They are really good for edge devices. We have a Raspberrypi 5 running as a thermal imaging enabled wildlife camera with local AI. I don't think you can find another platform that can support this as well.
Yes. The Orin Nano is amazing. We support that as well. Once I’ve added nvme SSD and a case on the pi, if I use an 8GB pi then it’s getting up to 1/2 the price of the jetson. But the jetson does inference 10x faster. Plus it has fantastic hardware video support.
None the less, the Pi is still faster and physically smaller. If I’m spinning up lots of them for test scenarios having the pi as a platform still is really handy.
But absolutely. The Jetson is brilliant. The Pi is none the less a fantastic platform.
I think so of the reactions toward the Pi here are a bit harsh. The price hike is not their fault, it’s all the AI companies that will continue to suck your tax dollars and give them to billionaires. Raspberry Pi have done an astonishing job for the whole communities and we should be thankful for that company.
Harsh commentary on the pi is nothing new. I’ve been a user since the 256mb pi b, so I’ve developed a sense of humor about it. I look back with amusement (and annoyance) at episodes like the Spanish hacker who heaped abuse on the pi team on Twitter for how they handled the drivers for their camera, even telling them they had pulled a “full Apple,” prior to dedicating years of his life to developing Linux drivers for macs. Or pretentious hn users with bios that include vc and starting their own companies complaining about the cost of a Pi b.
In what way is the pi5 faster than the Orin nano? Not gpu or inference, so… integer or floating point?
That’s a little black and white though. It’s ruling out that the AI could make a pretty good guess as to what you likely wanted to convey. If it gets that right, then you didn’t need to understand it nearly much.
When I observe its reasoning I see it doing this all the time.
Sure, but at that point it hardly makes sense to discuss any of the details of the implementation since we are just treating the LLM as a black box. For all we know, it could be maintaining its own specialized programming language made just for the application.
Oh there no denying that they are a threat to live stock that are not protected. And sometimes even when they, as the sometimes can get over fences.
In the Netherlands there are 14 wolf packs and 144 wolves in a geographical small area where the surrounding farms are often full of a lot sheep.
But it’s also easy to imagine that if you can become aware of the presents of wolves sufficiently early enough and reliably enough before they attack your animals then you have a chance to prevent an attack. And if you prevent enough attacks it becomes maybe harder than sticking to natural prey.
Wolves don't like taking risks. Interacting with humans is definitely risky. Detecting wolves reliably early and responding, either in person if that can be done quick enough, or setting off responses, ideally followed up by a human visit could well result in the wolves avoiding that farm in any case. The extend to which this can work in practice is something that needs to be determine through research and pilots. Wolves are very smart, so if tech is going to be able to help here, then the tech has to be very capable. The start with pilots and research two elements are needed. People with tech, but also people in the affected areas with a willingness to collaborate.
The current highly polarized environment is throwing road blocks against such collaboration.
The current situation is such that owners are unaware of the presence of wolves at their borders, so there is currently also not a lot of perceived risk associated with attacking live stock. A useful goal is to start to change that. And the people who read hacker news I'm sure that think of many ways to change the status quo. My system is one example.
Sadly, a lot time went by without trying these methods. None the less, one should try.
I the past I have offered free equipment to people to test who had been publicly expressing concern about the arrival of the wolf but they would not accept it.
So far these packs are monitored, but they contain problem wolves that have been known to attack people. Examples in Utrecht and veluwe, and it takes just a little to learn from those; the issue is that the wolves do not see people as a threat. So far there have been incidents, like bite marks, until they will attack in a group. They should adopt the German method to scare them away proactively
Only two occasions though out of more than 60. All the other times they were just traveling through.
And I also don’t believe those two times they were a threat to the horses, more likely just curious. But I didn’t have the full visibility enough to be sure as there was just one camera. So I called the farmer to be in the safe side.
Now I think there are no more wolves there. It’s been weeks since Insaw them. They have either dispersed or killed on the roads.
We have a solar powered thermal local AI wolf detection system that’s been in a field in Belgium for more than 8 months now, monitoring the wolves there. Thermal image motion detection can detection them at much greater distances than PIR sensors and of course in complete darkness.
The biggest barrier at the moment to living with wolves, if you true away the human resistance to the idea, is reliable detection for early warning prevention.
This is not such a hard problem with computer vision at the level that it is these days. We’ve made it more effective by coupling it with thermal imaging camera modules.
Now we just need people willing to want to want to work with such pilot project, which turns out to be a pretty big problem.
However, it’s possible that a few influential success stories can bring change to the situation. But there has to be a start first and certainly when this started to be a problem, anything along the lines of prevention was seen as a barrier towards killing them which is what the solution in many people’s eyes is.
That is a glorified trail camera, really does very little to help people live with wolves. What does is a strict hunting quota; the government decides how many wolves there should be, rather than nature.
And no, it's not a glorified trail camera. Very different.
Everyone wants to kill before trying measures to keep people and their animals apart.
I don't think you can say it does nothing to help people live with wolves because it's not being tried. It's not being tried, because people being convince the only thing that helps is killing everything that requires a bit of effort to live with. I realize it's very human to destroy and kill, but I think we should strive to do better.
Humans have already stolen so much from nature and the animals, don't you think a little effort to do better would be a good thing?
Sure they are. But try mentally reversing the situation. If any other single species occupied and changed the landscape as much as humans they would be viewed by humans as the most horrific plague on earth.
So I'm just a little empathy would be well placed here. A little effort to try and co-exist with the other inhabitants would not be out of place and as a bonus you can think a little better of yourself.
That's a fine attitude for squirrels which are ultimately harmless, but wolves can attack people, although that's somewhat rare, not like bears which attack you every time.
Wikipedia says wolves are more likely to attack humans if they've been living around humans for a while - they learn that humans are viable prey, not that humans are harmless.
Bear attacks depend on the (sub)species. European brown bears are very shy and much less likely to attack people than wolves.
Finnish people traditionally respected bears but were not afraid of them. Historical records mostly tell of two kinds of bear attacks. Either children / old people herding cattle in the forest accidentally got too close to a bear cub, or adult men were hunting bears and something went wrong.
There is a lot not to be proud of about human behavior. Still.. one has to try to be a little positive about there being hope as hopelessness is contagious.
They kill and eat things. Brutally and sometimes primarily for fun. Unless you somehow consider yourself no longer an animal I'm not sure eradicating capable predators is deserving of 'no longer living on this planet'
Cameras do not actively protect people or livestock, it just monitors and retroactively actions are taken. I prefer the German method where they make the wolves fear people.
The systems I developed are. It based on microcontrollers. But secure boot Linux systems. They detect wolves within 2 seconds on pi based version, sub second on the jetson based ones.
Then websocket connected subscribers speak a loud alert to wake the person up and auto load the camera view over the permanently up VPN.
The person monitoring can then verify on how ever many cameras are connected and illicit many different kinds of responses.
This is not a standard wild camera setup. I’m using high resolution 640x512 res thermal modules to see night and day.
Our system is already being used in Greenland for polar bears and has detected bears three times now from a long distance and the safety group were able to chase it away. There it’s monitoring 22 cameras with a single jetson.
And indeed, the idea is very rapid detection and validation and then potentially remote active response. It has all the necessary controls.
