def test_from_grayscale(self):
brisque = BRISQUE()
for idx, path in enumerate(self._paths):
image = cv2.imread(path, 0)
self.assertAlmostEqual(brisque.get_score(image),
self._score[idx],
places=3)
def test_from_path(self):
"""Test directly from the path."""
brisque = BRISQUE()
for idx, path in enumerate(self._paths):
self.assertAlmostEqual(brisque.get_score(path),
self._score[idx],
places=3)
def test_scale_feature(self):
"""Test scale the feature from Matlab original code."""
brisque = BRISQUE()
for idx, feat in enumerate(self._brisque_feats):
scaled_feat = brisque._scale_feature(feat)
np.testing.assert_array_almost_equal(scaled_feat,
self._scaled_feats[idx],
decimal=5)
def test_get_feature(self):
"""The feature generated is equal."""
brisque = BRISQUE()
for idx, path in enumerate(self._paths):
image = cv2.imread(path, 0)
feature = brisque.get_feature(image)
np.testing.assert_array_almost_equal(feature,
self._brisque_feats[idx],
decimal=0)
def test_brisque_values_grey(device: str) -> None:
img = imread('tests/assets/goldhill.gif')
x_grey = torch.tensor(img).unsqueeze(0).unsqueeze(0)
score = brisque(x_grey.to(device), reduction='none', data_range=255)
score_baseline = BRISQUE().get_score(img)
assert torch.isclose(score, torch.tensor(score_baseline).to(score), rtol=1e-3), \
f'Expected values to be equal to baseline, got {score.item()} and {score_baseline}'
def test_brisque_values_rgb(device: str) -> None:
img = imread('tests/assets/I01.BMP')
x_rgb = (torch.tensor(img, dtype=torch.float32).permute(2, 0, 1).unsqueeze(0))
score = brisque(x_rgb.to(device), reduction='none', data_range=255.)
score_baseline = BRISQUE().get_score(x_rgb[0].permute(1, 2, 0).numpy()[..., ::-1])
assert torch.isclose(score, torch.tensor(score_baseline).to(score), rtol=1e-3), \
f'Expected values to be equal to baseline, got {score.item()} and {score_baseline}'
class BRISQUE_Metric_gen(Callback):
def __init__(self):
super().__init__()
self.name = "brisque_gen"
self.brisque = BRISQUE()
def on_epoch_begin(self, **kwargs):
self.values = []
def on_batch_end(self, last_output, last_target, **kwargs):
for img in last_output:
score = self.brisque.get_score(img.permute(1, 2, 0).cpu().numpy())
self.values.append(score)
def on_epoch_end(self, last_metrics, **kwargs):
return add_metrics(last_metrics, sum(self.values) / len(self.values))
def main(args):
# get the reference to the webcam
camera = cv2.VideoCapture(0)
# ROI coordinates
img_w = 112
img_h = 112
top = 160
left = 300
bottom = top + img_h
right = left + img_w
image_num = 0
start_recording = False
# make a new folder storing images
today_str = datetime.now().strftime("%y%m%d%H%M%S")
saving_folder_path = "./" + today_str + '_' + args.file_prefix
os.mkdir(saving_folder_path)
saving_img_path = saving_folder_path + '/' + args.file_prefix + '_'
print("Save images to " + saving_img_path)
brisq = BRISQUE()
while(True):
# get the current frame
(grabbed, frame) = camera.read()
if (grabbed == True):
# fix the frame width to be 640
frame = imutils.resize(frame, width=640)
# clone the frame for rendering
clone = frame.copy()
# get the ROI
roi = frame[top:bottom, left:right]
# convert the roi to grayscale and blur it
gray = cv2.cvtColor(roi, cv2.COLOR_BGR2GRAY)
image_quality = brisq.get_score(gray)
# sometimes brisq outputs nan, we should ignore these anomolies
if math.isnan(image_quality):
continue
if start_recording and image_quality < float(args.ref_quality):
file_full_path = saving_img_path + str(image_num) + ".jpg"
cv2.imwrite(file_full_path, gray,[int(cv2.IMWRITE_JPEG_QUALITY), 100])
# make sure the image is saved
while(os.path.getsize(file_full_path) < 1):
cv2.imwrite(file_full_path, gray,[int(cv2.IMWRITE_JPEG_QUALITY), 100])
image_num += 1
print('%d images saved\r' % (image_num), end="")
# specify the font and draw the key using puttext
font = cv2.FONT_HERSHEY_SIMPLEX
cv2.putText(clone,str(int(image_quality * 10) / 10),(right - img_w, top), font, .8,(255,255,255),2,cv2.LINE_AA)
cv2.imshow("Gray", gray)
# draw the segmented hand
cv2.rectangle(clone, (right, top), (left, bottom), (0,255,0), 2)
# display the frame with segmented hand
cv2.imshow("Video Feed", clone)
# observe the keypress by the user
keypress = cv2.waitKey(1) & 0xFF
# if the user pressed "q", then stop looping
if keypress == ord("q") or image_num > int(args.img_num) - 1:
break
if keypress == ord("s"):
start_recording = not start_recording
else:
print("[Warning!] Error input, Please check your(camra Or video)")
break
# free up memory
camera.release()
cv2.destroyAllWindows()
from libsvm import svmutil # Trick to fix brisque import error in OSX
from brisque import BRISQUE
import torch
from utils import utils_image as util
from utils import utils_model
from models.network_dncnn import DnCNN as net
## Initiliaze Params ##
# Inference Device
device = torch.device('cuda' if torch.cuda.is_available() else 'cpu')
# Brisq Object
brisq = BRISQUE()
# Model zoo
model_pool = 'model_zoo'
def main(args):
# Model name
if args['model'] is None:
model_name = 'dncnn_50' # 'dncnn_25' | 'dncnn_50' | 'dncnn_gray_blind' | 'dncnn_color_blind' | 'dncnn3'
else:
model_name = args['model']
# RGB or Gray Scale mode
if args['color'] == 'rgb':
# plt.suptitle(self.supTitle[idx], fontsize="x-large")
# for j in range(6):
# if j <= 2:
# ax[0][j].imshow(result[j])
# ax[0][j].axis('off')
# ax[0][j].set_title(self.subTitle[j])
# else:
# ax[1][j - 3].imshow(result[j])
# ax[1][j - 3].axis('off')
# ax[1][j - 3].set_title(self.subTitle[j])
#
# mng = plt.get_current_fig_manager()
# mng.full_screen_toggle()
# if os.path.exists('examples/%s' % self.images[imIdx]) == False:
# os.makedirs('examples/%s' % self.images[imIdx])
# plt.savefig(os.path.join('examples/%s' % self.images[imIdx], self.supTitle[idx] + '.png'))
# figSaveIndex += 1
# plt.close()
toc = time.clock()
progress += 1
print('Progress % d / %d with time: % .4f' % (progress, self.sampleNumber[1] - self.sampleNumber[0], toc - tic))
import dicttoxml
xl = dicttoxml.dicttoxml(resultDict, attr_type=False)
f = open("xml/kadis_distortions_fixed_%d_%d.xml" % (self.sampleNumber[0], self.sampleNumber[1]), "wb")
f.write(xl)
np.save("npy/kadis_distortions_fixed_%d_%d.npy" % (self.sampleNumber[0], self.sampleNumber[1]), npyData)
np.save("brisque/kadis_distortions_brisque_%d_%d.npy" % (self.sampleNumber[0], self.sampleNumber[1]), npyBRQ)
brq = BRISQUE()
DG = DistortGenerate()
DG.image_generate()
import glob
from os.path import join
from brisque import BRISQUE
score_calculator = BRISQUE()
path = "/home/zenit/Pictures/Export"
pattern = 'P1000306*.jpg'
full_path = join(path, pattern)
matching_files = glob.glob(full_path)
def calculate_score(image_path):
score = score_calculator.get_score(image_path)
return score
file_score_tuples = map(lambda file: (file, calculate_score(file)), matching_files)
max_score_file = max(file_score_tuples, key=lambda item: item[1])
print("max score file is " + max_score_file[0])
def __init__(self):
super().__init__()
self.name = "brisque_in"
self.brisque = BRISQUE()
self.final_score = 0.
def __init__(self):
super().__init__()
self.name = "brisque_gen"
self.brisque = BRISQUE()
def test_score_from_scaled_feature(self):
"""Test score from scaled feature from Matlab original code."""
brisque = BRISQUE()
for idx, scaled_feat in enumerate(self._scaled_feats):
score = brisque._calculate_score(scaled_feat)
self.assertAlmostEqual(score, self._expected_score[idx], places=2)
