def test_prediction(app, last_feature_dim):
file = os.path.join(os.path.dirname(__file__), "fixtures", "elephant.jpg")
img = keras.preprocessing.image.load_img(file)
img = keras.preprocessing.image.img_to_array(img)
img = lqz.preprocess_input(img)
model = app(weights="imagenet")
preds = model.predict(np.expand_dims(img, axis=0))
# Test correct label is in top 3 (weak correctness test).
names = [p[1] for p in lqz.decode_predictions(preds, top=3)[0]]
assert "African_elephant" in names
def preprocess(data):
return lqz.preprocess_input(data["image"])
def test_image(request):
file = os.path.join(os.path.dirname(__file__), "fixtures", "elephant.jpg")
img = keras.preprocessing.image.load_img(file)
img = keras.preprocessing.image.img_to_array(img)
img = lqz.preprocess_input(img)
return np.expand_dims(img, axis=0)
def preprocess(data):
img = lqz.preprocess_input(data["image"])
label = tf.one_hot(data["label"], 1000)
return img, label
def test_numpy_input():
image = np.random.randint(0, 255, size=(300, 300, 3), dtype="uint8")
prepro = preprocess_input(image)
assert isinstance(prepro, np.ndarray)
def test_wrong_input():
with pytest.raises(ValueError, match="Input must be of size .*"):
preprocess_input(np.random.randint(0, 255, size=(4, 32, 32, 3), dtype="uint8"))
def test_tensor_input():
image = np.random.randint(0, 255, size=(300, 300, 3), dtype="uint8")
tf_image = tf.constant(image)
prepro = preprocess_input(tf_image)
assert isinstance(prepro, tf.Tensor)