def graph_fn():
(y_grid, x_grid) = ta_utils.image_shape_to_grids(height=3, width=5)
y_coordinates = tf.constant([1.5, 0.5], dtype=tf.float32)
x_coordinates = tf.constant([2.5, 4.5], dtype=tf.float32)
sigma = tf.constant([0.1, 0.5], dtype=tf.float32)
channel_onehot = tf.constant([[1, 0, 0], [0, 1, 0]], dtype=tf.float32)
channel_weights = tf.constant([1, 1], dtype=tf.float32)
heatmap = ta_utils.coordinates_to_heatmap(y_grid, x_grid, y_coordinates,
x_coordinates, sigma,
channel_onehot, channel_weights)
return heatmap
def test_coordinates_to_heatmap(self):
(y_grid, x_grid) = ta_utils.image_shape_to_grids(height=3, width=5)
y_coordinates = tf.constant([1.5, 0.5], dtype=tf.float32)
x_coordinates = tf.constant([2.5, 4.5], dtype=tf.float32)
sigma = tf.constant([0.1, 0.5], dtype=tf.float32)
channel_onehot = tf.constant([[1, 0, 0], [0, 1, 0]], dtype=tf.float32)
channel_weights = tf.constant([1, 1], dtype=tf.float32)
heatmap = ta_utils.coordinates_to_heatmap(y_grid, x_grid, y_coordinates,
x_coordinates, sigma,
channel_onehot, channel_weights)
# Peak at (1, 2) for the first class.
self.assertAlmostEqual(1.0, heatmap.numpy()[1, 2, 0])
# Peak at (0, 4) for the second class.
self.assertAlmostEqual(1.0, heatmap.numpy()[0, 4, 1])
