:param output: sigmoid output [512,512]
:param lamda: weights of heatmap
:param mode: mode=0 heatmap+RGB image
mode =1 heatmap
:return:
'''
heatmap = np.uint8(255 * output)
heatmap = cv2.applyColorMap(heatmap, cv2.COLORMAP_JET)
superimposed_img = heatmap * lamda + gt
superimposed_img = superimposed_img / (255 * (1 + lamda)) * 255
superimposed_img = np.uint8(superimposed_img)
if (mode == 0):
cv2.imshow('heatmap', superimposed_img)
if (mode == 1):
cv2.imshow('heatmap', heatmap)
cv2.waitKey(0)
return output
if __name__ == '__main__':
dataSet = DataSet()
v3 = Deeplabv3(input_shape=(512, 512, 1), classes=1, activation='sigmoid')
v3.load_weights('v3_std20.h5')
dice_list = []
name_list = []
RGB_image = dataSet.get_test_RGB_image()
now_image, now_gt, now_name = dataSet.get_test_data_obo(True)
now_gt = now_gt.reshape([512, 512])
now_output = v3.predict(now_image)
now_output = now_output.reshape([512, 512])
get_heat_map(RGB_image, now_output, lamda=0.4, mode=0)
from Unet import Unet
import utils
import numpy as np
from DataSet import DataSet
import config
if __name__ == '__main__':
dataSet = DataSet()
uNet = Unet.build_model()
uNet.load_weights('unet.h5')
dice_list = []
for x in range(0, config.test_num):
now_image, now_gt = dataSet.get_test_data_obo()
now_output = np.round(uNet.predict(now_image).reshape([-1]))
dice = utils.get_dice(now_gt, now_output)
dice_list.append(dice)
dice_list = np.array(dice_list)
dice = np.mean(dice_list)
print('dice :%.4f' % dice)
