def openImageDialog(self):
filename = QFileDialog.getOpenFileName(
self, "Open file", None, 'Image Files (*.png *.jpg *.bmp)')
if filename[0]:
# Read the image
self.imageSrc = image_read(filename[0])
# Compute the Max-Tree
(self.maxtree_p,
self.maxtree_s) = maxtree_union_find_level_compression(
self.imageSrc, connection8=True)
self.imageSrcFlat = self.imageSrc.flatten()
self.maxtree_a = compute_attribute_area(self.maxtree_s,
self.maxtree_p,
self.imageSrcFlat)
# Update slider and spinbox in the image resolution range
self.areaThresholdSlider.setRange(
0, (self.imageSrc.shape[0] * self.imageSrc.shape[1]))
self.areaThresholdSpinbox.setRange(
0, (self.imageSrc.shape[0] * self.imageSrc.shape[1]))
# Update the image label
self.updateImages()
def cam(self):
img_size = self.flag.image_size
batch_size = self.flag.batch_size
epochs = self.flag.total_epoch
config = tf.ConfigProto()
config.gpu_options.allow_growth = True
set_session(tf.Session(config=config))
with open(os.path.join(self.flag.ckpt_dir, self.flag.ckpt_name, 'model.json'), 'r') as json_file:
loaded_model_json = json_file.read()
#model = model_from_json(loaded_model_json)
model = models.vgg_like(self.flag)
weight_list = sorted(glob(os.path.join(self.flag.ckpt_dir, self.flag.ckpt_name, 'weight*')))
model.load_weights(weight_list[-1])
print '[*] model load : %s'%weight_list[-1]
label_list = [os.path.basename(path) for path
in sorted(glob(os.path.join(self.flag.data_path, '*')))]
dict_name_list = dict()
for name in label_list:
dict_name_list[name] = [path for path
in sorted(glob(os.path.join(self.flag.data_path, name, '*')))]
dict_name_list[name] = dict_name_list[name][:50]
data_list = []
[data_list.append(image_read(name, color_mode=1, target_size=128))
for name in dict_name_list['dog']]
# print len(data_list)
np_Inference_data_list = np.array(data_list)[:,0,:,:,:]
np_original_data_list = np.array(data_list)[:,1,:,:,:].astype(np.uint8)
# print np_Inference_data_list.shape
result = model.predict(np_Inference_data_list, self.flag.batch_size)
# print result.shape
prediction_labels = np.argmax(result, axis=1)
classmap = get_classmap_keras(self.flag, model, np_Inference_data_list, prediction_labels)
assert classmap.shape[0] == np_original_data_list.shape[0] == prediction_labels.shape[0]
cv2.namedWindow("show", 0)
cv2.resizeWindow("show", 500, 500)
for idx in range(classmap.shape[0]):
predicted_label = prediction_labels[idx]
print "[*] %d's label : %s"%(idx, label_list[predicted_label])
img_original = np_original_data_list[idx,:,:,:]
img_classmap = classmap[idx,:,:,predicted_label]
color_classmap = cv2.applyColorMap(img_classmap, cv2.COLORMAP_JET)
img_show = cv2.addWeighted(img_original, 0.8, color_classmap, 0.5, 0.)
# cv2.imshow("show", img_show)
# cv2.imwrite("./result/dog_%d.png"%idx, img_show)
# if cv2.waitKey(1) == 27:
# break
print "[*] done"
def openImageDialog(self):
filename = QFileDialog.getOpenFileName(
self, "Open file", None, 'Image Files (*.png *.jpg *.bmp)')
if filename[0]:
# Read the image
self.imageSrc = image_read(filename[0])
self.maxtree = maxtree(self.imageSrc)
# Update slider and spinbox in the image resolution range
self.areaThresholdSlider.setRange(
0, (self.imageSrc.shape[0] * self.imageSrc.shape[1]))
self.areaThresholdSpinbox.setRange(
0, (self.imageSrc.shape[0] * self.imageSrc.shape[1]))
# Update the image label
self.updateImages()
import sys
import time
import matplotlib.pyplot as plt
import numpy as np
from utils import image_read
from maxtree import maxtree_berger, maxtree_berger_rank, maxtree_union_find_level_compression, compute_attribute_area, direct_filter
file_path = sys.argv[1]
area_filter = int(sys.argv[2])
img1 = image_read(filename=file_path)
print("Resolution: {}".format(img1.shape[0] * img1.shape[1]))
print("{}x{}".format(img1.shape[0], img1.shape[1]))
flatten_image = img1.flatten()
# Original image
fig, (fig1, fig2, fig3, fig4) = plt.subplots(1, 4)
fig1.set_axis_off()
fig2.set_axis_off()
fig3.set_axis_off()
fig4.set_axis_off()
fig1.set_title("Original")
fig1.imshow(img1, cmap="gray")
# Algorithms
start = time.time()
(parents, s) = maxtree_berger(img1, connection8=True)
end = time.time()
attr = compute_attribute_area(s, parents, flatten_image)
out1 = direct_filter(s, parents, flatten_image, attr, area_filter)
import tree_of_shape
"""
I. Max-Tree
"""
#image_input = image_read("examples/images/circuit_small_small.png")
#image_output = max_tree.area_filter(image_input, 100)
#plt.imshow(image_input, cmap="gray")
#plt.show()
#plt.imshow(image_output, cmap="gray")
#plt.show()
"""
II. Tree of shapes
"""
image_input = image_read("examples/images/circuit_small_small.png")
plt.imshow(image_input, cmap="gray")
plt.show()
image_output = tree_of_shape.area_filter(image_input, 5)
plt.imshow(image_output, cmap="gray")
plt.show()
image_output = tree_of_shape.area_filter(image_input, 20)
plt.imshow(image_output, cmap="gray")
plt.show()
"""
III. Binary partition tree
"""