def anomaly_detection(test_img, g=None, d=None):
model = anogan.anomaly_detector(g=g, d=d)
ano_score, similar_img = anogan.compute_anomaly_score(model, test_img.reshape(1, 28, 28, 1), iterations=500, d=d)
# anomaly area, 255 normalization
np_residual = test_img.reshape(28,28,1) - similar_img.reshape(28,28,1)
np_residual = (np_residual + 2)/4
np_residual = (255*np_residual).astype(np.uint8)
original_x = (test_img.reshape(28,28,1)*127.5+127.5).astype(np.uint8)
similar_x = (similar_img.reshape(28,28,1)*127.5+127.5).astype(np.uint8)
original_x_color = cv2.cvtColor(original_x, cv2.COLOR_GRAY2BGR)
residual_color = cv2.applyColorMap(np_residual, cv2.COLORMAP_JET)
show = cv2.addWeighted(original_x_color, 0.3, residual_color, 0.7, 0.)
return ano_score, original_x, similar_x, show
idx_normal = np.any(y_train[..., None] == np.array([0])[None, ...], axis=1)
X_train = X_train[idx_normal]
perm_test = np.random.permutation(len(X_train))
X_train = X_train[perm_test]
## 3. other class anomaly detection
test_imgs = X_train[0:500]
score_for_2 = []
# test_img = np.random.uniform(-1,1, (28,28,1))
model = anogan.anomaly_detector(g=None, d=None)
for i in range(np.shape(test_imgs)[0]):
# start = cv2.getTickCount()
score = anogan.compute_anomaly_score(model,
test_imgs[i].reshape(1, 32, 32, 3),
iterations=500,
d=None)
# time = (cv2.getTickCount() - start) / cv2.getTickFrequency() * 1000
print('%d label, %d : done' % (1, i), '%.2f' % score)
score_for_2.append(score)
score_for_2 = np.sort(score_for_2)[::-1]
print(score_for_2)
import pickle
pickle.dump(score_for_2, open("cifar10_score.pickle", "wb"))
### compute outlier scores on CIFAR-100
(X_train, y_train), (X_test, y_test) = cifar100.load_data()
X_train = (X_train.astype(np.float32) - 127.5) / 127.5
X_test = (X_test.astype(np.float32) - 127.5) / 127.5
# X_train = X_train[:,:,:,None]