def _convert_data(self, imgs):
""" this contains operations on images that make image become input data (cannot be shown) """
# convert from RGB to CIE-LAB color space
print "RGB to LAB conversion"
imgs = [imnormalize(im) for im in imgs]
# roll axis to put channel axis to the first
print "Roll channel axis"
imgs = [im.transpose((2, 0, 1)) for im in imgs]
# flatten images
print "Flatten images"
imgs = np.asarray(imgs)
imgs = imflatten(imgs)
# normalize to be zero mean and unit std
print "Zero-mean & unit std normalization"
imgs = normalize01(imgs)
return imgs
def convert2pkl(self, pklfile):
if not os.path.isfile(pklfile):
dataset_dir = "/home/rzhao/Projects/deep-saliency/data/"
thus10000 = dataset_dir + "THUS10000_Imgs_GT/Imgs"
msra5000 = dataset_dir + "MSRA5000/Imgs"
msra5000_test = dataset_dir + "MSRA5000/MSRA-B-test"
img_ext = ".jpg"
msk_ext = ".png"
augX = 10
trn_img = []
trn_msk = []
for single_image in sorted(glob(thus10000 + "/*" + img_ext)):
rsb = glob(msra5000_test + "/*_" + ntpath.basename(single_image)[:-4] + "_smap" + msk_ext)
if len(rsb) == 0:
trn_img.append(single_image)
trn_msk.append(single_image[:-4] + msk_ext)
tst_img = []
tst_msk = []
for single_image in sorted(glob(msra5000_test + "/*" + msk_ext)):
tst_img.append(msra5000 + "/" + ntpath.basename(single_image)[: -len("_smap.png")] + img_ext)
tst_msk.append(msra5000 + "/" + ntpath.basename(single_image)[: -len("_smap.png")] + msk_ext)
# read images
print "reading ..."
train_img = [imread(fname) for fname in trn_img]
train_msk = [imread(fname) for fname in trn_msk]
test_img = [imread(fname) for fname in tst_img]
test_msk = [imread(fname) for fname in tst_msk]
# preprocessing
print "preprocessing ..."
train_x, train_y = self.preprocessing(train_img, train_msk, augx=augX)
test_x, test_y = self.preprocessing(test_img, test_msk, augx=0)
# shuffle training data
print "shuffle data ..."
np.random.seed(123)
np.random.shuffle(train_x)
np.random.seed(123)
np.random.shuffle(train_y)
# flattern and dtype conversion
print "flatten data ..."
train_x = np.asarray(train_x, dtype=np.float32)
train_y = np.asarray(train_y, dtype=np.float32)
test_x = np.asarray(test_x, dtype=np.float32)
test_y = np.asarray(test_y, dtype=np.float32)
train_x = imflatten(train_x)
train_y = imflatten(train_y)
test_x = imflatten(test_x)
test_y = imflatten(test_y)
# normalize data to have zero mean and unit std
train_x = normalize01(train_x)
test_x = normalize01(test_x)
# split into train and valid
nValid = np.int(len(train_img) * 0.1) * augX
train = [train_x[:-nValid], train_y[:-nValid]]
valid = [train_x[-nValid:], train_y[-nValid:]]
# train = [train_x[0:7000], train_y[0:7000]]
# valid = [train_x[7000:], train_y[7000:]]
test = [test_x, test_y]
data = [train, valid, test]
self.save(data, pklfile)
else:
print "History pickle file exists!"
