def test_generate_embedding_data_with_bottom_k_boxes(self):
saved_model_path = self._export_saved_model()
top_k_embedding_count = 0
bottom_k_embedding_count = 2
inference_fn = generate_embedding_data.GenerateEmbeddingDataFn(
saved_model_path, top_k_embedding_count, bottom_k_embedding_count)
inference_fn.start_bundle()
generated_example = self._create_tf_example()
self.assertAllEqual(
tf.train.Example.FromString(generated_example).features
.feature['image/object/class/label'].int64_list.value, [5])
self.assertAllEqual(
tf.train.Example.FromString(generated_example).features
.feature['image/object/class/text'].bytes_list.value, ['hyena'])
output = inference_fn.process(generated_example)
output_example = output[0]
self.assert_expected_example(output_example, botk=True)
def test_generate_embedding_data_fn(self):
saved_model_path = self._export_saved_model()
top_k_embedding_count = 1
bottom_k_embedding_count = 0
inference_fn = generate_embedding_data.GenerateEmbeddingDataFn(
saved_model_path, top_k_embedding_count, bottom_k_embedding_count)
inference_fn.setup()
generated_example = self._create_tf_example()
self.assertAllEqual(tf.train.Example.FromString(
generated_example).features.feature['image/object/class/label']
.int64_list.value, [5])
self.assertAllEqual(tf.train.Example.FromString(
generated_example).features.feature['image/object/class/text']
.bytes_list.value, [b'hyena'])
output = inference_fn.process(('dummy_key', generated_example))
output_example = output[0][1]
self.assert_expected_example(output_example)
