def fun(x): return to_scalar(np.rot90(x)) d_fun = lambda x : to_scalar(grad(fun)(x))
def fun(x): return to_scalar(np.rot90(x)) d_fun = lambda x : to_scalar(grad(fun)(x))
def fun(x):
return np.rot90(x)
def fun(x):
return to_scalar(np.rot90(x))
def callback_Test(mvable_pts, iter, g):
# img, vesselImage, vesselPts = updateVesselImg(vesselImage, vesselSecretion, mvable_pts, img)
# Loss
ax_loss.cla()
ax_loss.set_title('Train Loss')
ax_loss.set_xlabel('t')
ax_loss.set_ylabel('loss')
nowLoss = fitness(mvable_pts, iter)
all_loss.append(nowLoss)
time = np.arange(0, len(all_loss), 1)
ax_loss.plot(time, all_loss, '-', linestyle='solid',
label='loss') #, color = colors[i]
ax_loss.set_xlim(time.min(), time.max())
ax_loss.legend(loc="upper left")
ax_img.cla()
ax_img.set_title('Flow Image')
ax_img.tick_params(axis='x',
which='both',
bottom=False,
top=False,
labelbottom=False)
ax_img.tick_params(axis='y',
which='both',
left=False,
right=False,
labelleft=False)
imgplot = ax_img.imshow(np.rot90(np.array(img)))
ax_diffused_img.cla()
diffused_img_plt1 = diffusion(mvable_pts, vesselPts, img, vesselImage)
ax_diffused_img.set_title('Diffused Image')
ax_diffused_img.tick_params(axis='x',
which='both',
bottom=False,
top=False,
labelbottom=False)
ax_diffused_img.tick_params(axis='y',
which='both',
left=False,
right=False,
labelleft=False)
diffusedplot = ax_diffused_img.imshow(np.rot90(diffused_img_plt1))
ax_loss_map.cla()
loss_map = create_loss_map(np.array(diffused_img_plt1), vesselPts,
threshold, iter)
ax_loss_map.set_title('Diffusion Loss Map')
ax_loss_map.tick_params(axis='x',
which='both',
bottom=False,
top=False,
labelbottom=False)
ax_loss_map.tick_params(axis='y',
which='both',
left=False,
right=False,
labelleft=False)
loss_plot = ax_loss_map.imshow(np.rot90(np.array(loss_map)))
plt.draw()
saveImageOne(iter)
plt.pause(0.001)
return 3
img = np.zeros((imageSize, imageSize))
vesselImage = np.zeros((imageSize, imageSize)) * vesselSecretion
# movablePts = [(imageSize/2, imageSize/2.0), (imageSize/2 + 0.5, imageSize - 0.5)]
# movablePts = [(2.01,2.5)]
movablePts = [(2.0, 2.5)]
# movablePts = [(imageSize/2 + 0.5, imageSize - 5.5)]
# movablePts = [(imageSize/2.0, imageSize/2.0)]
vesselPts = [(int(x), int(y)) for (x, y) in movablePts]
for (xi, yi) in vesselPts:
vesselImage[xi, yi] = vesselSecretion
print('Starting Initial Diffusion')
img = initial_diffusion(movablePts, vesselPts, img, vesselImage)
print('Ended Initial Diffusion')
plt.imshow(np.rot90(np.array(img)[1:, 1:]))
plt.show()
# os.sys.exit()
def fitness(movablePts, iter):
# vesselPts, img, vesselImage are obtained through dynamic binding
diffused_img = diffusion(movablePts, vesselPts, img, vesselImage)
return loss_health(diffused_img, vesselPts, threshold, iter)
# Setup display figures
fig = plt.figure(figsize=(16, 4), facecolor='white')
ax_loss = fig.add_subplot(151, frameon=True)
ax_node_graph = fig.add_subplot(152, frameon=True)
ax_img = fig.add_subplot(153, frameon=True)
def conv2(x, y, mode='same'):
return np.rot90(convolve2d(np.rot90(x, 2), np.rot90(y, 2), mode=mode), 2)