The San Francisco data is copyright 2007 Google. The Immersive Media imaging system is currently slightly lower resolution but we also are taking one shot every 4 inches (30fps) of travel rather than every 100 feet. I do not think that the resolution difference will be permanent feature. All it takes is development time, money and a whole lot of storage.
All of the non-San Francisco Street View data is provided by a company
called Immersive Media. They
have a special omnidirectional video sensor with 11 elements that shoots
30 frames per second. The 11 cameras do a great job rejecting glare from
the sun. Compare the SF footage with the Las Vegas footage and look for sun
glare overriding the sensor. At street speeds, there is about 1 image every 3 to 5
inches. Street View is showing you one frame every 30 to 100 or so.
The Teleatlas camera car doesn't shoot panoramas, the cameras are too far
away to avoid massive parallax errors and their cameras are pretty narrow
field of view. I'm sure the collect very good POI data, though. The survey
vehicles used for the Immersive Media dataset are actually Volkswagon Beetles,
there is a tiny picture on the Immersive Media homepage. The camera can actually
see down most of the way to the road and anything other than a Beetle has a
pretty big footprint in the image. The camera system also see straight up
even though the Flash viewer in Street View does not. It's actually the warping
of the pixels to make the view that is the weakest link in the distribution chain.
The vehicles have the camera system and a special inertial positioning
system that provides survey grade coordinates as the vehicle moves down
the road even underground. That system is made by Applanix and it's the
same type of system used by many of the Darpa Grand Challenge Candidates.
All this adds up to many TBs of data and although it isn't easy to stream on the
web, they have figured out how to do it. If you visit the demo page
you can see full motion video panoramas that you can drag and look up,
down, left and right in! Requires Shockwave from Adobe. The streaming isn't
as sharp as the original product but it gives you an idea of navigating
an Immersive movie. Sort of like Quicktime VR but it is really a movie!
Immersive Media has collected data all over North America, you can see the complete
extent of their collects
and browse some clips. We also just announced a major expansion into Europe
so we'll see you blokes over the pond soon!
Full Disclosure: I wire the systems on the Beetles and write post-processing
software for Immersive Media. I've trained a lot of drivers in how to run
inertial positioning systems and I'm really pleased that data I support
is finally being seen by people! And feel free to Slashdot the demo page,
the servers are waiting to show you our movies. Remember to click and drag to
look around, this isn't boring old static web video where you look where we
tell you too.
You could do it with two projectors in
the center provided they had ultra wide angle
lenses like an
Elumen's projector
but those are somewhat expensive. One could
easily spend $24k on precision wide angle lenses that
would blend well at the edges.
The smaller "planetoids" globe in the lower
depths of the Sci-Fi mueseum in Seattle uses
a single Elumen's lens/projector hidden in the
wall projecting from the hidden back side of
the globe.
I have access to one of these projectors and
I've long thought about trying to find someone that
could make a bubble of reverse projection material
plastic.
Robosaurus is worn, not driven
on
Robosaurus
·
· Score: 2, Interesting
I've built one of these too, and all things being equal, I think you would be better off spending $120 to get one of the Steady cam clones. True, he has some cool shots on his page but those are not nearly as easy as he makes it
out to be. Maybe I am just clumsy.
When I walk forward my system wants to behave
like a pendulum causing the camera to rock
forward and back around the horizonal fulcrum.
If things aren't perfectly balanced it is very
difficult to keep the cameras tilt at a given
attitude. Your left hand (if you were the
author in the photo on the page) will not be
able to keep the attitude without pendulum
style oscillation. It's also difficult to
make the camera turn around the camera of
the horiontal bar and the fact that the
rotational inertia of the person-pipe-camera system is not appropriate for turning around
the camera.
Beyond those basic problems: it's also hard to hold on to and I tend to smack into door frames and innocent bystanders with the horizonal pipe.
One of the key parts to a steady cam rig is a gimbal joint that isolates tilt/tip motions of your hand from the "mass" that has the camera. Without this isolation it's really hard to get good shots without Zen master balance or just
being lucky.
If anyone out there wants to make a Steady-cam
like rig, I suggest they copy something like
the Flowpod. Note the gimbal connecting the handle
to the body of the device.
I watching it and it looks to me like they've got something wrong with their video feed. Moving edges are all hazy/swimmy. I don't know what it looks like to have the fields reversed but this is what I would guess it might look like. It really blurs things together under motion. Watching the crowd behind a panning camera is painful.
Could this have something to do with the down-conversion of progessive scan HDTV to NTSC?
All you ham radio operators out there in Slashdot land will remember packet radio. At design time Packet radio was the cheapest, simplest and least robust hack TAPR could come up with. It was basically a Bell 202 phone modem with the X.25 protocol strapped on top. Lack of forward error correction and low bitrate for the bandwidth were some of the highlights of this protocol.
Amazingly, hams managed to turn this turkey into a globe-spanning email network. I ran a node for this system for a number of years and tried to carefully baby the outmoded and weak communications links it was based on. In the later part of the 90s this network just sort of collapsed with the advent of the infinitely superior Internet.
I believe that the invention of packet radio set back amatuer radio for years. We could of been designing elegant PSK modems and high performance P2P networking protocols for years, but, instead, all of our efforts were concentrated on 1200 baud FSK keying. We wouldn't of made the Internet but we could of had a robust network for handling emergencies when the normal wideband systems failed.
I've tried for three or four years to do 1394 style video editing under Windows. I've fiddled with every hardware configuration and used every capture program under the sun and I still can't capture more than a few minutes of video without loosing frames. I read the various forums occasionally and it seems to me that a weegie board has more relevant things to say about video editing.
It's not your motherboard. It's not software X. It's all Microsoft. I dual booted RedHat (so my other box is debian, I was lazy) and low and behold I can capture for HOURS and nary a dropped frame. When it did drop a frame, dvgrab politely told me why. This stuff works. Too bad I can't edit under linux yet. When Cinelerra has the stability and feature set of something like Sonic Foundry's Video Vegas desktop video will finally stop being an aggravating trip through the worst that personal computing has to offer.
By the way, if you are a Windows user frustrated with your editing app crashing get Video Vegas. Despite the crazy name it has plenty of professional features and it's rock solid. Unlike Premiere, which I can crash just by blowing on the case gently, Vegas let's me get through hours of footage with no back talk.
The whales? Really?
on
Wireless Wales
·
· Score: 2, Funny
I can't even get DSL out here in the boonies but the Orcas are all runnin 802.11 on the World Whale Web.
As a current Reedie I get a big kick out of
everytime I hear The Center for Advanced Computation At Reed College. Basically a renovated two-bedroom residential house that was annexed because it was too close to the campus this unassuming building is even unknown to most Reed community members. It's absolutely hilarious when the media come by looking for The Center for Advanced Computing and we point them to a small house behind some trees. As far as I know Crandall is the only faculty there.
Crandall is a wonderful lecturer whose seminar talks never fail to befuddle the rest of the faculty (let alone the undergraduates). Personally, I have made a solemn vow stay away from the areas of physics that will likely degenerate into number theory. Regardless of my preferences, his class in scientific computing is more than worthwhile.
It's sad that Public Key Cryptology, as it relates to the web, has become a way of distributing money from web authors to a single company. What do we get for our money when we get a CA certification? A token verification that we have a credit card in many cases. What does the critical certification prove in terms of a trust relationship? Zip. A signed cert doesn't mean I have a secure server that protects my subscriber's credit information, nor does it imply that I'm in any way honest. I like the PGP web of trust model, but I don't know how it could be implemented on the web. There ought to be someway to have PKI without big silly corp in charge.
Anyone find it strange that the R and D department and the customer relations people were so close that it was possible to swap outgoing products?
If I was running a company that was building game consoles or 'high tech' stuff the R/D guys would work in an unmarked building on the outskirts of some town in Arizona. There would be lots of dish antennas, razor wire fences, and rabid guard dogs with names like "pookie" and "raul." Lastly, all one million dollar engineering prototypes would not go out UPS. There great and all, and I'm sure they do great with liver transplants and stuff, but this is my prototype. Hell, I wouldn't even trust the NYPD. My delivery people would be the Secret Service.
They rent out, don't they?
If not I'll call the Swedish Ski Commandos of Doom and Questionable National Alligence.
Slashdot Mirror
The San Francisco data is copyright 2007 Google. The Immersive Media imaging system is currently slightly
lower resolution but we also are taking one shot every 4 inches (30fps) of travel rather than every 100
feet. I do not think that the resolution difference will be permanent feature. All it takes is development
time, money and a whole lot of storage.
All of the non-San Francisco Street View data is provided by a company called Immersive Media. They have a special omnidirectional video sensor with 11 elements that shoots 30 frames per second. The 11 cameras do a great job rejecting glare from the sun. Compare the SF footage with the Las Vegas footage and look for sun glare overriding the sensor. At street speeds, there is about 1 image every 3 to 5 inches. Street View is showing you one frame every 30 to 100 or so.
The Teleatlas camera car doesn't shoot panoramas, the cameras are too far away to avoid massive parallax errors and their cameras are pretty narrow field of view. I'm sure the collect very good POI data, though. The survey vehicles used for the Immersive Media dataset are actually Volkswagon Beetles, there is a tiny picture on the Immersive Media homepage. The camera can actually see down most of the way to the road and anything other than a Beetle has a pretty big footprint in the image. The camera system also see straight up even though the Flash viewer in Street View does not. It's actually the warping of the pixels to make the view that is the weakest link in the distribution chain.
The vehicles have the camera system and a special inertial positioning system that provides survey grade coordinates as the vehicle moves down the road even underground. That system is made by Applanix and it's the same type of system used by many of the Darpa Grand Challenge Candidates.
All this adds up to many TBs of data and although it isn't easy to stream on the web, they have figured out how to do it. If you visit the demo page you can see full motion video panoramas that you can drag and look up, down, left and right in! Requires Shockwave from Adobe. The streaming isn't as sharp as the original product but it gives you an idea of navigating an Immersive movie. Sort of like Quicktime VR but it is really a movie!
Immersive Media has collected data all over North America, you can see the complete extent of their collects and browse some clips. We also just announced a major expansion into Europe so we'll see you blokes over the pond soon!
Full Disclosure: I wire the systems on the Beetles and write post-processing software for Immersive Media. I've trained a lot of drivers in how to run inertial positioning systems and I'm really pleased that data I support is finally being seen by people! And feel free to Slashdot the demo page, the servers are waiting to show you our movies. Remember to click and drag to look around, this isn't boring old static web video where you look where we tell you too.
You could do it with two projectors in the center provided they had ultra wide angle lenses like an Elumen's projector but those are somewhat expensive. One could easily spend $24k on precision wide angle lenses that would blend well at the edges.
The smaller "planetoids" globe in the lower depths of the Sci-Fi mueseum in Seattle uses a single Elumen's lens/projector hidden in the wall projecting from the hidden back side of the globe.
I have access to one of these projectors and I've long thought about trying to find someone that could make a bubble of reverse projection material plastic.
Check out the Waldo style user interface this thing has.
I've built one of these too, and all things being equal, I think you would be better off spending $120 to get one of the Steady cam clones. True, he has some cool shots on his page but those are not nearly as easy as he makes it out to be. Maybe I am just clumsy.
When I walk forward my system wants to behave like a pendulum causing the camera to rock forward and back around the horizonal fulcrum. If things aren't perfectly balanced it is very difficult to keep the cameras tilt at a given attitude. Your left hand (if you were the author in the photo on the page) will not be able to keep the attitude without pendulum style oscillation. It's also difficult to make the camera turn around the camera of the horiontal bar and the fact that the rotational inertia of the person-pipe-camera system is not appropriate for turning around the camera.
Beyond those basic problems: it's also hard to hold on to and I tend to smack into door frames and innocent bystanders with the horizonal pipe.
One of the key parts to a steady cam rig is a gimbal joint that isolates tilt/tip motions of your hand from the "mass" that has the camera. Without this isolation it's really hard to get good shots without Zen master balance or just being lucky.
If anyone out there wants to make a Steady-cam like rig, I suggest they copy something like the Flowpod. Note the gimbal connecting the handle to the body of the device.
I watching it and it looks to me like they've
got something wrong with their video feed. Moving
edges are all hazy/swimmy.
I don't know what it looks like to have
the fields reversed but this is what I
would guess it might look like. It really
blurs things together under motion.
Watching the crowd behind a panning camera
is painful.
Could this have something to do with the
down-conversion of progessive scan HDTV to
NTSC?
All you ham radio operators out there in Slashdot land will remember packet radio. At design time Packet radio was the cheapest, simplest and least robust hack TAPR could come up with. It was basically a Bell 202 phone modem with the X.25 protocol strapped on top. Lack of forward error correction and low bitrate for the bandwidth were some of the highlights of this protocol.
Amazingly, hams managed to turn this turkey into a globe-spanning email network. I ran a node for this system for a number of years and tried to carefully baby the outmoded and weak communications links it was based on. In the later part of the 90s this network just sort of collapsed with the advent of the infinitely superior Internet.
I believe that the invention of packet radio set back amatuer radio for years. We could of been designing elegant PSK modems and high performance P2P networking protocols for years, but, instead, all of our efforts were concentrated on 1200 baud FSK keying. We wouldn't of made the Internet but we could of had a robust network for handling emergencies when the normal wideband systems failed.
I've tried for three or four years to do 1394
style video editing under Windows. I've fiddled
with every hardware configuration and used every
capture program under the sun and I still can't
capture more than a few minutes of video without
loosing frames. I read the various forums
occasionally and it seems to me that a weegie
board has more relevant things to say
about video editing.
It's not your motherboard. It's not software X.
It's all Microsoft. I dual booted RedHat (so my
other box is debian, I was lazy) and low and
behold I can capture for HOURS and nary a dropped
frame. When it did drop a frame, dvgrab politely
told me why. This stuff works. Too bad I can't edit
under linux yet. When Cinelerra has the stability
and feature set of something like Sonic Foundry's
Video Vegas desktop video will finally stop being
an aggravating trip through the worst that personal
computing has to offer.
By the way, if you are a Windows user frustrated
with your editing app crashing get Video Vegas.
Despite the crazy name it has plenty of
professional features and it's rock solid. Unlike
Premiere, which I can crash just by blowing on the
case gently, Vegas let's me get through hours of
footage with no back talk.
I can't even get DSL out here in the boonies but the Orcas are all runnin 802.11 on the World Whale Web.
As a current Reedie I get a big kick out of everytime I hear The Center for Advanced Computation At Reed College. Basically a renovated two-bedroom residential house that was annexed because it was too close to the campus this unassuming building is even unknown to most Reed community members. It's absolutely hilarious when the media come by looking for The Center for Advanced Computing and we point them to a small house behind some trees. As far as I know Crandall is the only faculty there.
Crandall is a wonderful lecturer whose seminar talks never fail to befuddle the rest of the faculty (let alone the undergraduates). Personally, I have made a solemn vow stay away from the areas of physics that will likely degenerate into number theory. Regardless of my preferences, his class in scientific computing is more than worthwhile.
It's sad that Public Key Cryptology, as it relates to the web, has become a way of distributing money from web authors to a single company. What do we get for our money when we get a CA certification? A token verification that we have a credit card in many cases. What does the critical certification prove in terms of a trust relationship? Zip. A signed cert doesn't mean I have a secure server that protects my subscriber's credit information, nor does it imply that I'm in any way honest. I like the PGP web of trust model, but I don't know how it could be implemented on the web. There ought to be someway to have PKI without big silly corp in charge.
Anyone find it strange that the R and D department and the customer relations people were so close that it was possible to swap outgoing products?
If I was running a company that was building game consoles or 'high tech' stuff the R/D guys would work in an unmarked building on the outskirts of some town in Arizona. There would be lots of dish antennas, razor wire fences, and rabid guard dogs with names like "pookie" and "raul." Lastly, all one million dollar engineering prototypes would not go out UPS. There great and all, and I'm sure they do great with liver transplants and stuff, but this is my prototype. Hell, I wouldn't even trust the NYPD. My delivery people would be the Secret Service.
They rent out, don't they?
If not I'll call the Swedish Ski Commandos of Doom and Questionable National Alligence.