def test_inference_init_yes_tflite_model_no_edgemodel():
model = {
"tflite": _good_tflite_model(),
}
tf_engine = TFInferenceEngine(model=model, labels=_good_labels())
assert tf_engine
assert tf_engine._tf_interpreter
def test_inference_init_no_tflite_model_yes_edgemodel():
model = {
'edgetpu':
'some_edgetpu.tflite'
}
with pytest.raises(KeyError):
TFInferenceEngine(model=model, labels='no_data/no_labels.txt')
def test_inference_init_yes_models_yes_labels():
model = {
"tflite": _good_tflite_model(),
"edgetpu": _good_edgetpu_model(),
}
tf_engine = TFInferenceEngine(model=model, labels=_good_labels())
assert tf_engine
assert tf_engine._model_tflite_path == _good_tflite_model()
assert tf_engine._model_edgetpu_path == _good_edgetpu_model()
def test_inference_init_other_params():
model = {
"tflite": _good_tflite_model(),
"edgetpu": _good_edgetpu_model(),
}
tf_engine = TFInferenceEngine(
model=model,
labels=_good_labels(),
confidence_threshold=0.876,
top_k=678,
)
assert tf_engine
assert tf_engine._model_tflite_path == _good_tflite_model()
assert tf_engine._model_edgetpu_path == _good_edgetpu_model()
assert tf_engine.confidence_threshold == 0.876
assert tf_engine.top_k == 678
assert tf_engine.is_quantized
assert tf_engine._model_labels_path == _good_labels()
def test_inference_init_no_params():
with pytest.raises(AssertionError):
TFInferenceEngine()
def test_inference_init_no_tflite_model_yes_edgemodel():
model = {"edgetpu": "some_edgetpu.tflite"}
with pytest.raises(KeyError):
TFInferenceEngine(model=model, labels="no_data/no_labels.txt")
def test_inference_init_no_model_yes_labels():
with pytest.raises(AssertionError):
TFInferenceEngine(model=None, labels="no_data/no_labels.txt")