ADAS Driver Drowsiness detection using i.MX RT1170

[Sharathembdes]

Hi,

 We have the following requirements 

1) Vehicle driver's face detection or if driver is absent

2) To count the number of driver's eye blinks

3) To check the driver's mouth opening (Yawn Detection)

4) To check the driver's head position (Head Tilt) 

For the above requirements is the MX RT1170 is sufficient  to detect drowsiness ?
If it's not then please suggest suitable part number.

Thanks & Regards,

Sharath

[david_piskula]

Hello Sharath,

it's difficult to say whether your usecase is viable to run on RT1170 or not. This is always very model dependent.

Attributes that decide whether or not you can use a given device:

1. Complexity of the model - more complex models require more computation and more memory space. This can be a limiting factor when you need fast inference, low power consumption or when the model doesn't fit inside the available memories on your device.
2. Input size - similar to the previous point, the size of input (amount of pixels in your case) also affects both computation speed and memory size requirements.
3. Inference time - when your model fits in your memory and successfully runs on the device, you also have to evaluate whether the time it takes to compute the results and the FPS you are able to run at are sufficient for your usecase or not.

It's possible that the RT1170 would be sufficient for you but I can't say for sure until an actual experiment with an actual model is done.

Other devices that could be of interest are the MCX MCU (in preproduction phase at the moment) https://www.nxp.com/products/processors-and-microcontrollers/arm-microcontrollers/general-purpose-mc... - this MCU features an NPU to accelerate neural networks or some of our application processors, mainly in the i.MX 8 family, which would offer a lot more power and memory than an MCU.

Best Regards,

David