Hi! I'm the CEO of Bounce Imaging. Now sure exactly how to post here but excited to see ourselves on here after someone sent me the link. Happy to show some sample video as requested: https://youtube.com/shorts/PlmG9HdzltU?feature=share
(I posted this quickly on my own Youtube as I'll need my colleagues at work to put up a better on our official page)
As noted, indeed our near-IR cameras (prior generations - the Recce360 Mini etc) are much easier to interpret in-flight than thermal just because it is hard to get your head around panoramic thermal as easily. If you want to see fully 360 video of that, just download our app (Bounce Viewer) and in the History section scroll down to demo videos and you can pan around as it is thrown in the air or shooting around on the back of a dog. Note that the stabilization is along multiple axes!
Indeed Jonas' setup was pretty cool but this had actually been tried many times before THAT- including cool designs by the Brits, US Navy, and others from decades ago and a cool conceptual design by Franziska Faro (spelling, sorry!). ALL however suffered from the challenge of doing real-time processing with low-latency and automatic stabilization in flight without melting the camera through too much processing. The way we cracked it, first for our visual cameras and now for thermal, is through a stitching method that is 200X-2000X more efficient and noise-insensitive because it is not based on SURF/SIFT feature detection (if you're into the nerdy side of things).
Have you looked at producing 3D reconstructions over the thrown trajectory? And/or something like a gaussian splat-based representation for viewing the whole trajectory at once?
Throwable 360 cameras owe their heritage to Jonas Pfeil, who built one for his thesis at TU Berlin [0], and later commercialized the idea in a camera called Panono [1]. They never really went mainstream, but it was a really cool idea with lots of buzz back then (10+ years ago).
I would love to see some footage off that in a throw - I would expect a blurry mess.
With panono, they captured a single shot at the highest point in a throw. In daylight, short exposure time. Only when throwing straight up without strong rotation.
With (uncooled) microbolometers you can't have a short exposure times and might end up with 20-30msec effective exposure. This thing might be harder to throw up straight without rotation. Throwing through a windows or rolling? Please show me that this does not give a blurry mess, but gives useful images!
Demo footage suspiciously missing...
Edit: I stand corrected - the spot video does not contain thermal but shows NIR images. One can see some point projections from a depth sensor?
So I commented (Bounce CEO) with some explanations, and posted a sample video of the thermal that I quickly put on youtube while my team can access our official account. Posting here too: https://youtube.com/shorts/PlmG9HdzltU?feature=share
Domestically (not on exported cameras that are restricted to 9FPS), you can actually hit ~27FPS reasonably easily and with our built-in IMU this is in fact usable in flight. However, thermal is hard enough to interpret when stationary so the main use case for the THERMAL variant is once it is in the middle of a room or on a pole to search attics etc. However our visual cameras are super easy to work with in flight - if you download our app (Bounce Viewer) and go to the demo videos in History you can see it working on the back of a dog or thrown in the air.
And I mean thousands of these systems are in use around the world (of the visual, the thermal Pit Viper we just announced) and have been thrown through windows, down stairs, on ropes, poles, vehicles, dogs....
Perhaps the use case is to toss it into a potentially dangerous position, let it settle, and then take the shot? I don't see any claims that this is intended to operate while in motion.
Midway-point between the authorities, or overshoot and see what's behind them. Depending on the scenery/layout placing 2-3 of these can make a big difference.
Also why not drop them from a small (less-noisy) drone? Drone with a camera can help surveil the situation from above, it can drop one of these balls, so you increase visibility by a lot.
These are intended to be used like grenades but for intelligence. If it fails or lands wrong it’s not a big deal the idea is they will pump these things out by the millions.
The only thing preventing this is lack of usefulness, you can do this now with a drone and Lidar but there isn't much profit in precisely mapping unknown underground structures. Or is there anything preventing this?
This is cool but I was hoping to see one (or more) thrown, bounce around, then see the software piece together a 3D sort of layout of everything it captured. Excited to see v2!
Whilst this seems cool. You won't be just throwing this. This will be a controlled item and you'll need to recover it, so eventually someone will lose one and then tears. Lots and lots of tears.
The idea is to throw this in instead of sending a person or a dog, reducing the odds they get hurt or civilians are unnecessarily hurt. And we keep costs as low as possible to reduce the tears (but honestly the camera is usually ok - pretty tough)
Rainbow Six Siege used to have an "operator" that could throw sticky cameras that looked a lot like the ones in TFA, though smaller. Sometimes real life imitates video games.
I didn't work directly for it, but I remember my team working on parts of the software and back-end stuff for Nokia Ozo[1], a 360° virtual reality camera. I feel that that camera was ahead of its time.
I don’t think many people would have thrown the OZO up to take cute panoramas though, as it went for something like $60,000.. At the time (~2017), the OZO was the benchmark quality-wise, and it hit the community pretty hard, when Nokia cancelled the project over night and disbanded/fired the team behind it.
At the time I was working for a startup called VideoStitch, later Orah, which had launched the Orah 4i, a much smaller, 4-lens, 360 degree 4K live streaming camera. Expectations were huge, that 360 degree video streaming would go through the roof, with a camera at every wedding. Didn’t really work out that way, I wonder whether stereo 180 degree “Spatial Videos” in the Apple ecosystem will go the same way.
What made it ahead of its time? Seems like Google Daydream and that wave of 3DOF headsets came out in a year later and had their 15 minutes of stereo 360.
That's funny that the wiki article didn't put a picture of the camera itself and instead wasted an image on a generic logo. The duckbill look of it was definitely odd. The real-time preview was useful to directors that could not figure out, "what does the camera see" <facepalm> It SEES EVERYTHING!!!! Just stand there and look around. That's what VR cameras see. Why is that so hard?
The Google camera that used 16 GoPros with more interesting to me even though it only did a cylinder rather than a complete sphere, but it did generate stereo images. We appended additional GoPros to the top/bottom of the unit so we could fill in the zenith/nadir of the final image. 22 total cameras. 22 mircoSD cards. DIT was a nightmare.
We also played around with the prototype from FB that was closer to Ozo, but that was right at the end of my time with live action VR.
None of these cameras were "throwable" though. There was a consumer unit though that would snap a 360° image at the apex of the toss.
> "what does the camera see" <facepalm> It SEES EVERYTHING!!!!
This is such a bad take, haha. Its as bad as telling the cinematographer to look forward. Obviously there's the exact point of view of the device but also the exposure, and other camera intrinsics you don't want to guess at when you're about to shoot a take.
Your take is also not a good one either as you are ignoring my personal experience, and it seems you're assuming I'm just making stuff up. The question isn't "is the exposure good", "did the actor hit their mark", or anything else that a direct might be concerned. It was "what does it see".
Yes, it is hard to direct from around the corner and not being able to see anything. The Ozo did a live stitch that could be viewed in a headset. Guess what, directors didn't like wearing a headset all day just like the general public. Pushing the stitched feed to a monitor also caused confusion as some just can't wrap their head around how the rectilinear image translates. It came down to use attaching a dedicated GoPro to the stand of the VR camera that looked "straight" just so directors/video village could "see" something. The directors were now much happier than all of the other attempts.
I know we've made significant progress wit on-device compute but I would be surprised to find out they are stiching in real time in that amount of space... Even doing this in real time on a commodity laptop would be pretty impressive to me.
It’s plausible to do this on device. We were discussing this product concept 8 years ago at one of the video/AI companies I was at. Between the relatively low res of thermal cameras and the cues you can get from various sensors about orientation it should be doable.
When the camera settings do not change, the positions are locked, etc, you can quickly do a "stitch" with a template that doesn't really take a lot of compute at all. It's by no means a clean stitch. Even in the demo video, you can see the edges from each camera. However, this is a product embracing the "know your audience" mantra, and it serves the purpose as designed.
Thermal cameras normally have a much lower resolution and only a 4 or 8 bits color depth. More than enough for locating people in a situation, I assume
“The man from Box 500 [UK MI5] proudly announced that the thermal-imaging camera at the rear had detected a pattern of movement. Every morning, he reported, at exactly 08:00, the hostages gathered in the commercial section on the ground floor. An SAS officer tartly pointed out the central heating was timed to go on every morning at that hour: the “hostages” detected by MI5’s machine were radiators warming up."
YES! That's the neat trick. Clever of you to get that without prompting! So I figure you must have some imaging background. We're hundreds to thousands of times more efficient than traditional SIFT/SURF or other feature detection methods, which is the only way we can do this at the edge on the camera itself (without melting it or running out of battery in five minutes).
Surveillance/recon equipment and drones seem to be the next "2A" things to buy. And anti-drone tooling. Directed RF jammers? Drone-deployable castnets? Shotguns with turkey choke? Heh.
Their visible cameras are $7k, I'd be surprised if this is less than $15k.
Assuming the Berlin patents have expired, a visible light or "night vision" near IR camera made from commodity sensors and on-device stitching is about $100-200 BOM which could be a $500 premium cat toy (and thus actually affordable as a tactical tool for users other than funded US/EU police). I've been finding gear for community safety guys in South Africa and it is amazing how much can get done with cheap consumer gear now.
I personally think the m4 is way overrated. Compared to its two cousins the 1301 and 300 patrol it’s just larger and heavier and more powerful expensive.
But… I do think it’s funny that basically in a year shotguns have been made relevant again.
I’ve considered buying a couple of the $5 Alibaba drones to “test”.
If I could hook the cheapo alivaba drones up to my better DJI equipment, I think that might make for a fun 3gun course this year.
Would be fun to do "smart clays" -- I wonder if you could make a clay which just shifted weights to affect flight path, something cheaper thna a full drone. Or you could fly a $300-500 non-expendable but not-end-of-the-world-if-hit drone trailing a target on a long tether, with no-shoot above a certain angle to keep the drones themselves mostly safe?
(and yeah, I have an M4 and 3 Turkish knockoff M4s (1 great out of the box, one I fixed up, one which doesn't love to eject but I can probably fix), and the 2 x 1301s I own are better.)
For me it’s large I’m not a small guy and the grip on the m4 is just big in my hands, then manipulating the safety is tough. At least the Turkish clones are cheap!
> Or you could fly a $300-500 non-expendable but not-end-of-the-world-if-hit drone trailing a target on a long tether, with no-shoot above a certain angle to keep the drones themselves mostly safe?
I had considered that. I’ll try hanging something off an avata2 but my concern will definitely be that I’ll need some rigid structure to land and take off. If the line is too close I think I could suck it up into the props
Interesting. I assumed it was just sloppy journalism but it sounds like you’ve seen this confusion before. The name did not strike me as hard to remember.
Terminology-wise, what’s the better way to communicate that something captures a truly omnidirectional spherical view, not just 360° in a circle? Seems this article has gone for ‘360° panoramic’ but to me that implies blind spots above and below, but the camera placements on this device seem to have full spherical coverage. Shouldn’t they call it a 41,253 square degree camera?
SO we looked at this and it adds so much weight and battery drain. There are a million micro-robots - believe it or not this form factor and simplicity is what makes us a go-to tool (contrary to what you might think, under a lot of stress first responders may not want to drive something around given the additional cognitive load)
You already need someone watching the screens. I don't see the driving being a big increase from that. Weight and battery life don't seem to be a concern. There are drive mechanisms that use mass distribution or inertia to drive it like a wheel and don't use much energy. This is especially true if you're throwing it and just need to retrieve it because you made a bad toss, or need to clear an additonal angle.
As noted, indeed our near-IR cameras (prior generations - the Recce360 Mini etc) are much easier to interpret in-flight than thermal just because it is hard to get your head around panoramic thermal as easily. If you want to see fully 360 video of that, just download our app (Bounce Viewer) and in the History section scroll down to demo videos and you can pan around as it is thrown in the air or shooting around on the back of a dog. Note that the stabilization is along multiple axes!
Indeed Jonas' setup was pretty cool but this had actually been tried many times before THAT- including cool designs by the Brits, US Navy, and others from decades ago and a cool conceptual design by Franziska Faro (spelling, sorry!). ALL however suffered from the challenge of doing real-time processing with low-latency and automatic stabilization in flight without melting the camera through too much processing. The way we cracked it, first for our visual cameras and now for thermal, is through a stitching method that is 200X-2000X more efficient and noise-insensitive because it is not based on SURF/SIFT feature detection (if you're into the nerdy side of things).
reply