Vehicle Detection

Hi all,

What is the most accurate method of detecting a moving vehicle and using the draw_rectangle() function to draw a rectangle around the vehicle?

The OpenMV camera will be mounted onto the back of a car and is required to accurately draw a rectangle around any moving vehicle behind it.

Thanks in advance.

I don;t think our product can really do that. Vehicle detection requires really powerful CNN support which we don’t have quite yet.

Mostly out of curiosity, I wonder if the recent work on adding CNNs would actually make this possible.

Yes and no, the camera can now detect if a car is in an ROI or not but there’s no network architecture yet to track multiple cars at once. Edge Impulse will create these soon.

Hi,
we also need to detect cars, but we only need to know, whether there is a car within a given ROI or not.
So currently, we do not have to track multiple cars.

I went through the examples but I am not sure what would be the best approach here.
Could you give me some hint which example could be used, maybe after some modification ?

Kind Regards, Jan

Hi Jstoefer,

You need to collect data of cars and then train a CNN using Edge Impulse to do this. Our examples cover standard CV stuff. Not CNN’s directly which require data to be collected to train the CNN.

1 Like