How to accelerate the person detction model on the STM32F7456G_DICOVERY board

I see that the tflite library has support for CMSIS_NN , but i want to know how i can leverage it for my model.

so far im able to run the person detection model at 0.2 FPS on the STM32F7 using tflite libraries and inovoking the model. Any suggestion to improve the FPS?

