def augument(data_path, label, image_name, save_path, size=224, training = True):
image_path = os.path.join(data_path, image_name)
(name, extension) = splitfilename(image_name)
extension = extension.lower()
if extension not in IMG_EXTS:
print('filered image: %s' % image_name)
return
try:
img = image.imdecode(open(image_path, 'rb').read()).astype('float32')
except Exception as ex:
print("error: ", ex)
return
if label is not None:
label_path = os.path.join(save_path, label)
else:
label_path = save_path
mkdir(label_path)
if training:
aug1 = image.HorizontalFlipAug(0.5)
aug2 = image.HorizontalFlipAug(.5)
img = image.resize_short(img, size=384, interp=2)
center_crop, _ = image.center_crop(img, size=(size, size))
new_name = "%s_%s%s" % (name, "0", extension)
cv.imwrite(os.path.join(label_path, new_name), center_crop.asnumpy())
random_crop, _ = image.random_crop(img, size=(size, size))
new_name = "%s_%s%s" % (name, "1", extension)
cv.imwrite(os.path.join(label_path, new_name), random_crop.asnumpy())
random_crop, _ = image.random_crop(img, size=(size, size))
new_name = "%s_%s%s" % (name, "2", extension)
cv.imwrite(os.path.join(label_path, new_name), random_crop.asnumpy())
random_crop, _ = image.random_crop(img, size=(size, size))
new_name = "%s_%s%s" % (name, "3", extension)
cv.imwrite(os.path.join(label_path, new_name), random_crop.asnumpy())
img_aug1 = aug1(random_crop).clip(0,255)
new_name = "%s_%s%s" % (name, "4", extension)
cv.imwrite(os.path.join(label_path, new_name), img_aug1.asnumpy())
img_aug2 = aug2(center_crop).clip(0, 255)
new_name = "%s_%s%s" % (name, "5", extension)
cv.imwrite(os.path.join(label_path, new_name), img_aug2.asnumpy())
img_resize = image.imresize(img, w=size, h=size, interp=2)
new_name = "%s_%s%s" % (name, "6", extension)
cv.imwrite(os.path.join(label_path, new_name), img_resize.asnumpy())
else:
img = image.resize_short(img, size=size)
img, _ = image.center_crop(img, size=(size, size))
new_name = "%s%s" % (name, extension)
cv.imwrite(os.path.join(label_path, new_name), img.asnumpy())
def main():
net = models.resnet50_v2(pretrained=True)
url = 'http://data.mxnet.io/models/imagenet/synset.txt'
fname = download(url)
with open(fname, 'r') as f:
text_labels = [' '.join(l.split()[1:]) for l in f]
url2 = 'https://upload.wikimedia.org/wikipedia/commons/thumb/b/b5/\
Golden_Retriever_medium-to-light-coat.jpg/\
365px-Golden_Retriever_medium-to-light-coat.jpg'
fname2 = download(url2)
x = image.imread(fname2)
x = image.resize_short(x, 256)
x, _ = image.center_crop(x, (224, 224))
plt.imshow(x.asnumpy())
plt.show()
prob = net(transform(x)).softmax()
idx = prob.topk(k=5)[0]
for i in idx:
i = int(i.asscalar())
print('With prob = %.5f, it contains %s' %
(prob[0, i].asscalar(), text_labels[i]))
def predict_with_models_from_gluon_model_zoo_example():
# Gluon model zoo provides multiple pre-trained powerful models.
# We can download and load a pre-trained ResNet-50 V2 model that was trained on the ImageNet dataset.
net = models.resnet50_v2(pretrained=True)
# Download and load the text labels for each class.
url = 'http://data.mxnet.io/models/imagenet/synset.txt'
fname = download(url)
with open(fname, 'r') as f:
text_labels = [' '.join(l.split()[1:]) for l in f]
# Randomly pick a dog image from Wikipedia as a test image, download and read it.
url = 'https://upload.wikimedia.org/wikipedia/commons/thumb/b/b5/Golden_Retriever_medium-to-light-coat.jpg/365px-Golden_Retriever_medium-to-light-coat.jpg'
fname = download(url)
x = image.imread(fname)
# Use the image processing functions provided in the MXNet image module.
x = image.resize_short(x, 256)
x, _ = image.center_crop(x, (224, 224))
plt.imshow(x.asnumpy())
plt.show()
def transform(data):
data = data.transpose((2, 0, 1)).expand_dims(axis=0)
rgb_mean = nd.array([0.485, 0.456, 0.406]).reshape((1, 3, 1, 1))
rgb_std = nd.array([0.229, 0.224, 0.225]).reshape((1, 3, 1, 1))
return (data.astype('float32') / 255 - rgb_mean) / rgb_std
prob = net(transform(x)).softmax()
idx = prob.topk(k=5)[0]
for i in idx:
i = int(i.asscalar())
print('With prob = %.5f, it contains %s' %
(prob[0, i].asscalar(), text_labels[i]))
def transform_val(data, label):
im = data.astype('float32') / 255
im = image.resize_short(im, 256)
im, _ = image.center_crop(im, (224, 224))
im = nd.transpose(im, (2,0,1))
im = mx.nd.image.normalize(im, mean=(0.485, 0.456, 0.406), std=(0.229, 0.224, 0.225))
return (im, nd.array([label]).asscalar())
def transform_val(data, label):
im = data.astype('float32') / 255
im = image.resize_short(im, 256) #对数据按照短边进行crop为256*256
im, _ = image.center_crop(im, (224, 224)) #对数据进行中心裁剪为224*224
im = nd.transpose(im, (2,0,1)) #
im = mx.nd.image.normalize(im, mean=(0.485, 0.456, 0.406), std=(0.229, 0.224, 0.225))#归一化操作
return (im, nd.array([label]).asscalar()) #返回图像和标签
def __getitem__(self, idx):
data, label = self.paths[idx]
img_d = self._imread(data)
if self._annot:
img_an = self._imread(label, flag=0)
img_an = img_an.astype('float32')
img_an = img_an.transpose(axes=(2, 0, 1))
img_an = img_an.reshape(1, 1, 360, 480)
img_an = img_an.pad(mode="constant",
constant_value=0.0,
pad_width=(0, 0, 0, 0, 12, 12, 0, 0))
img_an = img_an.reshape(1, 384, 480)
img_an = img_an.transpose(axes=(1, 2, 0))
img_an, _ = image.center_crop(img_an, (472, 376))
img_an = img_an.transpose(axes=(2, 0, 1))
imag_d = self._transform(img_d)
imag_d = imag_d.astype('float32')
imag_d = imag_d.reshape(1, 3, 360, 480)
imag_d = imag_d.pad(mode="constant",
constant_value=0.0,
pad_width=(0, 0, 0, 0, 12, 12, 0, 0))
imag_d = imag_d.reshape(3, 384, 480)
if self._annot:
return imag_d, img_an
return imag_d
def self_designed_transform(x, args):
if args.CenterCrop is True:
x, _ = center_crop(x, args.CenterCropSize)
x = two_type_factory[args.sumbol_or_ndarray].image.to_tensor(x)
if args.Normalize is True:
x = two_type_factory[args.sumbol_or_ndarray].image.normalize(
x, mean=args.mean, std=args.std)
return x
def transform_image(img_path):
img = image.imread(img_path)
data = image.resize_short(img, 256)
data, _ = image.center_crop(data, (224, 224))
data = data.transpose((2, 0, 1)).expand_dims(axis=0)
rgb_mean = nd.array([0.485, 0.456, 0.406]).reshape((1, 3, 1, 1))
rgb_std = nd.array([0.229, 0.224, 0.225]).reshape((1, 3, 1, 1))
data = (data.astype("float32") / 255 - rgb_mean) / rgb_std
return data
def load_image(img_path, long_side_length):
x = image.imread(img_path)
x = image.resize_short(x, long_side_length)
x, _ = image.center_crop(x, (448, 448))
x = x.astype('float32')
x = x / 255
x = image.color_normalize(x,
mean=nd.array([0.485, 0.456, 0.406]),
std=nd.array([0.229, 0.224, 0.225]))
x = x.reshape((1, 3, 448, 448))
return x
def classify(image_file_path):
x = image.imread(image_file_path)
x = image.resize_short(x, 256)
x, _ = image.center_crop(x, (224, 224))
plt.imshow(x.asnumpy())
plt.show()
prob = net(transform(x)).softmax()
idx = prob.topk(k=5)[0]
print(' prob | name')
print(' ------------------')
for i in idx:
i = int(i.asscalar())
print(' %.3f | %s' % (prob[0, i].asscalar(), text_labels[i]))
def transform_filter(data, label):
"""
用于图片/标签前处理的函数。
Transform function for image/label pre-processing.
:param data: image data, mx.ndarray
:param label: image label, mx.ndarray
"""
im = filter_image(data, filter_level) # compress image to certain bit
im = im.astype("float32") / 255
im = image.resize_short(im, RESIZE_SIZE)
im, _ = image.center_crop(im, (INPUT_SIZE, INPUT_SIZE))
im = nd.transpose(im, (2, 0, 1))
im = mx.nd.image.normalize(im,
mean=(0.485, 0.456, 0.406),
std=(0.229, 0.224, 0.225))
return (im, nd.array([label]).asscalar())
from mxnet.gluon.utils import download
from mxnet import image
net = models.resnet50_v2(pretrained=True)
url = 'http://data.mxnet.io/models/imagenet/synset.txt'
fname = download(url)
with open(fname, 'r') as f:
text_labels = [' '.join(l.split()[1]) for l in f]
url = 'https://upload.wikimedia.org/wikipedia/commons/thumb/b/b5/Golden_Retriever_medium-to-light-coat.jpg/365px-Golden_Retriever_medium-to-light-coat.jpg'
fname = download(url)
x = image.imread(fname)
x = image.resize_short(x, 256)
x, _ = image.center_crop(x, (224, 224))
plt.imshow(x.asnumpy())
plt.show()
def transform(data):
data = data.transpose((2, 0, 1)).expand_dims(axis=0)
rgb_mean = nd.array([0.485, 0.456, 0.406]).reshape((1, 3, 1, 1))
rgb_std = nd.array([0.229, 0.224, 0.225]).reshape((1, 3, 1, 1))
return (data.astype('float32') / 255 - rgb_mean) / rgb_std
prob = net(transform(x)).softmax()
idx = prob.topk(k=5)[0]
for i in idx:
i = int(i.asscalar())
def load_vgg16_image(img_path):
x = image.imread(img_path)
x = image.resize_short(x, 256)
x, _ = image.center_crop(x, (224, 224))
return x
def load_vgg16_image(img_path, image_width=224, image_height=224):
x = image.imread(img_path)
x = image.resize_short(x, 256)
x, _ = image.center_crop(x, (image_width, image_height))
return x
