Hi, would like to understand if possible once the objects recognized by your AI camera (such as OpenMV cam H7 plus), can I use like Arduino board via Bluetooth to save the result what objects detected to an iOS app made by myself? Try to understand if any guidance that I can use output from the camera board for above purpose.
Hi, you want to use the OpenMV Cam RT1062 with a Bluetooth radio onboard. It can directly talk to your phone app. Using the H7 Plus with an Arduino will be more work.
As for what you can recognize, we recently upgraded our TensorFlow Lite library. As such, we can run pretty much any model. However, you’ll want to stick to models generated by EdgeImpulse right now on our current units if you want any FPS greater than 1. Desktop models take seconds to execute right now.
We will have future camera models coming out that will be much faster, but the current line of products lack NPUs. They can still do quite a bit from just RAW CPU performance, but something like YOLO will not run in real-time right now. We can do image classification and run EdgeImpule’s FOMO model, which does object localization in real-time, though.
Hi Kwagyeman, thanks for the reply. Very helpful. I’m working on some very early prototyping so latency is not the No 1 concern just now. Just try to make something work in the first place.
Can you give me a bit more guidance on this? Let’s say i use Xcode build an app in iPhone (or if iOS doesn’t work I can switch to Ardiod). At high level, how that work to connect the OpenMV Cam RT1062 object recognition and send to object list to the phone app (via bluetooth)? Do I need any API? Any example I can follow to start from? Try quickly look into the OpenMV doc can’t find anything with direct guidance on this.
Thanks.
Hi, our bluetooth support is based on MicroPython’s standard Bluetooth library. As such, any MicroPython Bluetooth tutorial should work.
E.g: Raspberry Pi Pico W and Bluetooth in Micropython. As easy as 123! - Coffeebreakpoint
We use the same radio as the Pi Pico W. As for the aioble library. That’s built into the firmware already. No need to install it.
If you want to demo Bluetooth, see the ble blinky example, and then use Nordic’s Bleutooth scanner app. This will let you find the OpenMV Cam, connect to it, and then you can write the LED’s state as an int from the app to turn the LED off or on. Note that the app will let you send strings/ints/bools, select the int datatype when sending a command from it.