Thanks. The music met the main criteria of being free, which is what mattered to me!

Red Green show never gets old 😎

Thanks! Glad you enjoyed them.

Thank you!

Thanks Rob, glad you enjoyed it. We do a lot of community fence projects and bears are endless fascinating so that will be an ongoing feature but I am keen for any fun video project in the future. Let me know if you ever have a fun idea. 

A lot of interesting questions on fencing and we can chat for hours on this and maybe we should! We primarily approach specifics of what works for any given electric fence from the perspective of a tradesperson, rather than a scientist. What works is difficult to quantify and changes with specifics; we have thousands of hours of field work in electric fencing for bears. 

Tom Smith's paper was looking at portable fences - these are a challenge as what makes a good fence (tight wires, good grounding, high powered energizer) are all heavy. With respect to his work, I would never recommend using an aluminium tent peg as grounding nor an energizer outputting 0.1 joules for bears, even for a backpacker fence. They just are not sufficient to deliver an effective and painful shock to the hide, this is compounded by loose wires that tend to roll over fur. I am skeptical a bear would even feel a 1.1kv shock, the Bella Coola fence in that video was putting out over 12kV. Camera traps to verify bear interaction would be a valuable addition to a study like this as we have no quantifiable information about each event. Another conclusion to his study could be that while bears are curious, they tend to be cautious around human encampments and generally aversive to novel barriers, regardless of whether they recieve a strong shock. 

We rarely mark or usual visual indicators on bear fences, even standalone wire fences, unless wolves are also a target species. We do insist on a break in vegetation of a minimum of 2 meters to allow wildlife to slowly approach the fence rather than stumble onto it. This reduces all wildlife stumbling into it, but breaking fences happens for sure, most often it is ungulates. As a rule bears are cautious approaching electric fences. They rarely challenge a well-built electric fence but can break or tear down poorly constructed fences that don't deliver an adequate shock. One question I think would be cool is a movement study on how a population of bears utilize unprotected human attractants in areas where electric fences were recently introduced. 

For sampling electric fences, there are quite a few approaches solutions in the market. A few of the modern energizers have this feature built-in and there are basic lights that flash when a fence shorts out. Both of these though are aimed at alerting someone that there is a problem with their fence like a down wire. Catching an animal event is a lot harder as it is typically going to be a short-to-ground on a single pulse only. We built our own alarm for this express purpose, it samples the potential and pulse-rate and then trips on a single low pulse. We don't sell a ton of them, mostly for polar bear fences where people really want to know if a bear is challenging a fence.  If you wanted to add it to a project, I am sure the two of us could connect it to an IOT modem.

 

Hi @jared 

Thanks for all the information - really, really interesting. 

And I'd definitely be interested in hearing more about your pulse-rate trip alarm! This is exactly what we need! The project will likely start with a 'pilot' area being assessed first, probably closest to the town so it's easiest for people to go and physcially check for problems and assess alerts. The total site is a 50km+ highway that is unfortunately killing hundreds of koalas and if the pilot assessment (maybe up to a few kms from the town) was successful, I think the aim would be to fence more of the highway. The pilot assessment would involve a 3-6 month assessment of a fence system including any fault detections and potential shock event alerts. Would you guys potentially be interested in collaborating on this?

One major issue for us more generally is using electric fencing around humans and human settlements, where people or pets may be shocked. Gates like the ones you used in the video are a must I think for sites requiring pedestrian access. Regular warning signs etc., are also a requirement by law over here I believe. Another idea we are considering is a 'smart' fence that uses AI to actually switch on only when animals are detected - I am hoping to work with @kimhendrikse on this one at some point soon. 

Cheers,

Rob

 

That's a great idea Rob. Reminds me a bit of  the fences that are open to allow free travel for wild life but close for cattle. But Jared, what do you perse mean by an international collaboration? Are you wanting to put your expertise to practise in European locations or are you mean meaning joining of forces with International folks, potentially anywhere where the problem might need solving ?

Clearly, I can see plenty of opportunity for interesting experiments with some of my tech over your side of the world. I bet Rob would have a wealth of extra applicability suggestions.

Hi Kim, We already work internationally and I am more than happy to collaborate anywhere there is good idea or a problem we can help solve! Without getting too far ahead of myself, I also expect to be able to work more directly as well in both yours and Rob's areas in the future ;)

Well that sounds pretty good. In the future we could have a chat to catch up with the latest developments.

We deployed our first training run of our new polar bear model in the zoo on a Denmark in the weekend and it correctly detected a polar bear (as opposed to a brown bear which is also in the model) at greater than 130m. And that’s with a model that can process one high quality inference per second on a Raspberry Pi.

This is a detection at 130-150 distance.

but it’s actually a dynamic digital zoom from an 8MP camera with this view.

So we are pretty excited.

Hat's off to @Lars_Holst_Hansen  for all his hard work building a quality dataset and annotating thousands of images. It's paid off.

Kim

This is awesome @kimhendrikse ! Getting a polar bear at 130m! and well done @Lars_Holst_Hansen on the hard work paying off!

@jared I'd be super keen to discuss options for test fences for koalas when you have time, including the fence/shock event monitoring you mentioned - this will be vital for this pilot project I think. And combining this in the future with AI to make smart fences (let's face it, for dumb people - I include myself in this too as I'll definitely touch it!) would mean we could safely use it urban areas where we commonly have koala road strike problems...

Cheers,

Rob

 

Just here to reiterate, WOW! Hell of a job snapping that polar bear - congrats! 

Thanks man!

And what I'm happy about is that this model even runs on my Raspberry Pi version, which I can ship now to anyone who has an interest in bear early warning systems btw. It can check at a rate of once per second with a state of the art 20M parameter yolov9-gelan model :)

(Or deer, or about 19 other animals at the moment, whatever interests you I can train a model for).

And if you thought the above was good, I can even ship the device with 640x512 thermal cameras (or 1280x1024), such as this miniature one here (Actually, that's both a thermal and 8MP optical camera, it can scan both at the same time).



And they come in a variety of lenses, with the biggest one here, 100mm, I'm sure that I could reliably detect a bear at night with no starlight available at over 300m when it's pitch black outside. I estimate between 500m-1km actually, but I haven't tested that.  Now I just need someone to want to do that so we can find out :)

@kimhendrikse That is pretty darn cool. We do a bit of polar bear work each year - mostly ship around mid-summer due to sea-lift schedule for northern communities. With work I am sure we could help find a test location in Canada, I also did a shipment to Svallbard last summer for deterrents. There has been a lot of interest in polar bear detection systems in the past - I can think of a few contacts that might be interested. Will chat to @BrettMargoSupplies but maybe we can all get together and look at some test ideas if you guys are interested. 

Do you have estimate on power usage? I was astonished recently when I learned how power-hungry those pi's can be when running visual models. I am guessing it may need mains energy rather than a solar/battery array.

That’s great! Definitely interested!

Think in terms of 10W continuously.