Inference time very high

[AkkiSony]

Hello Everyone,

I have converted yolov3 model into tflite and now, when I run the script, there are two inferences as an output on the console.

Net forward-pass time: 1677.3169040679932 ms.
Box computation time: 0.9748935699462891 ms.

Does "Net forward-pass time" include the time of loading the model as well? If yes, I would like to know if there is a way, where I can measure inference after the model is loaded into the memory.

Because, my inference without plugging in USB accelerator for Yolov3 on PC was around 350ms.