Stereo Vision and Time of Flight setups for Face Scanning
- Thread starter onQ
- Start date
Sony weren't waiting on hardcore gamers to decide whether to bother or not. They needed to make that choice before PS4 was released, and get games in development. They didn't, because they weren't interested. As I've said all along, devs, not even first party ones, just aren't going to try anything new and worthwhile beyond party games with these things. Face customisation is as good as it'll probably get.
There are many ideas they could design that can appeal to the hardcore. A combination of controller and camera can work. The camera can compliment/enhance the main hardcore experience nstead of becoming the primary controller.
the camera as a feature is a problem when it increases price, sacrifices performance and games are targeting casual experiences.
The PS4 has the perfect ecosystem to offer something appealing. Stereoscopic cameras, a controller with a light bar and motion detection, and the ability to detect voice.
I think this might be the case. With FW 2.0 background removal is being done when you live stream which tells me that it's being down in hardware.
http://www.gamepur.com/news/16756-p...ere-are-list-hidden-features-and-changes.html
Or it's being done by the video capture+stream software. Note this says chroma keying, not background removal, so it has zero baring on the game in question. And chroma keying is very easy. In the video feed, for any pixels around a given value (RGB[0,255,0], or a hue of ~ 120 degrees), replace it with the alternative image. Nothing complicated at all, nothing warranting a second camera.
If anything this shows the opposite of your theory, as if background removal is available in hardware, why bother with a cheap-ass green screen technique?
Huh? I think that's just the websites interpretation of the background removal to explain it better.
By the way I'm sure that the video capture+stream is also being done using secondary hardware.
Why? They call it 'green screen' before explaining what chroma keying is.
Why use that as a description if in reality it's background removal based on depth without needing a coloured background?article said:
It's easily tested though. Someone with a PS4 fire up the streams and see what options there are for background removal.
A Google for Twitch suggests this is basically a Twitch feature and the PS4 FW would just be updating the Twitch software to support it, I expect. Presumably it'll allow a person to be overlaid over the game play without needing a whole box partition.
The videos in this link are hilarious. Feature is not ready for consumers.
http://www.polygon.com/2014/10/7/6939155/nba-2k15-face-scan-glitch-disaster-impossible
http://www.polygon.com/2014/10/7/6939155/nba-2k15-face-scan-glitch-disaster-impossible
I see mostly user errors with his attempts. like the bricks with all the different colors plus the white ceiling with all the light shinning from the side as the background, then he place the camera on top of the TV I'm guessing & not in front of it where it could use the light from the TV to get a better scan. he also never really calmed down.
How is having a brick wall and a white ceiling and windows a user error? Applications like this need to work in a variety of environments. Once you start telling people they have to modify their rooms, or move their tv into a different room, you know your application is destined to fail as a general consumer product. His head and face are clearly visible at all times, but it wouldn't even register for a second. It was 100 fail from the software. There were plenty of times he was still, the square were turn green and then immediately report that it had lost track of his head.
I'm also guessing these videos were made because he had already had many issues trying the feature before. Looking at his camera capture, I'm not sure you could ask for much better than that generally. His face and head are very clearly visible, and actually well lit.
It's the fact that he had it all working against him that makes it a user error. it might have worked if he lowered the camera angle so that the background didn't include the bright ceiling at the top of his head & bricks at the lower part.
It's user error no matter how good a product is some people are just not going to be able to use it right. Some people take bad pictures with $1,000 DSLRs while others take great pictures with a smartphone & the person with the DSLR will still blame the product instead of taking the time to learn what they are doing wrong.
Furthermore, regards the OP, background removal wouldn't have a problem with background scenery being bricks as it'd be removed. That could be a problem with a moving, handheld camera (although then the error message should be to keep the camera stationary) but when it's stationary, the background can be captured and removed (excepting of course this is a far from accurate process as I've repeated!). Failure to extract the head shows that PS4 Eye does not have a technical advantage for head scanning.
Also, the error reporting was telling the user he was using it effectively. If his face was supposed to fill the square, why was the frame turning green? If you want the face a certain size, you need a face template to align with. The choice of a square shows the software was trying to be more flexible and the algorithm would be adapting to the head size in the view. It's just not very good at it!
Also, the error reporting was telling the user he was using it effectively. If his face was supposed to fill the square, why was the frame turning green? If you want the face a certain size, you need a face template to align with. The choice of a square shows the software was trying to be more flexible and the algorithm would be adapting to the head size in the view. It's just not very good at it!
Why? 1) Background removal shouldn't be confused. 2) Face recognition is a commodity tech these days and there's nothing in the background that's suggestive of 'face' to confuse the system.
Except he used it as instructed. Was he told to cover up the background with a green screen? Did he persist despite never getting a green square? The square kept turning green telling him everything was ready, and then, without him changing a thing, it lost tracking. If the user's action work one moment and then, remaining the same, fail, that's clearly a software fault.
There is no way you can watch that vid and suggest the guy is using it poorly. He sits incredibly still with a well let face inside the square and it doesn't work for more than a second. It's best attempt it finally started to scan his face, but it failed at 0%. If you start asking people to make changes to their rooms when the video feed is already very good, then your application basically sucks.
As a photographer I can see a lot wrong with what he was doing.
He didn't need to make changes to his room he could have simple\y placed the camera about 12in from his face & aimed it in a way that his background didn't go from bright ceiling to brick textures. The camera lost his face as it adjusted to the bright ceiling & he was never really stable with the camera eye level & 6 to 12 inches from his face like the game ask.
This official demonstration from the dvelopers suggest you're well wrong and he was using it precisely as he was supposed to (contrary to the indications of the game that you should fill the square with your face).
Another one:
