def test_show_image_text_np(self):
text = "apple"
fig, axis = plt.subplots(1, 1, figsize=(6.4, 4.8))
fe.util.show_image(text, fig=fig, axis=axis)
obj1 = fig_to_rgb_array(fig)
obj2 = self.text_img_ans
self.assertTrue(check_img_similar(obj1, obj2))
def test_show_image_title_np(self):
img = np.ones((150, 150), dtype=np.uint8) * 255
fig, axis = plt.subplots(1, 1, figsize=(6.4, 4.8))
fe.util.show_image(img, fig=fig, axis=axis, title="test title")
obj1 = fig_to_rgb_array(fig)
obj2 = self.title_img_ans
self.assertTrue(check_img_similar(obj1, obj2))
def test_tf_output(self):
cutmix = CutMixBatch(inputs='x', outputs=['x', 'lam'])
cutmix.beta = MockBetaDistribution('tf')
cutmix.uniform = MockUniformDistribution('tf')
mixed_images = cutmix.forward(data=self.tf_input, state={})
images = mixed_images[0].numpy()
lam = mixed_images[1].numpy()
with self.subTest('First mixed image'):
self.assertTrue(
check_img_similar(images[0],
img_to_rgb_array(self.cutmix_output1)))
with self.subTest('Second mixed image'):
self.assertTrue(
check_img_similar(images[1],
img_to_rgb_array(self.cutmix_output2)))
with self.subTest('lambda value'):
self.assertEqual(round(float(lam), 2), 0.65)
def test_show_image_height_width_np(self):
img = np.zeros((150, 100))
fig, axis = plt.subplots(1, 1, figsize=(6.4, 4.8))
fe.util.show_image(img, fig=fig, axis=axis)
obj1 = fig_to_rgb_array(fig)
obj2 = self.hw_ratio_img_ans
self.assertTrue(check_img_similar(obj1, obj2))
def test_saliency(self):
fe.estimator.enable_deterministic(200)
label_mapping = {
'airplane': 0,
'automobile': 1,
'bird': 2,
'cat': 3,
'deer': 4,
'dog': 5,
'frog': 6,
'horse': 7,
'ship': 8,
'truck': 9
}
batch_size = 32
train_data, eval_data = cifar10.load_data()
pipeline = fe.Pipeline(test_data=train_data,
batch_size=batch_size,
ops=[Normalize(inputs="x", outputs="x")],
num_process=0)
weight_path = os.path.abspath(
os.path.join(__file__, "..", "resources", "lenet_cifar10_tf.h5"))
model = fe.build(model_fn=lambda: LeNet(input_shape=(32, 32, 3)),
optimizer_fn="adam",
weights_path=weight_path)
network = fe.Network(
ops=[ModelOp(model=model, inputs="x", outputs="y_pred")])
save_dir = tempfile.mkdtemp()
traces = [
Saliency(model=model,
model_inputs="x",
class_key="y",
model_outputs="y_pred",
samples=5,
label_mapping=label_mapping),
ImageSaver(inputs="saliency", save_dir=save_dir)
]
estimator = fe.Estimator(pipeline=pipeline,
network=network,
epochs=5,
traces=traces,
log_steps=1000)
estimator.test()
ans_img_path = os.path.abspath(
os.path.join(__file__, "..", "resources", "saliency_figure.png"))
ans_img = img_to_rgb_array(ans_img_path)
output_img_path = os.path.join(save_dir, "saliency_test_epoch_5.png")
output_img = img_to_rgb_array(output_img_path)
self.assertTrue(check_img_similar(output_img, ans_img))
def test_show_image_check_float_0_to_1_np(self):
img = np.zeros((256, 256, 3), dtype=np.float32)
for x in range(256):
img[x, :, :] = x / 255
fig, axis = plt.subplots(1, 1, figsize=(6.4, 4.8))
fe.util.show_image(img, fig=fig, axis=axis)
obj1 = fig_to_rgb_array(fig)
obj2 = self.float_img_ans
self.assertTrue(check_img_similar(obj1, obj2))
def test_show_image_color_arbitrary_range_np(self):
img = np.zeros((256, 256, 3), dtype=np.float32)
for x in range(256):
img[x, :, :] = x * 0.2
fig, axis = plt.subplots(1, 1)
fe.util.show_image(img, fig=fig, axis=axis)
obj1 = fig_to_rgb_array(fig)
obj2 = self.float_img_ans
self.assertTrue(check_img_similar(obj1, obj2))
def test_show_image_color_tf(self):
img = np.zeros((90, 90, 3), dtype=np.uint8)
img[:, 0:30, :] = np.array([255, 0, 0])
img[:, 30:60, :] = np.array([0, 255, 0])
img[:, 60:90, :] = np.array([0, 0, 255])
img = tf.convert_to_tensor(img)
fig, axis = plt.subplots(1, 1, figsize=(6.4, 4.8))
fe.util.show_image(img, fig=fig, axis=axis)
obj1 = fig_to_rgb_array(fig)
obj2 = self.color_img_ans
self.assertTrue(check_img_similar(obj1, obj2))
def test_show_image_mixed_figure_layer_np(self):
bg_img = np.ones((150, 150, 3), dtype=np.uint8) * 255
boxes = np.array([[0, 0, 10, 20], [10, 20, 30, 50], [40, 70, 200,
200]])
fig, axis = plt.subplots(1, 1, figsize=(6.4, 4.8))
fe.util.show_image(bg_img, fig=fig, axis=axis)
fe.util.show_image(boxes, fig=fig, axis=axis)
fe.util.show_image("apple", fig=fig, axis=axis)
obj1 = fig_to_rgb_array(fig)
obj2 = self.mixed_img_ans
self.assertTrue(check_img_similar(obj1, obj2))
def test_show_image_color_torch(self):
img = np.zeros((90, 90, 3), dtype=np.uint8)
img[:, 0:30, :] = np.array([255, 0, 0])
img[:, 30:60, :] = np.array([0, 255, 0])
img[:, 60:90, :] = np.array([0, 0, 255])
img = torch.from_numpy(img.transpose((2, 0, 1)))
fig, axis = plt.subplots(1, 1)
fe.util.show_image(img, fig=fig, axis=axis)
obj1 = fig_to_rgb_array(fig)
obj2 = self.color_img_ans
self.assertTrue(check_img_similar(obj1, obj2))
def test_show_image_bounding_box_np(self):
bg_img = np.zeros((150, 150))
boxes = np.array([[0, 0, 10, 20, "apple"], [10, 20, 30, 50, "dog"],
[40, 70, 200, 200, "cat"],
[0, 0, 0, 0, "shouldn't shown"],
[0, 0, -50, -30, "shouldn't shown2"]])
fig, axis = plt.subplots(1, 1, figsize=(6.4, 4.8))
fe.util.show_image(bg_img, fig=fig, axis=axis)
fe.util.show_image(boxes, fig=fig, axis=axis)
obj1 = fig_to_rgb_array(fig)
obj2 = self.bb_img_ans
self.assertTrue(check_img_similar(obj1, obj2))
def test_show_image_color_np(self):
img = np.zeros((90, 90, 3), dtype=np.uint8)
img[:, 0:30, :] = np.array([255, 0, 0])
img[:, 30:60, :] = np.array([0, 255, 0])
img[:, 60:90, :] = np.array([0, 0, 255])
fig, axis = plt.subplots(1, 1)
fe.util.show_image(img, fig=fig, axis=axis)
# Now we can save it to a numpy array.
obj1 = fig_to_rgb_array(fig)
obj2 = self.color_img_ans
self.assertTrue(check_img_similar(obj1, obj2))
def test_paint_numpy(self):
output_test = self.img_data.paint_numpy()
output_test = np.squeeze(output_test, axis=0)
output = img_to_rgb_array(self.output_img)
self.assertTrue(check_img_similar(output, output_test))
def test_paint_figure(self):
fig = self.img_data.paint_figure()
output = img_to_rgb_array(self.output_img)
output_test = fig_to_rgb_array(fig)
self.assertTrue(check_img_similar(output, output_test))
