def profile(model_path, model_type):
cfg = dk.load_config('config.py')
model_path = os.path.expanduser(model_path)
model = dk.utils.get_model_by_type(model_type, cfg)
model.load(model_path)
count, h, w, ch = 1, cfg.TARGET_H, cfg.TARGET_W, cfg.TARGET_D
seq_len = 0
if "rnn" in model_type or "3d" in model_type:
seq_len = cfg.SEQUENCE_LENGTH
#generate random array in the right shape
img = np.random.rand(int(h), int(w), int(ch)).astype(np.uint8)
if seq_len:
img_arr = []
for i in range(seq_len):
img_arr.append(img)
img_arr = np.array(img_arr)
#make a timer obj
timer = FPSTimer()
try:
while True:
'''
run forward pass on model
'''
if seq_len:
model.run(img_arr)
else:
model.run(img)
'''
keep track of iterations and give feed back on iter/sec
'''
timer.on_frame()
except KeyboardInterrupt:
pass
def main():
parser = argparse.ArgumentParser()
parser.add_argument('--model',
help='File path of Tflite model.',
required=True)
parser.add_argument('--image',
help='File path of the image to be recognized.',
required=True)
args = parser.parse_args()
# Initialize engine.
engine = InferenceEngine(args.model)
# Run inference.
img = Image.open(args.image)
result = engine.Inference(np.array(img))
print("inference result", result)
timer = FPSTimer()
while True:
engine.Inference(np.array(img))
timer.on_frame()
def profile(model_path, model_type):
cfg = dk.load_config('config.py')
model_path = os.path.expanduser(model_path)
model = dk.utils.get_model_by_type(model_type, cfg)
model.load(model_path)
h, w, ch = cfg.TARGET_H, cfg.TARGET_W, cfg.TARGET_D
# generate random array in the right shape in [0,1)
img = np.random.randint(0, 255, size=(h, w, ch))
# make a timer obj
timer = FPSTimer()
try:
while True:
# run inferencing
model.run(img)
# time
timer.on_frame()
except KeyboardInterrupt:
pass
in_model = os.path.expanduser(args['--model'])
# Load TFLite model and allocate tensors.
interpreter = tf.lite.Interpreter(model_path=in_model)
interpreter.allocate_tensors()
# Get input and output tensors.
input_details = interpreter.get_input_details()
output_details = interpreter.get_output_details()
# Test model on random input data.
input_shape = input_details[0]['shape']
input_data = np.array(np.random.random_sample(input_shape), dtype=np.float32)
interpreter.set_tensor(input_details[0]['index'], input_data)
interpreter.invoke()
#sample output
for tensor in output_details:
output_data = interpreter.get_tensor(tensor['index'])
print(output_data)
#run in a loop to test performance.
print("test performance: hit CTRL+C to break")
timer = FPSTimer()
while True:
interpreter.set_tensor(input_details[0]['index'], input_data)
interpreter.invoke()
timer.on_frame()