def run_predict(self):
#load the trained model from .hdf5 file
model = load_model(
r"C:\Users\10536\Desktop\MSc_Project\py\Unet\my_test\my_unet.hdf5")
model.summary()
pack = np.zeros((self.frames, self.col, self.row, 1))
#model.predict requires the input in a format of [frames,size,size,1]
for i in range(self.frames):
#resize each frame to 512*512
img_h = figure_processer.hist_eq(self.fin_sq[:, :, i] * 255)
#enhance the picture
pack[i, :, :, 0] = img_h / (255)
pre_y = model.predict(pack, batch_size=1, verbose=1)
return pre_y
def save_data(file_name,path):
fin = read_bin.read_bin2(file_name,path)
row = fin.shape[0]
col = fin.shape[1]
frames = fin.shape[2]
for i in range(frames):
label = fin[:,:,i]
img_h = figure_processer.hist_eq(label*256)
pic_new = image.new("L", (row, col))
for j in range(row):
for k in range(col):
pic_new.putpixel((j,k), int((img_h[j][k]+1)))
new_img = pic_new.resize((512,512),image.BILINEAR)
new_img = new_img.transpose(image.ROTATE_270)
new_img = new_img.transpose(image.FLIP_LEFT_RIGHT)
new_img.save(r"C:\Users\10536\Desktop\MSc_Project\py\data_set\%s__%d.jpg"%(file_name,i), "JPEG")
print("save %d"%(i))
def mean_pro(file_name, path):
fin = read_bin.read_bin2(file_name, path)
scan70 = fin[:, 70, :].astype(np.float32)
img_h = figure_processer.hist_eq(scan70 * 255)
img_gamma = figure_processer.gamma(img_h, 0.8)
img_bright = figure_processer.bright(img_gamma, 2.5, 0 / 255)
img_m = figure_processer.m_filter(img_bright.astype(np.float32))
img_hp = figure_processer.highpass_filter(img_m * 255) / 255
rate = figure_processer.get_rate1(img_hp)
np.savetxt(r"%s_rate_1.txt" % (path), rate.astype(np.int), fmt='%d')
rate_f = rate_processer.rate_proc(rate)
np.savetxt(r"%s_rate_f.txt" % (path), rate_f.astype(np.int), fmt='%d')
rate_2 = Kmeans_pro.Kmeans_pro(img_h, 5)
plt.figure()
plt.title('%s' % (file_name))
plt.imshow(rate_2)
plt.show()
rate2 = rate_processer_2.get_rate2(rate_2)
np.savetxt(r"%s_rate_2.txt" % (path), rate2.astype(np.int), fmt='%d')
print("%s done" % (file_name))
return rate, rate_f, rate2, img_h
def save_data_masked(file_name,path):
fin = read_bin.read_bin2(file_name,path)
row = fin.shape[0]
col = fin.shape[1]
frames = fin.shape[2]
z = 0
sel = np.linspace(120,frames-140,5)
sel = sel.astype(np.int)
for i in sel:
z = z+1
label = fin[:,:,i]
img_h = figure_processer.hist_eq(label*256)
img_mask = Kmeans_pro.Kmeans_pro(img_h,3)
img_m = figure_processer.m_filter(img_mask.astype(np.float32))
pic_new = image.new("L", (row, col))
for j in range(row):
for k in range(col):
pic_new.putpixel((j,k), int((img_m[j][k]*255/3+1)))
new_img = pic_new.resize((512,512),image.BILINEAR)
new_img = new_img.transpose(image.ROTATE_270)
new_img = new_img.transpose(image.FLIP_LEFT_RIGHT)
new_img.save(r"C:\Users\10536\Desktop\MSc_Project\py\data_Kmeans\%s__%d_mask.jpg"%(file_name,z), "JPEG")
print("save %d"%(i))
import numpy as np import read_bin import figure_processer import rate_processer import Kmeans_pro import rate_processer_2 import matplotlib.pyplot as plt import get_file path = r"C:\Users\10536\Desktop\MSc_Project\files\example_dia_motion_B" file_name = "example_dia_motion_B" fin = read_bin.read_bin2(file_name, path) scan70 = fin[:, 45, :].astype(np.float32) img_crop = figure_processer.crop(scan70) img_h = figure_processer.hist_eq(img_crop * 255) ####img_gamma = figure_processer.gamma(img_h,1.2) ####img_bright = figure_processer.bright(img_gamma,1.5,0/255) ####img_m = figure_processer.m_filter(img_bright.astype(np.float32)) ####img_hp = figure_processer.highpass_filter(img_m*255)/255 ####rate = figure_processer.get_rate1(img_hp) ####rate_m = figure_processer.sigma(rate) ####np.savetxt(r"%s_rate_1.txt"%(path),rate_m.astype(np.int),fmt='%d') #rate_1 = figure_processer.get_rate1(img_gamma) rate_1_1 = figure_processer.get_rate1(img_h) rate_1_1_m = figure_processer.sigma(rate_1_1) #return the cropped value back rate_1_1_m = rate_1_1_m + 0.4 * scan70.shape[0] np.savetxt(r"%s_rate_2.txt" % (path), rate_1_1_m.astype(np.int), fmt='%d') rate_1_2 = figure_processer.get_rate1(img_crop)