OpenVINO+NCS2/NCS+MutiModel(FaceDetection, EmotionRecognition)+MultiStick+MultiProcess+MultiThread+USB Cam/PiCam. RaspberryPi3

[Hyodo__Katsuya]

Hello everyone.

I tried implementing Multi-model asynchronous sample of Face Detection + Emotion Detection in Python. (+ MultiStick)

Movie playback and reasoning are executed completely asynchronously, so if you move the camera quickly, the frame will shift greatly.

The MultiStick function is being verified for stability improvement.

https://github.com/PINTO0309/OpenVINO-EmotionRecognition.git

[Hyodo__Katsuya]

Corei7 + NCS2 x2

[Truong__Dien_Hoa]

Hi @Katsuya. Very interesting project. I'm thinking about how to use async and multiprocess to increase the FPS and will definitely take a look on how you do it. How many batches you use in your model ? I was thinking might be increase number of batches can help but for resnet50, t(batch=4) is very close to 4*t(batch).

[Hyodo__Katsuya]

Hi, Truong, Dien Hoa.

> How many batches you use in your model ?

It's x1 batch.

↓

btw, x1 batch + x4 request + x2 models + x2 NCS2, In parallel.