Virtual Forest | Hacker News Reader

Virtual Forest

virtualforest.io

198 pointshabi60 comments

Great stuff! I'm a little frazzled by the fact that you're leaving 350$ cameras in the woods unattended, though. Figures they're not too far from your home given the CAT5 length you mention, but for anyone interested in building one in a more remote location (IoT SIMs with decent data plans are a thing now!), you might want to consider camouflage, and/or setting up in a tree.

> leaving 350$ cameras in the woods unattended

Hunters do it all the time.

I worked on a revamp of the system for controlling those cameras remotely, viewing all your pictures, etc... (the cameras had cell modems) at my last job.

Its insane. Not only are the cameras a little pricey (not crazy), but HOLY SHIT the data plans are expensive. Granted, that's where they make money, but still.

Wha? Not anymore. You can get some really cheap data plans these days. Granted I don't know what the reception would be like out in the wilderness...

you can get cheap dataplans as a consumer, straight from the cell providers. Most of the game cameras do not let you just plug in a sim card.

Most of the game camera companies act as resellers of data plans from places like Verizon and AT&T; so they 100% control the data plan rates.

You'd be surprised how much untouched wilderness there is out there. And generally opportunistic criminals don't tend to be wandering through the wilderness looking for cameras to steal. And even then they're face will be recorded for all they know.

Yeah. There are many trails in the US where you can hike for days without seeing another human.

Maybe if you go in circles. If you cover 20 miles/day, even in places like Alberta or Wyoming, it's hard to go much more than two days without bumping into another one in one form or fashion.

On the trails, sure, but a short way off of even the busiest trails you could go an extremely long time without seeing anyone, especially in areas that don't allow hunting.

For example, this woman died after weeks spent lost just near the Appalachian Trail: https://www.bostonglobe.com/metro/2016/05/25/hiker-who-died-...

Yeah - we used to have a few of those unfortunate events reported every year when I was growing up in Wyoming. Sometimes even when it wasn't cold (dehydration is a thing, apparently). As young kids, we didn't know anything about anything, and always wondered, "why didn't they walk in a straight line? they were never more than 75 miles away from help."

It's really hard to walk in a straight line in dense woods. A compass helps, but even then you have to know how to use it. It's usually best to find a stream, follow it down to a river, and then follow that down to some sort of settlement (hopefully).

75 miles is a lot to cover on foot, too!

It is! Almost two days on open range! (Which is why dehydration is such an issue)

When I was a kid, we were amazed at the people who apparently got lost on Hell's Half Acre, about 40 miles west of Casper:

  - https://www.google.com/search?q=hells+half+acre+wyoming&num=20&source=lnms&tbm=isch&sa=X&ved=0ahUKEwi2iqHTsOXPAhWCsFQKHegtCo0Q_AUICCgB&biw=1151&bih=818

Broke an ankle or got stuck in a gully - sure, but lost?

Now that I'm officially old, and far less energetic, I'd be willing to believe that they just sat down and hoped that help would find them in time.

here in California there are people even in the woods.

>IoT SIMs with decent data plans are a thing now!

[Citation Needed]

---

What do you mean by decent data plans? I'm seeing plans that are between $10 and $20 usd per GiB. That's definitely not good enough of live video, but I suppose you could release new video every month.

It looks like it's just a single 360 camera with one viewpoint so it isn't really a "true immersive experience" on VR headsets, it's just a 2D image projected on a sphere. For true immersion in VR you need stereoscopic 360 degree recording, which can be achieved with using two of the 360 cameras or with this mount for GoPros:

http://izugar.com/product/z6x3d.html

Surely if you have two 360° cameras, one of them will be capturing the other one. I would think you'd need one 360° camera rig with stereo pairs of cameras.

With just two cameras, not only would you capture the other camera in view, but in those co-linear directions I don't believe it would be possible to get any depth information (or it would be all weird, like the same view, but magnified in one eye).

See: Lytro, which has struggled to find a good use for its lightfield cameras, but might actually have something now. Lightfield cameras can extract depth information directly, which lets them synthesize stereo views with a single camera.

Duct-tape enough of them together and you can synthesize stereo views in every direction, which is pretty cool.

  this mount for GoPros

Six GoPros, to be specific.

Pitch black. Not sure if nighttime, or not working correctly :P

Looks like something is working, the browser is receiving an image and it's not a solid black colour, turning up the contrast to the max you can see some difference:

http://i.imgur.com/Tm6J5iI.jpg

I'm not sure if it's just image compression artefacts or that's what the camera is actually seeing before it's processed and sent to users.

Does anyone else find it weird that you click and drag in the opposite direction you (well I) would expect, considering that the cursor becomes a hand that grabs?

Yes. I came here to say this.

Filed a PR. https://github.com/aframevr/aframe/pull/1987

I don't have a cardboard to properly see if the image really is "3D", but it seems like it's just a photosphere and therefore doesn't really have depth information. Is that right?

That is correct.

Sometime in the future, consumers will have software that can turn 2D images into simulated 3D [0]. It's still in dev mode right now, though.

[0] https://www.youtube.com/watch?v=Oie1ZXWceqM

That video is extremely impressive! Colour me excited

Yeah, and there are some other in-dev camera rigs that use multiple cameras plus software interpolation to simulate 3d as well.

The thing that's most intriguing to me are the experiments using multiple depth cameras set up around a space and using software to build a live, 3d model and overlay the video data as a texture on top of the models. It's all very rudimentary and low-res at the moment but it's the sort of thing that can eventually become 3d/VR telepresence and that just strikes me as awesome.

I'm very new to this sort of thing, but it seems like "stereoscopic video" isn't nearly as common as I expected it to be. 2D panoramas seem to be very much the norm. I'm not sure if this is down to production difficulty, delivery difficult or stereoscopy just not being impressive enough.

One thing is that it seems to me that anything that's recorded by a camera (rather than rendered in realtime) is going to be "wrong" once you tilt your head, as your eyes are now on top of each other rather than next to each other.

Reminds me of Quicktime VR. Does anyone remember Quicktime VR? I mean...we had this tech... in the early 90s.

Now I have the nightmares again. (I got to implement spherical projections in realtime on a Pentium 90. It's... an interesting problem. Yeah. That's what I'll call it)

But yes, we had it. We had VR helmets, too. Consumer-grade. Remember the VFX-1? (https://en.wikipedia.org/wiki/VFX1_Headgear)

But this time, VR is going to be different. Honest.

Yup. The Ricoh Theta S captures spherical panoramas. The VR terminology is just overselling it. Neat idea though. It would be cool if there was a sonic component.

I think you're on to something. Say we record monaural audio with directional mics on/beside each cam, then encode and compress each stream, allowing for realtime stereo mixing during playback determined by view angle. Add a compass, accelerometer, gyro to track orientation. Couldn't we then achieve the desired effect and even simulate spatial audio effects, 6DoF movement in scene, blend UI sounds and add 3D sound to the environment? AR anyone? With a small peripheral you could emit a few chirps at diff frequencies and measure them using same mic rig to create a virtual map of the environment's acoustic characteristics and use it to render sound effects for composite elements, generated UI, nav feedback, similar to the way image based lighting is used today to make artificially generated objects appear as if they were really present in the scene.

Sounds like a good open hardware/software project but I'm short on cameras and mics for something like that. Anyone see potential there?

Combine with laser rangers and filters for their wavelength on the cams, and you can sample 3d point cloud data too and render the environment as a 3D (4D) scene, use it for composite reference, or slap a small LiDAR scanner under the whole thing for precise measurement.

Pretty cool. Make it live and add some sound and you could almost feel like you're there.

Simple nitpick: when I drag to the left, I expect to go to the right. (I'm just seeing this through my browser though, maybe it makes sense in VR, idk.)

There is no consensus on drag direction. I've built several projects and have alternated the drag direction on each and have always received vociferous complaints about both.

If there is no consensus on drag direction, Littlstar's player should be the consensus. It gives a major wedgie to every other player in the space, at the moment. Especially when it comes to motion and drag.

Or just make it a preference, like what Apple does with trackpad scrolling direction.

Making it a preference means creating a settings UI. People have their preference, but they are also adaptable. It's not immediately clear that a "VR" experience (not that I'd call 360° photo "VR") would benefit from a preferences UI.

Doesn't really affect VR, VR is absolute. Those are just for desktop controls...sort of up to the developer which way they want it to drag, but the default is left/left...may have to update that since I do hear your preference is more common.

Drag direction in 3D depends on user background and kind of application. It depends on if you expect to be moving the world or a virtual character. In the case of a panorama viewer I understand that some people might expect to move the picture (the world) as you do on a phone photo gallery or map application. For a first person interactive application it's more likely to expect that the character will turn to the direction you are dragging the mouse. You're controlling the character and not the world.

This is cool, it's a DIY 360 degree camera attached to a Raspberry Pi that you can view with a VR headset.

This page is the build and parts list but the web VR using A-frame is at: http://virtualforest.io/vrforest.html

Demo is just a static image... isn't this supposed to be a webcam or something?

> Images stream live every 15 minutes from 4 - 22h in the EST (GMT-5) time zone

There is a large 360 YouTube timelapse video here: http://virtualforest.io/vrtimelapse.html

If you want to do non-360 timelapse with the regular (and much cheaper) Pi camera then I wrote a guide: https://unop.uk/time-lapse-photography-with-a-raspberry-pi

> the web VR using A-frame is at: http://virtualforest.io/vrforest.html

And it works better with WebGL enabled, else you get a rather empty page.

On Chrome on Nexus 6 with Android 7 get black screen regardless of orientation or screen touching/dragging.

This is incorrect:

    On mobile devices the A-frame framework will use the device sensors to figure out it's relative position.

It will only figure out orientation, not position. Position is impossible to integrate from the low-end IMUs used in smartphones.

Doesn't seem to work for me on my HTC Vive but I've never used any A-Frame site before so I might just be doing it wrong. I tried on both Chrome and Firefox.

For starters, you need an experimental build of Chromium that Google provides in a ZIP file, or you need Firefox Nightly, and in both cases you have to enable a flag.

After that, I do not know what the current status is for A-Frame supporting the Vive. And of course, there was recently a big change to the WebVR API, so A-Frame may not be caught up to that, either.

As far as I know, A-Frame's "known good configuration" is Oculus Rift + Firefox Nightly, or (HTC Vive|Oculus Rift) + not-latest Experimental Chromium.

And Windows 10, of course, because Facebook can't find a single person capable of doing graphics work on Linux out of 15,000 employees, or so they would have us believe.

You can find access to all of these things at http://www.webvr.info

> because Facebook can't find a single person capable of doing graphics work on Linux out of 15,000 employees

Either you muck around with NVidia binary Linux drivers (or worse, binary drivers for some ARM chipset in an Android phone), or you're working with Intel stuff (which isn't going to give you a cutting-edge VR experience).

A for-profit corporation has better things to spend its 20% time on.

A-Frame is caught up and supports the latest WebVR 1.1 changes. Rift/Nightly (until Vive support lands in FF), or Chromium builds.

The A-Frame framework that is mentioned can be found here: https://aframe.io/

What's the UBEC for?

Edit: does it just convert the 24v to 5v, and why is that better than just doing 5v to begin with? Is this so you can get more current?

If the power supply is far away it's better to have a higher voltage due to voltage drop due to cable resistance, 24V is pretty standard AFAIK.

Really cool! I'll have to test this with Cardboard.

The VR view seems to crash Firefox on Android 5 / 6 on two devices I tested though.

This is really cool, and I'm not complaining or feature creeping at all, but what about the sound of the forrest?

There doesn't seem to be anything to control condensation.

Do you think this might become a problem?

Doesn't seem to work well with custom orientation in Chrome devtools: http://i.imgur.com/fp2kSO8.png

So what? The idea that the developer should care about this is beyond comprehension.

I recommend https://chrome.google.com/webstore/detail/webvr-api-emulatio...

Reply on news.ycombinator.com