def test_inverted(neural_net, testing_file, render=False):
test_pattern = image.load_binary_image(testing_file)["data"]
inverted_pattern = invert_pattern(test_pattern)
image.render_image(inverted_pattern, neural_net.rows, neural_net.cols)
input("Press enter to continue...")
# Evaluate
refreshed = neural_net.evaluate_net(inverted_pattern, 'async', 20,render=render)
image.render_image(refreshed, neural_net.rows, neural_net.cols)
input("Press enter to continue...")
def refresh_synchronic(self, s, render):
refreshed = [sgn(x) for x in np.dot(self.W, s)]
# TODO: ver eso de render image
if render:
im, bitmap = image.render_image(refreshed, self.rows, self.cols)
return np.array(refreshed)
def test_combination(neural_net, testing_set, render=False):
log.info("Testing patterns...")
L = len(testing_set)
# Combine patterns
combined_pattern = image.load_binary_image(testing_set[0])["data"]
for i in range(1,L):
next_pattern = image.load_binary_image(testing_set[i])["data"]
# Load test pattern
combined_pattern = sum_patterns(combined_pattern, next_pattern)
image.render_image(combined_pattern, neural_net.rows, neural_net.cols)
input("Press enter to continue...")
# Evaluate
refreshed = neural_net.evaluate_net(combined_pattern, 'async', 20,render=render)
image.render_image(refreshed, neural_net.rows, neural_net.cols)
input("Press enter to continue...")
def test(neural_net, testing_set, render=False):
log.info("Testing patterns...")
for test in testing_set:
# Load test pattern
test_pattern = image.load_binary_image(test)["data"]
# Add noise to test pattern
test_pattern_noisy = add_noise(test_pattern, 0.25)
image.render_image(test_pattern_noisy, neural_net.rows,
neural_net.cols)
input("Press enter to continue...")
refreshed = neural_net.evaluate_net(test_pattern_noisy,
'async',
20,
render=render)
image.render_image(refreshed, neural_net.rows, neural_net.cols)
input("Press enter to continue...")
def refresh_asynchronic(self, s, render): # Asynchronic
if render:
im, bitmap = image.render_image(s, self.rows, self.cols)
refreshed_s = np.copy(s)
for i in (np.random.permutation(self.N)):
refreshed_s[i] = sgn(sum(self.W[i, :] * refreshed_s))
if render and refreshed_s[i] != s[i]:
im, bitmap = image.render_pixel(im, bitmap, refreshed_s[i],
i % self.cols,
int(np.floor(i / self.cols)))
return np.array(refreshed_s)
def plot():
output = render_image()
return Response(output.getvalue(), mimetype='image/png')
def draw(self):
render_image(self.title, self.im, self.fps, False)