Want to Run this Dlstreamer models (Age,Face)on Hailo Raspberry Pi

[Avinash-Gahlowt]

I had ran this models for age,gender,face on my Intel laptop using gst-launch-1.0 v4l2src device=/dev/video0 ! decodebin ! gvadetect model=~/intel/models/intel/face-detection-adas-0001/FP32/face-detection-adas-0001.xml ! gvaclassify model=~/intel/models/intel/age-gender-recognition-retail-0013/FP32/age-gender-recognition-retail-0013.xml model_proc=/opt/intel/dlstreamer/samples/gstreamer/model_proc/intel/age-gender-recognition-retail-0013.json ! gvawatermark ! xvimagesink

Now I want to use that same model and everything on Hailo 8 AI Accelerator Raspberry Pi so if i use Hailo Data Flow Compler This is used to convert Onnx Based model to Hailo compatible format which can directly be pluged in the pipeline with it post-processing and everything is good to go .
So ./models/public/aclnet/aclnet_des_53.onnx can i do something similar to that so that ill be able to use Hailonet element to run this model in HEF format.

If any other way of doing the same im open to work on that , My end goal is to run Age,gender,emotion on Hailo 8 Ai accelerator Raspberry Pi using gstreamer .

[brmarkus]

OpenVINO uses its own format, Intermediate-Representation (IR). OpenVINO supports various other formats like ONNX.
Once a model (e.g. in ONNX-format) was loaded (and compiled) it can be exported - and imported again (without compiling it again).

But I'm not aware of the other way around - converting a model in IR format into ONNX. But I might simple miss a tool.

Have you found other models in e.g. ONNX format?

E.g. (no priority, no preference, not tested, not checking license, not checking accuracy, just a few search-hots from Google):

• https://github.com/onnx/models
• https://github.com/Ali-Fayzi/deepface-onnx

I don't know the "Hailo Raspberry Pi"... you might find some models being too big, requiring to much memory, supporting too old or too new operations...