I’ve just shared a new project: "Rooberry Pi"
In the area I live, there are often kangaroos that jump in front of traffic, often from blind spots, and can get themselves into harm’s way. I wanted to build a low-cost solution that can alert drivers to kangaroos in these blind spots to reduce the chance of a collision.
The system works by:
A Raspberry Pi with a camera, monitoring an area
Using Tensorflow Lite the Pi camera images are processed and the inference engine gives a confidence level a kangaroo is present in the frame
If there is a sighting, save the photo to cloud storage and send a POST request to a server
Read more here!
Hi Ben,
What a cool project, and so useful! I will be interested to watch for future iterations of this project - really has the potential to be used in the wider community.
Hi Ben,
Quite a project, well done.
Hows the Pi handling the weather? any problems with overheating in a sealed enclosure?
Also can you give me apart number for the DC socket?
Thanks
Dave
Hi Dave,
It looks a bit like this one: Panel Mount 2.1mm DC barrel jack | Adafruit ADA610 | Core Electronics Australia
All of the different systems working together, so sick!
Great idea and application.
What would be great would be to recognise Tasmanian devils who are known to head onto the road for roadkill and then become roadkill themselves.
Hi all,
Thanks for the feedback -
In terms of the panel mount connector used I picked it up from a local jaycar, unsure if core-electronics stock these:
In terms of weather, yeah its been a scorcher summer in Perth with temperatures above 40 for days. I noticed telemetry dropping out and I presume it has been too hot in the enclosure since its mostly air tight. I also underclocked the Rasperry Pi to 800 MHz as the default 1.2 GHz was getting quite hot too.
The next iteration will need to address these concerns, probably need a larger solar panel, bigger battery, and a ventilated enclosure.
And yeah in terms of identifying other objects, the default tensor model I used had kangaroos from the get-go, I wonder if it also has tasmanian devils, or other Australian wildlife. I was only triggering on kangaroos but I suppose I could trigger on any of these labels. I believe this is the model:
Thanks all, Ben
Great project! I love the idea of this.
Also I forgot to mention - I would consider using something with hardware accelerated TPU cores in the next iteration too, so all the heavy lifting is offloaded and the scan rate could be much higher.
I had my eye on this Dev Board | Coral which has onboard TPU and could probably scan for objects just as fast as the camera can feed it photos.
I think there are USB based versions of these or similar that plug into RPi.
A couple of suggestions: Log the RPi SoC (CPU) to confirm that temperature is the problem. It would also be worth logging the temperature and humidity inside the enclosure.
The problem with ventilated enclosure is the humidity can become very high and cause corrosion. There is also the problem with insects invading and damaging the boards. An alternative might be to put a heat sink that is exposed on the inside and the outside, but sealed where it passes through the enclosure.
I would suggest using UV stabilised plastic over the as much of the enclosure as is possible. A source of this material is the trays from many frozen meals. One problem with this material is that it doesn’t easily adhere to things.
Also ensure the enclosure is shaded. Typically outdoor enclosures have a top cover, then either an air gap or insulation.
It doesn’t appear that you are using any external IR lights. I would have thought this would have been needed.
Thanks again for sharing this project and hope it can be put to good use. Having been a volunteer wildlife rescuer in the past, I know how much help this could be.
Hi Craig,
Thanks for the feedback. I was looking for a sealed enclosure with an external heatsink but unable to find anything within a reasonable price - have you used any good ones before?
And correct I didn’t use the IR lights that came with the camera as the battery life outside of direct sunlight was quite lacking due to the small battery. That will also be a next revision thing and very important due to the activity of kangaroos mostly around dusk.
How is this project going? Did you manage to work out the thermal issues?
I’d be keen to see if any govt. departments would be willing to put effort into this to help reduce the wildlife death toll.
Howdy, project hasn’t really progressed. I did end up purchasing a Coral dev board but haven’t really looked into it since, such is the way of all hobby projects i guess