GPU acceleration in eiq Face Recogintion

[sharat2096]

Hi,

I executed a tflite based Face Recognition application on imx8mqevk with bsp 5.4.3.1 ,i could see a part of GPU coming into picture.I didn't get anything on how GPU is running even when UseNNapi(false) in the existing code base.Any inputs on running ".tflite"  model on GPU making this code base as reference??.

https://source.codeaurora.org/external/imxsupport/eiq_sample_apps/tree/examples-tflite/face_recognit...

 Sources for compilation were taken from the mentioned link.

[sharat2096]

Hi,

Hoping to hear ,any update for the concern raised.

https://community.nxp.com/t5/i-MX-Processors/GPU-acceleration-in-eiq-Face-Recogintion/m-p/1168365/hi...

[Bio_TICFSL]

Hello Sharat2096,

The face recognition is part of the eiq and openCV please check this page:

https://vanity-freescale.jiveon.com/docs/DOC-343740

Regards

[sharat2096]

Hi,

Actually my point of interest is how GPU is handling  the tflite model,I tried a similar approach but the i could see only CPU execution and nothing was running on GPU. Also interested to know  how tflite interpreter is enabling the execution on GPU for the EIQ Sample Code. Few cases of my trial gave some segmentation faults while calculating tensors following EIQ face recognition cpp application as reference .