Deploying Face Detection model with Yocto Hardknott using imx8mp-lpddr4-evk

[ejmfrancisco]

Hi! I am new to building images with Yocto. My goal is to deploy a real-time face detection model in the imx8mp-lpddr4-evk board with Yocto Hardknott image.

In my code, the following are in the requirements.text file.

torch==1.12.1

torchvision==0.13.1

numpy==1.21.2

opencv-contrib-python==4.6.0.66

tensorboardX==1.8

matplotlib==3.5.3

scipy==1.7.3

pillow==8.4.0

My concern is, do I have to add this in the local.conf file manually or I can install it later on when the image is already built.

Also, if I have to add it in local.conf file, how do I add those? What specific syntax will I have to append in the local.conf?

[Zhiming_Liu]

You shouldn't change these version in Yocto.

There is no need to sync the PC development environment and Yocto development environment.

You can refer this document to deploy your model, you need export torch model to onnx/tflite.

https://www.nxp.com/docs/en/user-guide/IMX-MACHINE-LEARNING-UG.pdf