str(len(features)) + " items loaded.")
# print("Normalizing features")
# modified_feature_arr = hf.normalize_array(features)
# No_Of_Test_Items = len(modified_feature_arr)
l_channel_paths = hf.load_luminance_paths('train')
print("Loading Luma")
l_channel_chromas = hf.load_luminance(l_channel_paths)
print(str(len(l_channel_chromas)) + " items loaded.")
modified_feature_arr = l_channel_chromas
No_Of_Test_Items = len(modified_feature_arr)
if mode == 'a':
a_channel_paths = hf.load_a_channel_chroma_paths('train')
print("loading a channel chroma...")
a_channel_chromas = hf.load_a_channel_chroma(a_channel_paths)
# print(a_channel_chromas[0])
# print(a_channel_chromas[1])
print(str(len(a_channel_chromas)) + " items loaded.")
train_y_channel = np.array(a_channel_chromas).reshape(
No_Of_Test_Items, -1)
else:
b_channel_paths = hf.load_b_channel_chroma_paths('train')
print("loading b channel chroma...")
b_channel_chromas = hf.load_b_channel_chroma(b_channel_paths)
print(str(len(b_channel_chromas)) + " items loaded.")
train_y_channel = np.array(b_channel_chromas).reshape(
No_Of_Test_Items, -1)
train_y_channel = train_y_channel + 128
train_y_channel = train_y_channel / 256.0
print("modifying the shape of input and output")
train_x = np.array(modified_feature_arr).reshape([
No_Of_Test_Items, modified_feature_arr[0].shape[0],
modified_feature_arr[0].shape[1], 1
])
print("Pickling shapes")
hf.pickle_shape(train_x, train_y_channel)
print("train_x shape: ", train_x.shape)
print("train_y shape: ", train_y_channel.shape)
return train_x, train_y_channel
def main():
args = parse_arguments()
DATASET_PATH = args.path
mode = 'NONE'
if not (args.a or args.b or args.ab): #Check if only one case is true
print("ERROR: use -h for HELP")
exit()
return
# print(args)
if args.a:
print("Testing model based on a-channel")
mode = 'a'
if args.b:
print("Testing model based on b-channel")
mode = 'b'
if args.ab:
print("Testing model based on a&b-channel")
mode = 'ab'
# FOR AKHEEL
# print("processing images")
# pre_works.process_images(DATASET_PATH)
if (mode != 'ab'):
test_x = prereq_load_and_compute(SIFT=True)
predict_and_dump(test_x, mode)
return
# IF AB MODE
print("Loading a_chroma")
predictions_A = hf.load_from_pickle("predicted_a_chroma")
# print("a_channel_chroma", predictions_A[0][1])
# print("a_channel_chroma", predictions_A[1][1])
# print("a_channel_chroma", predictions_A[2][1])
print("Loading b_chroma")
predictions_B = hf.load_from_pickle("predicted_b_chroma")
luminance_paths = hf.load_luminance_paths('test')
print("loading luminance...")
luminance = hf.load_luminance(luminance_paths)
for i in range(len(predictions_A)):
a_channel_chroma = hf.scale_image(predictions_A[i])
# print("a_channel_chroma", a_channel_chroma)
b_channel_chroma = hf.scale_image(predictions_B[i])
hf.reconstruct(luminance[i], a_channel_chroma, b_channel_chroma, i,
'AB')
# to reconstruct grayscale images
empty_array = np.zeros((200, 200))
hf.reconstruct(luminance[i], empty_array, empty_array, i, 'G')
def start(mode, callbackFn):
if (mode != 'ab'):
test_x = prereq_load_and_compute(SIFT=True)
predict_and_dump(test_x, mode)
if (mode == 'a'):
print("A COMPLETED")
return callbackFn('b')
else:
print("B COMPLETED")
return callbackFn('ab')
# IF AB MODE
print("Loading a_chroma")
predictions_A = hf.load_from_pickle("predicted_a_chroma")
# print("a_channel_chroma", predictions_A[0][1])
# print("a_channel_chroma", predictions_A[1][1])
# print("a_channel_chroma", predictions_A[2][1])
print("Loading b_chroma")
predictions_B = hf.load_from_pickle("predicted_b_chroma")
luminance_paths = hf.load_luminance_paths('test')
print("loading luminance...")
luminance = hf.load_luminance(luminance_paths)
empty_array = np.zeros((200, 200))
b_channel_paths = hf.load_b_channel_chroma_paths('test')
print("loading b channel chroma...")
b_channel_chromas = hf.load_b_channel_chroma(b_channel_paths)
a_channel_paths = hf.load_a_channel_chroma_paths('test')
print("loading a channel chroma...")
a_channel_chromas = hf.load_a_channel_chroma(a_channel_paths)
for i in range(len(predictions_A)):
a_channel_chroma = hf.scale_image(predictions_A[i])
# print("a_channel_chroma", a_channel_chroma)
b_channel_chroma = hf.scale_image(predictions_B[i])
hf.reconstruct(luminance[i], a_channel_chroma, b_channel_chroma, i,
'AB')
# to reconstruct grayscale images
hf.reconstruct(luminance[i], empty_array, empty_array, i, 'G')
hf.reconstruct(luminance[i], a_channel_chromas[i],
b_channel_chromas[i], i, 'GT')
return callbackFn(mode)
def predict_and_dump(test_x, mode):
luminance_paths = hf.load_luminance_paths('test')
print("loading luminance...")
luminance = hf.load_luminance(luminance_paths)
if mode=='a':
b_channel_paths = hf.load_b_channel_chroma_paths('test')
print("loading b channel chroma...")
b_channel_chromas = hf.load_b_channel_chroma(b_channel_paths)
# try:
model = load_a_model()
# except:
# print("Error loading model")
if mode=='b':
a_channel_paths = hf.load_a_channel_chroma_paths('test')
print("loading a channel chroma...")
a_channel_chromas = hf.load_a_channel_chroma(a_channel_paths)
# try:
model = load_b_model()
# except:
# print("Error loading model")
predictions = model.predict(test_x)
print("Dumping predictions")
empty_array = np.zeros((200,200))
if(mode == 'a'):
hf.save_blob(predictions, 'predicted_a_chroma')
for i in range(len(predictions)):
a_channel_chroma = hf.scale_image(predictions[i])
print("a_channel_chroma", a_channel_chroma)
a_channel_chroma = a_channel_chroma*20
hf.reconstruct(empty_array, a_channel_chroma,empty_array, i, 'A')
if(mode == 'b'):
hf.save_blob(predictions, 'predicted_b_chroma')
for i in range(len(predictions)):
b_channel_chroma = hf.scale_image(predictions[i])
b_channel_chroma = b_channel_chroma*20
hf.reconstruct(empty_array, empty_array,b_channel_chroma, i, 'B')
