def _create_dataset():
"""Creates a very simple dataset from within this directory
by taking some training and testing sets, normalizes them
and crops them.
"""
# Import some sample images from within the directory
x_train = skio.ImageCollection(
"tests/example_dataset/images_prepped_train/*.png").concatenate()
y_train = skio.ImageCollection(
"tests/example_dataset/annotations_prepped_train/*.png"
).concatenate()
x_test = skio.ImageCollection(
"tests/example_dataset/images_prepped_test/*.png").concatenate()
y_test = skio.ImageCollection(
"tests/example_dataset/annotations_prepped_test/*.png"
).concatenate()
# Crop the images to 256x256 and convert them to float32
x_train = x_train[:, :256, :256, :].astype(np.float32)
y_train = y_train[:, :256, :256, None].astype(np.float32)
x_test = x_test[:, :256, :256, :].astype(np.float32)
y_test = y_test[:, :256, :256, None].astype(np.float32)
# Normalize the images
x_train /= 255.0
y_train /= 255.0
x_test /= 255.0
y_test /= 255.0
return x_train, x_test, y_train, y_test
def load_defect_data():
train_path = []
train_label = []
with open("./data/train/label.csv", "r") as file:
for line in file:
image_path, image_label = line.split(',')
train_path.append(image_path)
train_label.append(image_label.strip('\n'))
test_path = []
test_label = []
with open("./data/test/label.csv", "r") as file:
for line in file:
image_path, image_label = line.split(',')
test_path.append(image_path)
test_label.append(image_label.strip('\n'))
coll_train = io.ImageCollection(train_path, load_func=convert_gray)
coll_test = io.ImageCollection(test_path, load_func=convert_gray)
coll_train = np.asarray(coll_train)
coll_test = np.asarray(coll_test)
train_label = np.asarray(train_label).astype(np.int8)
test_label = np.asarray(test_label).astype(np.int8)
X_train = coll_train
X_test = coll_test
y_train = train_label
y_test = test_label
return (X_train, y_train), (X_test, y_test)
def test_collection():
pattern = [os.path.join(data_dir, pic)
for pic in ['camera.png', 'color.png', 'multipage.tif']]
images = sio.ImageCollection(pattern[:-1])
assert len(images) == 2
assert len(images[:]) == 2
images = sio.ImageCollection(pattern)
assert len(images) == 3
assert len(images[:]) == 3
def rotate_from_dir(imagepath, savepath, rotateflag=2):
# str=imagepath+'\\*.jpg'+':'+imagepath+'\\*.png'
str = imagepath + '\\*.jpg'
if (rotateflag > 4):
if (rotateflag == 5):
coll = io.ImageCollection(str, load_func=img_rotate140180)
if (rotateflag == 6):
coll = io.ImageCollection(str, load_func=img_rotate180140)
if (rotateflag == 7):
coll = io.ImageCollection(str, load_func=img_rotate14090)
if (rotateflag == 8):
coll = io.ImageCollection(str, load_func=img_rotate9040)
else:
coll = io.ImageCollection(str, load_func=img_rotate4000)
else:
if (rotateflag == 4):
coll = io.ImageCollection(str, load_func=img_rotate90140)
if (rotateflag == 3):
coll = io.ImageCollection(str, load_func=img_rotate6090)
if (rotateflag == 1):
coll = io.ImageCollection(str, load_func=img_rotate0030)
else:
coll = io.ImageCollection(str, load_func=img_rotate3060)
# 循环保存图片
for i in range(len(coll)):
io.imsave(
savepath + '\\rs_' + np.str(rotateflag) + '_' + np.str(i) + '.jpg',
coll[i])
def marcm_save_CSVs_RGB_images_overlapping_regions_three_channel(
Bin_C0_Dir="Binaries_C0",
Bin_C1_Dir="Binaries_C1",
Bin_C2_Dir="Binaries_C2",
C2_overlap_threshold=0.3,
C1_overlap_threshold=0.5,
pixel_size=0.4,
csv_save_dir="CSVs_C0_in_C1C2",
RGB_save_dir="RGB_C0_overlapping_regions",
):
ic_C0 = io.ImageCollection(
load_pattern=os.path.join(Bin_C0_Dir, "*.tif*"), load_func=read_image_2d
)
ic_C1 = io.ImageCollection(
load_pattern=os.path.join(Bin_C1_Dir, "*.tif*"), load_func=read_image_2d
)
ic_C2 = io.ImageCollection(
load_pattern=os.path.join(Bin_C2_Dir, "*.tif*"), load_func=read_image_2d
)
condition, img_names = test_imagecollections_same_files_and_order(
ic_C0, ic_C1, ic_C2
)
if not condition:
print(
f"""{Bin_C0_Dir}, {Bin_C1_Dir}, {Bin_C2_Dir} do not contain the same number of files or
they are not in the same order."""
)
return
if not os.path.isdir(csv_save_dir):
os.mkdir(csv_save_dir)
if not os.path.isdir(RGB_save_dir):
os.mkdir(RGB_save_dir)
Dict_DFs = OrderedDict()
for C0_img, C1_img, C2_img, names in zip(ic_C0, ic_C1, ic_C2, img_names):
DF = in_region_three_channel(
C0_img=C0_img,
C1_img=C1_img,
C2_img=C2_img,
C1_overlap_threshold=C1_overlap_threshold,
C2_overlap_threshold=C2_overlap_threshold,
pixel_size=pixel_size,
)
Dict_DFs[names] = DF
DF.to_csv(os.path.join(csv_save_dir, names + ".csv"))
RGB_img = create_RGB_image_overlapping_regions(DF, C0_img_input=C0_img)
io.imsave(os.path.join(RGB_save_dir, names + ".tiff"), RGB_img)
return Dict_DFs
def _predict_on_sample(sample, network, tiramisu_model, weights):
"""Predicts values on sample, using the chosen network and weights.
Parameters
----------
network : str
Network results to compare with the gold standard.
tiramisu_model : str or None (default : None)
Tiramisu model to be used.
predict_vars : str or None (default : None)
JSON file containing the variables 'folder', 'path', 'has_goldstd',
'path_goldstd', 'segmentation_interval', and 'registered_path'. If None,
values are based on Larson et al samples.
weights : str or None (default : None)
File containing weight coefficients to be used during prediction.
Returns
-------
None
"""
print(f"# Now reading sample {sample['path']}.")
# need to check how to deal with this better.
try:
if not sample['file_ext']:
sample['file_ext'] = const.EXT_SAMPLE
except KeyError:
sample['file_ext'] = const.EXT_SAMPLE
pattern = os.path.join(sample['path'], f"*{sample['file_ext']}")
data_sample = io.ImageCollection(load_pattern=pattern)
print('# Processing...')
folder = os.path.join(utils.prediction_folder(network, tiramisu_model),
sample['folder'])
utils.process_sample(folder,
data=data_sample,
weights=weights,
network=network)
if sample['registered_path']:
pattern = os.path.join(sample['registered_path'], '*' + const.EXT_REG)
data_sample = io.ImageCollection(load_pattern=pattern)
print('# Processing registered sample...')
folder = os.path.join(utils.prediction_folder(network, tiramisu_model),
f"{sample['folder']}_REG")
utils.process_sample(folder,
data=data_sample,
weights=weights,
network=network)
return None
def valData(self):
#print(io.ImageCollection(self.val_path+self.image_folder+'/*',load_func=self.imread_convert))
image_collection = np.expand_dims(
np.array(
io.ImageCollection(self.val_path + self.image_folder + '/*',
load_func=self.imread_convert)), 3)
mask_collection = np.expand_dims(
np.array(
io.ImageCollection(self.val_path + self.mask_folder + '/*',
load_func=self.imread_convert)), 3)
#print(image_collection.shape)
print(mask_collection.shape)
val_data = self.adjustData(image_collection, mask_collection)
return val_data
def polar2_from_dir(imagepath, savepath, flag=1):
# str=imagepath+'\\*.jpg'+':'+imagepath+'\\*.png'
str = imagepath + '\\*.jpg'
if (flag == 1):
coll = io.ImageCollection(str, load_func=img_polar2_1)
if (flag == 2):
coll = io.ImageCollection(str, load_func=img_polar2_2)
if (flag == 3):
coll = io.ImageCollection(str, load_func=img_polar2_3)
if (flag == 4):
coll = io.ImageCollection(str, load_func=img_polar2_4)
if (flag == 5):
coll = io.ImageCollection(str, load_func=img_polar2_5)
if (flag == 6):
coll = io.ImageCollection(str, load_func=img_polar2_6)
if (flag == 7):
coll = io.ImageCollection(str, load_func=img_polar2_7)
else:
coll = io.ImageCollection(str, load_func=img_polar2_1)
# scipy.misc.toimage(coll[0]).save('outfile.jpg')
# #循环保存图片
for i in range(len(coll)):
io.imsave(savepath + '\\plr2_' + np.str(i) + '.jpg', coll[i])
def load_data_from_folder(self, dir):
# read all images into an image collection
ic = io.ImageCollection(dir + "*.jpg", load_func=self.imread_convert)
# create one large array of image data
data = io.concatenate_images(ic)
# extract labels from image names
labels = np.array(ic.files)
for i, f in enumerate(labels):
m = re.search("_", f)
labels[i] = f[len(dir):m.start()]
return (data, labels)
def create_dict_of_binary_masks(
input_dir="Prob_Map_C1",
thresh_method=filters.threshold_otsu,
binary_method_list=[
[morphology.binary_erosion, morphology.selem.diamond(1)],
[morphology.binary_opening, morphology.selem.diamond(3)],
[scipy_morphology.binary_fill_holes, morphology.selem.diamond(4)],
[morphology.binary_erosion, morphology.selem.diamond(1)],
],
min_area=1000,
):
ic = io.ImageCollection(
load_pattern=os.path.join(input_dir, "*.tiff"), load_func=read_prob_image
)
dict_binary_masks = OrderedDict()
for i, files in enumerate(ic.files):
thresh_value = thresh_method(ic[i])
binary_image = ic[i] > thresh_value
edited_binary_image = apply_binary_methods(
binary_image, binary_method_list=binary_method_list
)
region_list = extract_regions_from_binary(
edited_binary_image, min_area=min_area
)
dict_binary_masks[files] = create_binary_image_from_region_list(
edited_binary_image, region_list=region_list
)
return dict_binary_masks
def create_img_dict_from_folder(label_img_re=r"\w\dg\d\d?", **kwargs):
img_collection = io.ImageCollection(**kwargs)
labelled_img_dict = {
re.search(label_img_re, file)[0]: img_collection[i]
for i, file in enumerate(img_collection.files)
}
return labelled_img_dict
def loadData():
csv_path = '../data/malignancy_labels.csv'
# csv_path = '../data/test_uid_img_roi_labels.csv' #TEST
# csv_path = '../data/uid_img_roi_labels.csv'
Y_train = []
with open(csv_path, 'r') as f1:
reader = csv.reader(f1)
for row in reader:
Y_train.append(row)
f1.close()
str = "../data/Test_img/*.jpg" #TEST
# str = "../data/img_data3/*.jpg"
coll = io.ImageCollection(str)
print(len(coll))
X_train = []
for i in range(0, len(coll)):
print(i + 1)
img = coll[i]
# print(img.shape)
# jishenbianhao3 = [] # 将单通道的灰度图像，变成三通道的灰度图像，因为这样的检测效果会更好
# for x in range(3):
# jishenbianhao3.append(img)
# jishenbianhao3 = np.array(jishenbianhao3).transpose([1, 2, 0])
# img1 = jishenbianhao3
# print(img1.shape)
X_train.append(img)
X = np.array(X_train)
Y = np.array(Y_train)
# index = np.arange(400)#TEST
_X_train, _X_test, _Y_train, _Y_test = train_test_split(X, Y, test_size=0.30, random_state=42)
del X_train,Y_train
print(gc.collect())
return _X_train, _Y_train,_X_test,_Y_test
"""
def copyFile(fileDir, tarDir):
for i in range(2, 6):
imagedir = fileDir + str(i) + '/'
labeldir = docDir + str(i) + '/'
pathDir = os.listdir(imagedir)
# print(pathDir)
# for filename in pathDir:
# print(filename)
coll = io.ImageCollection(suffix)
# print(len(coll)) # 打印图片数量
num = 500
# num = int ((2*len(coll))/10)
# print(num)
sample = random.sample(pathDir, num)
# print(sample)
for name in sample:
# print(name)
# print(fileDir + name)
label = name.replace('.png', '.txt')
print(name)
print(label)
shutil.copyfile(imagedir + name, tarDir + name)
shutil.copyfile(labeldir + label, tarDocDir + label)
def test_imread_collection_single_MEF():
io.use_plugin('fits')
testfile = os.path.join(data_dir, 'multi.fits')
ic1 = io.imread_collection(testfile)
ic2 = io.ImageCollection([(testfile, 1), (testfile, 2), (testfile, 3)],
load_func=fplug.FITSFactory)
assert _same_ImageCollection(ic1, ic2)
def copyFile(fileDir, tarDir):
pathDir = os.listdir(fileDir)
# for filename in pathDir:
# print(filename)
coll = io.ImageCollection(str)
# print(len(coll)) # 打印图片数量
num = 500
# num = int ((2*len(coll))/10)
# print(num)
sample = random.sample(pathDir, num)
# print(sample)
for name in sample:
# print(name)
# print(fileDir + name)
label = name.replace('.png', '.txt')
print(name)
print(label)
shutil.copyfile(fileDir + name, tarDir + name)
shutil.copyfile(docDir + label, tarDocDir + label)
os.remove(fileDir + name)
os.remove(docDir + label)
def test6():
def convert_gray(f):
rgb = io.imread(f)
gray = color.rgb2gray(rgb)
print("fname is:%s, shape is:%s, gray shape is:%s" %
(f, rgb.shape, gray.shape))
dst = transform.resize(gray, (256, 256))
return dst
input_path = data_dir + '/*.png'
coll = io.ImageCollection(input_path, load_func=convert_gray)
output_path = "./output"
if not os.path.exists(output_path):
os.makedirs(output_path)
for i in range(len(coll)):
io.imsave(output_path + '/' + np.str(i) + '.jpg', coll[i])
print("has %d images" % len(coll))
io.imshow(coll[10])
plt.show() # miss it at the first time
# TODO: the window behave is strange
plt.figure(num='collection')
plt.subplot(1, 2, 1)
plt.title('pic 10')
plt.imshow(coll[10])
plt.subplot(1, 2, 2)
plt.title('pic 11')
plt.imshow(coll[11])
plt.show()
def Test(opts):
#model loading..
model = load_model(opts.modelPath, compile=False)
#pathnames loading for test images and labels
directory = os.path.join(opts.dataDir, opts.dataType)
fnamesX = sorted(os.listdir(os.path.join(directory, 'X')))
pathnamesX = [
os.path.join(directory, 'X', f) for f in fnamesX
if f.split('.')[-1] in opts.ext
]
# fnamesY = sorted(os.listdir(os.path.join(directory,'Y')))
# pathnamesY = [os.path.join(directory,'Y',f) for f in fnamesY if f.split('.')[-1] in opts.ext]
#loading images and model
imgsX = io.ImageCollection(load_pattern=pathnamesX)
# print(type(imgsX))
# imgsY = io.ImageCollection(load_pattern=pathnamesY)
imgsX = img_as_float(imgsX.concatenate()[:, :, :, np.newaxis])
# imgsY = img_as_float(imgsY.concatenate()[:,:,:,np.newaxis])
model = load_model(opts.modelPath, compile=False)
print(imgsX.shape)
#predicting and saving predicted images..
predY = model.predict(imgsX,
batch_size=opts.batchSize,
verbose=opts.verbosity)
CheckAndCreate(opts.outDir)
# for predY_,fnameX in izip(predY,fnamesX):
for predY_, fnameX in zip(predY, fnamesX):
io.imsave(os.path.join(opts.outDir, fnameX), predY_[:, :, 0])
return
def polar1_from_dir(imagepath, savepath):
# str=imagepath+'\\*.jpg'+':'+imagepath+'\\*.png'
str = imagepath + '\\*.jpg'
coll = io.ImageCollection(str, load_func=img_polar)
# io.imshow(coll[0])
# ok ok
# scipy.misc.toimage(coll[0], cmin=0.0, cmax=255).save('outfile.jpg')
# scipy.misc.toimage(coll[0], mode='P').save('outfile.jpg')
# scipy.misc.toimage(coll[0]).save('outfile.jpg')
# output = scipy.misc.toimage(coll[0])
# output.convert('RGBA')
# scipy.misc.toimage(output).save('outfile.jpg')
# error
# imgp = Image.fromarray(coll[0])
# imgp.save('outfile.jpeg')
# ok ok 但是灰色图
# cv2.imwrite("filename.jpg", coll[0])
# error
# matplotlib.image.imsave('name.png', coll[0])
# 循环保存图片
for i in range(len(coll)):
# imgp = Image.fromarray(coll[i])
# error
# io.imsave(savepath + '\\plr1_' + np.str(i)+'.jpg',coll[i])
img_name = savepath + '\\plr1_' + np.str(i) + '.jpg'
scipy.misc.toimage(coll[i]).save(img_name)
def save_animation(data,
_title=None,
_path=None):
class IMGLoader:
ic = data
def __call__(self, frame):
return np.hstack([self.ic[frame]['oldData'], self.ic[frame]['newData']])
img_load = IMGLoader()
frm = range(len(data))
img_collection = io.ImageCollection(frm, load_func=img_load)
fig, ax = plt.subplots()
show_img = plt.imshow(img_collection[0], cmap='gray')
plt.axis('off')
def update(i):
show_img.set_data(img_collection[i])
return show_img
anim = FuncAnimation(fig, update, frames=frm, interval=10, repeat=True)
# Подготовка директории
try:
os.stat(_path)
except FileNotFoundError:
os.makedirs(_path)
filename = os.path.join(_path, _title + '.gif')
anim.save(filename, writer='imagemagick', fps=30)
def LoadData(number):
if number == 1:
path = '/Users/zhuxiaoxiansheng/Desktop/日常/数据集/yale_faces/*.bmp'
elif number == 2:
path = '/Users/zhuxiaoxiansheng/Desktop/日常/数据集/orl_faces_full/*.pgm'
elif number == 3:
path = '/Users/zhuxiaoxiansheng/Desktop/日常/数据集/jaffe/*.tiff'
elif number == 4:
path = '/Volumes/TOSHIBA EXT/数据集/YaleB/*.pgm'
pictures = io.ImageCollection(path)
data = []
for i in range(len(pictures)):
picture = pictures[i]
picture = skimage.color.rgb2gray(picture)
data.append(np.ravel(picture.reshape((1,picture.shape[0]*picture.shape[1]))))
label = []
if number == 1:
for i in range(len(data)):
label.append(int(i/11))
elif number == 2:
for i in range(len(data)):
label.append(int(i/10))
elif number == 3:
for i in range(len(data)):
label.append(int(i/20))
elif number == 4:
label = [0]*64+[1]*64+[2]*64+[3]*64+[4]*64+[5]*64+[6]*64+[7]*64+[8]*64+[9]*64+[10]*60+[11]*59+[12]*60+[13]*63+[14]*62+[15]*63+[16]*63+[17]*64+[18]*64+[19]*64+[20]*64+[21]*64+[22]*64+[23]*64+[24]*64+[25]*64+[26]*64+[27]*64+[28]*64+[29]*64+[30]*64+[31]*64+[32]*64+[33]*64+[34]*64+[35]*64+[36]*64+[37]*64
return np.matrix(data),np.matrix(label).T
def main(_):
# file_path = "/Volumes/TOSHIBA EXT/final/try"
file_path = "D:\\final\\thumb"
out_file_name = "D:\\final\\useful_data\\"
png = file_path + '/*.png'
# xml_path = "/Volumes/TOSHIBA EXT/final/final.xlsx"
xml_path = "D:\\final\\final.xlsx"
labels_xml = xlrd.open_workbook(xml_path)
labels_table = labels_xml.sheets()[0]
defect = labels_table.col_values(3)[1:]
labels = load_label(defect)
workbook = xlwt.Workbook(encoding='ascii')
worksheet = workbook.add_sheet("sheet1")
coll = io.ImageCollection(png)
for n in range(19):
worksheet.write(0, n, labels_table.row_values(0)[n])
m = 1
for i in range(16950):
print("i=", i, "m=", m, "n=", n)
if labels[i] != 1:
image = load_photo(coll, i)
cv2.imwrite(out_file_name + str(i) + ".png", image)
for n in range(19):
worksheet.write(m, n, labels_table.row_values(i + 1)[n])
m = m + 1
workbook.save('D:\\final\\useful_data\\useful_data.xlsx')
def shear_from_dir(imagepath, savepath, shearflag=2):
# str=imagepath+'\\*.jpg'+':'+imagepath+'\\*.png'
str = imagepath + '\\*.jpg'
if (shearflag == 3):
coll = io.ImageCollection(str, load_func=img_shear43)
if (shearflag == 1):
coll = io.ImageCollection(str, load_func=img_shear41)
else:
coll = io.ImageCollection(str, load_func=img_zoom42)
# 循环保存图片
for i in range(len(coll)):
io.imsave(
savepath + '\\sr_' + np.str(shearflag) + '_' + np.str(i) + '.jpg',
coll[i])
def getImage(str):
mat=io.ImageCollection(str)
print(len(mat))
return mat
def images(path):
x = []
for i in range(len(files)):
x = x + ["%s/%s/%s" % (path, labels[i], files[i])]
#print("%s/%sResized/%s.Bmp" % (path, name, i))
x = io.ImageCollection(x)
return x
def img_read(sub_path):
str1 = path + '/*.jpg'
dataX = io.ImageCollection(str1)
train_xdata = np.array([dataX])
train_xdata = train_xdata[0]
return train_xdata
def build_moment_dataset(folder):
from sklearn import model_selection
def just_alpha(file, img_num):
img = io.imread(file)
img = np.squeeze(img[:, :, 3])
img = np.bool_(img)
return img
ic = io.ImageCollection(folder, load_func=just_alpha)
X_pos = np.zeros(shape=(len(ic), 9), dtype='float')
y_pos = np.ones(shape=(len(ic)), dtype='int')
X_neg = np.zeros_like(X_pos)
y_neg = np.zeros_like(y_pos)
for i in range(len(ic)):
#calculate the positive and negative moments
alpha = ic[i]
X_pos[i, :] = gen_features(alpha)
X_neg[i, :] = gen_features(np.logical_not(alpha))
X = np.concatenate((X_neg, X_pos))
y = np.concatenate((y_neg, y_pos))
X_train, X_test, y_train, y_test = model_selection.train_test_split(
X, y, train_size=.5, shuffle=True, stratify=y)
return X_train, X_test, y_train, y_test
def piliangzhixing():
#用于批量处理
pic_name = file_name('./tes') #待识别图像文件名字的读取
for i in range(0, len(pic_name)):
picture_name = pic_name[i]
#批量读入图像并存在coll
coll = io.ImageCollection('./tes/*.jpg')
print '图片的数量', len(coll)
for i in range(0, len(coll)):
io.imsave('./linshi/test.jpg', coll[i])
zhixing()
os.remove('./linshi/test.jpg')
#print "全局变量YanSe",YanSe
#识别结果写进去txt
fp = open('./result/test.txt', 'a+')
fp.write(picture_name[i] + ' ' + YanSe + '\n')
fp.close()
#修改结果名字txt
result_txt_name = time.strftime('%Y%m%d_%H_%M',
time.localtime(time.time()))
result_name = result_txt_name + '.txt'
os.rename('./result/test.txt', './result/' + result_name)
def load_stack(inputDir):
"""Loads .jpg files from a directory and concatenates into an image stack\
using io.ImageCollection."""
imageFiles = glob.glob(os.path.join(inputDir, '*.jpg'))
imageVolume = io.ImageCollection(imageFiles, as_grey=True).concatenate()
return imageVolume
def main():
parser = argparse.ArgumentParser(usage=__doc__)
parser.add_argument('images', nargs='*', help='Images to view.')
parser.add_argument(
'--layers',
action='store_true',
help='Treat multiple input images as layers.',
)
parser.add_argument(
'-m',
'--multichannel',
help='Treat images as RGB.',
action='store_true',
)
args = parser.parse_args()
with gui_qt():
v = Viewer()
images = io.ImageCollection(args.images, conserve_memory=False)
if args.layers:
for image in images:
v.add_image(image, multichannel=args.multichannel)
else:
if len(images) > 0:
if len(images) == 1:
image = images[0]
else:
image = np.stack(images, axis=0)
v.add_image(image, multichannel=args.multichannel)
def images(name, path):
x = []
for i in labelsInfo.ID:
x = x + ["%s/%sResized/%s.Bmp" % (path, name, i)]
#print("%s/%sResized/%s.Bmp" % (path, name, i))
x = io.ImageCollection(x)
return x
