Support NeoGAF

[Oppo]

It's quite small, isn't it? By those dimensions. Equivalent to a little over 7 inches long.

Really not much more than a doubled up Eye, with better processing. Which as pointed out can't handle darkness all that well – need much bigger lenses and better gear overall for that. Then again the Move stuff glows for a reason.

I don't think this is after the Kinect style IR body tracking though. This seems more for doing the "usual" camera duties while providing some extra depth cueing to the console for some gaming applications. The Move stuff is very accurate, but the weak link was always the Eye camera, so I assume the fact that this looks like a "bar" incorporating two lenses and sensors (but no IR or time-of-flight style stuff) is more a function of necessity than anything else.

Remember that the Kinect system is largely a library with skeletal tracking and all that other nonsense it needs to work in any semi-reasonable way. (ignoring when "Kinect" also means voice, of course.)

Also they are fools, fools I say if they do not call this thing the "Four Eyes".

 

[phosphor112]

from 10ft away?

He's talking about the time it takes to process the image. It can do it at 120fps. We aren't talking about how far it can do that.

So can my phone camera. But it doesn't, unless you want to build an app for me.

Yes... and we're still talking about image processing latency. Sony has 120fps modes on PS Eye.

Curious, do you have a source that it'll be packed in? I kept hearing Sony call it a "peripheral". Sure the DS4 has support with it's light bar (smart decision), but as far as I could tell it's a peripheral that you'll need to buy separately.

I was under the impression it came with it.

That's not finger tracking, that's basic shape extraction and analysis.
Without knowing the developer of that demo, I can suggest an easy way to do a similar one :
- extract pixels in a given color range (using histogram models of skin color)
- labelize individual blobs
- for each blob : extract center and main directions (principal component analysis), draw a cross and two lines.

It's a nice and easy demo to do (all the previous operations are common functions in image processing libraries like OpenCV, there may even be tutorials for that). But it's not finger tracking because :
- it can't tell a finger from any other object of similar color
- it merges all neighbouring pixels in asingle object : it "works" for an open hand, but a closed fist will register as a single misshaped finger.
- it cannot estimate the articulation of a finger. Even if one is localized, it will have only one direction.

Not sure how onQ did it, but here is a tutorial on doing something probably similar.

http://createdigitalmotion.com/2009...webcam-best-cam-for-vision-augmented-reality/

 

[xxracerxx]

Can we all agree that it looks like crap?

http://www.noreset.net/blog/wp-content/uploads/2013/02/pseye4.jpg[IMG]

I'm not sure I want that bulky box on my tv, I expect better from Sony.[/QUOTE]

No worse than Kinect, in my opinion. Personally I do not like anything around my television, be that Wii sensor bar, Kinect, EyeToy, etc. Bleh.

 

[JaggedSac]

Silly to put that glowing bar on all controllers if it isn't included with all the consoles.

 

[xxracerxx]

Silly to put that glowing bar on all controllers if it isn't included with all the consoles.

How else am I to know which controller is mine?

 

[onQ123]

What you're not mentioning is that the output of kinect is a tracked skeleton at 30 fps. The output of PS4eye is two raw image streams, that need to be processed for any depth measurement.

no kinect outputs a point cloud that also has to be processed to get that skeleton tracking.


the PS4 could also have a chip that converts the depth info into skeleton tracking.


I'm sure that's the reason for a special PS4Eye port on the PS4.


from the 3D Sound patent it shows that what looks to be the PS4 has chips for head tacking & device tracking.

 

[Oppo]

the PS4 could also have a chip that converts the depth info into skeleton tracking.

I'm sure that's the reason for a special PS4Eye port on the PS4.

yep. that AUX port is sort of a giveaway that there is a dedicated hardware component to processing that stuff.

 

[quetz67]

Comparing Kinect and Stereo Vision is like comparing apples to oranges as they both you different ways to accomplish same tasks. Time will tell which is better overall, but I think they both have their strengths and weaknesses.

So Kinect has this IR tech that kinda replaces time-of-flight but still emits light while the new PS Eye is just a stereo camera - right?

With enough processing power a stereo camera is usually quite good for depth perception, but light situation is very critical. Lets say there is a huge store front window behind the players with sunlight shining in it will probably fail completely.

 

[joesiv]

Silly to put that glowing bar on all controllers if it isn't included with all the consoles.

Actually it's genius because developers will know that the barrier to entry is a lot less for their players. They also don't need to support some strange control scheme (ala move), since they can retain the standard DS control layout with layers of move support. If sixaxis is used, the move support could just aid much higher quality tracking, and everyone would have it once they get the peripheral.

Also people have talked about the light bar lighting up based on gameplay, if you're health is low it could glow red (in a dark room it'll probably illuminate the room red).

 

[Domino792]

so Dance Central 4 on the PS4?

 

[tinfoilhatman]

Anyone that thinks a "depth camera" can track 6 people's full skeletons in ANY lighting situation is drinking the kool-air........sorry folks not in a million years.

 

[joesiv]

yep. that AUX port is sort of a giveaway that there is a dedicated hardware component to processing that stuff.

not really, it could just be a USB3.0 connection with a proprietary connector (simliar to OG Xbox with it's controller ports) to make things dummy proof (and to make sure you don't use it for anything else). USB3.0 is more than fast enough for 2x800p@60fps even uncompressed. Besides how does a connector tell you anything about dedicated hardware components? A connector is just that, a connector, which is connected to traces on a PCB, which could be going anywhere, maybe right into the AMD APU...

 

[Oppo]

With enough processing power a stereo camera is usually quite good for depth perception, but light situation is very critical. Lets say there is a huge store front window behind the players with sunlight shining in it will probably fail completely.

You are right but that's a bad example, as Kinect would fail there too. Sunlight awash with IR, etc.

Anyone that thinks a "depth camera" can track 6 people's full skeletons in ANY lighting situation is drinking the kool-air........sorry folks not in a million years.

You could do it right now of course, given enough processing. More than we will realistically get in the PS4 case but it's hardly impossible.

 

[iceatcs]

Can we all agree that it looks like crap?

I'm not sure I want that bulky box on my tv, I expect better from Sony.

At least not that massive as Kinect.

 

[Alx]

no kinect outputs a point cloud that also has to be processed to get that skeleton tracking.

But the skeleton is tracked at 30 fps by the console... "kinect" is the sensor and the processing libraries.

the PS4 could also have a chip that converts the depth info into skeleton tracking.

It already needs resources to estimate the depth info. Then it needs some more for the skeleton tracking. And both of them are harder to achieve from pure stereovision than the Primesense way. I don't say it's impossible, but it's very CPU-GPU expensive, and less reliable in changing conditions.

I'm sure that's the reason for a special PS4Eye port on the PS4.

I don't see the relation between connectivity and processing, to be honest... the port is just a plug. Kinect1 also has a special port on the 360 slim, but it's still USB protocol.

 

[onQ123]

Anyone that thinks a "depth camera" can track 6 people's full skeletons in ANY lighting situation is drinking the kool-air........sorry folks not in a million years.

Why not? it's sending out it's own IR so lighting shouldn't be a problem.

 

[AgentP]

Anyone that thinks a "depth camera" can track 6 people's full skeletons in ANY lighting situation is drinking the kool-air........sorry folks not in a million years.

How do six people fit into that FOV? All this talk about skeletal tracking is a red herring. You don't need all that to dance around. Hell, most of the time you could have the camera unplugged and people would still think they were playing a game.

 

[Unknown One]

Putting the light bar on the controller makes sense. It can be used for aesthetic reasons and gives people more incentive to buy the PSEYE if the price is right and have it be compatible from the get go.

I won't be getting the PSEYE unless it's actually useful, but I like the light bar idea.

 

[InaudibleWhispa]

Anyone that thinks a "depth camera" can track 6 people's full skeletons in ANY lighting situation is drinking the kool-air........sorry folks not in a million years.

In the case of Kinects depth sensor, it replies on IR so it works in ANY lighting situation by default unless the game requires an RGB feed which most don't. You can play Kinect Sports in the dark if you want. As for tracking 6 players, Kinect Party does that on the current Kinect.

All this talk about skeletal tracking is a red herring. You don't need all that to dance around. Hell, most of the time you could have the camera unplugged and people would still think they were playing a game.

In the case of Kinect nothing would move if it isn't tracking a person. If you block the camera and have someone play Kinect Sports, they'll bowl a ball and the Avatar will be motionless since the character follows your every move so in every case anybody would instantly tell if they aren't actually playing the game. As for Dance Central, limbs flash when you're doing something wrong and it couldn't accurately score you without tracking. Just Dance is an example of a game where tracking isn't very good and people can just dance around pretending they're playing, but in my experience it's less fulfilling and most people just ask me to put Dance Central on because they like the feedback.

 

[Oni Jazar]

PlayStation®4 Eye senses the color of DUALSHOCK®4’s light bar to judge the positions of multiple players, and set characters in the same positions or let a character talk to a particular player, enabling players to enjoy games in new way

That's kind of awesome. Have the character look straight at the player. Knows when you get up to make a sandwich and yells at you.

 

[JaggedSac]

All this talk about skeletal tracking is a red herring. You don't need all that to dance around.

You are correct, you don't need skeletal data for dancing, Dance Central doesn't use the skeletal system, they use the raw depth data. The raw depth data has information about which pixel relates to which player and they have their own algorithms to determine matching moves.

 

[RSTEIN]

That's kind of awesome. Have the character look straight at the player. Knows when you get up to make a sandwich and yells at you.

That does sound awesome. Such a simple thing yet so effective.

 

[MikeE21286]

Does anyone else feel like they held this back intentionally *just* to counter Kinect 2.0

 

[Domino792]

Does anyone else feel like they held this back intentionally *just* to counter Kinect 2.0

yea,i think they want Microsoft to be the ones to start the motion control talk so they can counter it.

 

[onQ123]

But the skeleton is tracked at 30 fps by the console... "kinect" is the sensor and the processing libraries.



It already needs resources to estimate the depth info. Then it needs some more for the skeleton tracking. And both of them are harder to achieve from pure stereovision than the Primesense way. I don't say it's impossible, but it's very CPU-GPU expensive, and less reliable in changing conditions.



I don't see the relation between connectivity and processing, to be honest... the port is just a plug. Kinect1 also has a special port on the 360 slim, but it's still USB protocol.

if you have a perfectly cutout 3D image of the player it's not going to be hard to know where the body parts are it will all work together.

 

[yurinka]

Regarding that if it's better 2D camera + IR vs 2 2D cameras discussion I have some questions:
-Which system do the humans use to perceive 3D?
-Which system do the pro 3D cameras use to shoot 3D movies?
-How the 3DTVs do work?

Does anyone else feel like they held this back intentionally *just* to counter Kinect 2.0

I think they didn't have time to explain everything on the conference, so tried to choose the new things in order to don't have the fanboys that they copied Microsoft's social, streaming and 2nd screen focus. Because I'm sure they're working in the same.

Core gamers typically don't want to see motion sensor stuff in these public conferences, so I'd bet Sony will show the camera stuff in the GDC and then to split their big games through E3 (games focused to US/worldwide), Gamescon (games focused to EU) and TGS (games focused to Japan).

From Sony's side I'd like to see stuff like putting your face to the game main char, to log in with face recognition, game characters talking at the players and looking at him (just tracking the pad position) or 3D object rendering for Media Molecule's project. And 3D video chat.

 

[HoodWinked]

make the hardware, figure out the software later.

 

[joesiv]

Regarding that if it's better 2D camera + IR vs 2 2D cameras discussion I have some questions:
-Which system do the humans use to perceive 3D?
-Which system do the pro 3D cameras use to shoot 3D movies?
-How the 3DTVs do work?

- Humans use 2 2D cameras most call eyes. Humans also use what most call a brain, it's like an advanced computer much more powerful than the PS4, and it does it all with zero lag while simultaneously recording in long term memory for further object recognition and processing.
- 3D movies are either: 2 2D cameras, 1 2D camera and a 3D sensor, 1 2D camera and lots of processing and manual processing.
- 3D TV's are an output device and require no processing, so I don't know how this is even relevant to the input device such as Kinect/PSEye

 

[onQ123]

Regarding that if it's better 2D camera + IR vs 2 2D cameras discussion I have some questions:
-Which system do the humans use to perceive 3D?
-Which system do the pro 3D cameras use to shoot 3D movies?
-How the 3DTVs do work?

But the answer to these questions does not answer the question of which will be better on a console.


The Answer would be Stereo Camera but how it's done isn't as important as the end result & that's something we will have to wait and see.


1 thing that I do know for sure is that I would rather have a camera that's tracking at 60FPS & above than to have one that's limited to 30FPS.

 

[Number45]

Hmm, I wonder if the design of the PS Eye carries any hints to the possible design of the PS4?

 

[tha_devil]

Hmm, I wonder if the design of the PS Eye carries any hints to the possible design of the PS4?

6 gamecubes ducktaped together.

Now the only question is if it will get a a handle or not.

 

[Alx]

if you have a perfectly cutout 3D image of the player it's not going to be hard to know where the body parts are it will all work together.

That's a veeery big "if". And more than the cutout, you would need a full depth map for body part tracking, especially in Z direction.

 

[yurinka]

- 3D TV's are an output device and require no processing, so I don't know how this is even relevant to the input device such as Kinect/PSEye

I think it's fair to think that they may feature 3D video chat with it. In this case, the people watching their 3DTVs on the other side will receive 2 2D images, one for each eye.
So I think that for this PS4 Eye will be better because it's recording exactly this so wouldn't need to process anything.
Kinect would need additional processing to create an image for each eye from the 2D single image + depth, which would be less precise 3D (lower resolution in its depth tracking sensor, recreated images from a 2D single one instead of 2 2D images shooted with the proper distance).

That's a veeery big "if". And more than the cutout, you would need a full depth map for body part tracking, especially in Z direction.

Considering their great results tracking bodies with a 2D camera, I'm pretty sure that it will help them a lot. I bet they will have full body tracking, more precise than Kinect 2 and without problems of being sitting in a sofa. It would need calibration asking you to do the 'T' position.

 

[Alx]

But why would you want to do 3D video chat ?

 

[JaggedSac]

Can someone link me to good examples of stereo camera depth data? I have been looking it up and have found nothing impressive.

 

[AgentP]

I

In the case of Kinect nothing would move if it isn't tracking a person. If you block the camera and have someone play Kinect Sports, they'll bowl a ball and the Avatar will be motionless since the character follows your every move

I wasn't being literal. I can imagine some kids playing Kinect Stars Wars for an hour and finding out it was just a video. On screen actions and real life actions are loosely correlated in these casual games.

 

[Afrit]

PlayStation Eye works pretty good in low light I did this video only using the light from my laptop.

125 FPS hand tracking PlayStation Eye (Introducing the Leap?)

That's pretty cool.

That's a veeery big "if". And more than the cutout, you would need a full depth map for body part tracking, especially in Z direction.

Aside from using a stereo camera for depth tracking. There is good advantage of having 2 cameras instead of one for a Move game. By using one camera for tracking @ 120fps while the other camera will capture the scene in 720p HD, then combine the feeds for augmented reality. This way it won't be a grainy video on the screen like the current PSeye, while having a prety decent tracking.

 

[DBT85]

I'll be interested to see what deveoplers make of it if Sony do pack one in with every box. That certainty of knowing you have EVERYONE on the platform owning one means they can attempt more with it.

Also the most important question is this... Will we get an online multiplayer of Sports Champions Table Tennis FFS!

 

[BeEatNU]

Pretty sure it's just an enhanced version of Move.

exactly, I do not know why people are digging for nothing

 

[MightyHedgehog]

But why would you want to do 3D video chat ?

Yeah, that'd be awkward talking to each other with 3D glasses on.

 

[Oppo]

Hmm, I wonder if the design of the PS Eye carries any hints to the possible design of the PS4?

I was wondering this myself. But then again the Eye doesn't really look anything like a PS3, other than being black.

 

[onQ123]

But why would you want to do 3D video chat ?

Because Chatting in 3D with hot chicks is cool!

Pseudo-3D Video Conferencing with a Generic Webcam

 

[Green Yoshi]

Kinect has a CPU and was motorized, I'm assuming this is basically to it what those Chinese imitation consoles are to the real deal.

Given how useful the Kinect actually IS in practice (this gen anyway) though it's not exactly something I'm crying over.

I think PS Eye will be a lot cheaper (40$ or so) because it has no motor.

 

[onQ123]

I think PS Eye will be a lot cheaper (40$ or so) because it has no motor.

The new Kinect isn't going to have a motor either.

 

[onQ123]

Can someone link me to good examples of stereo camera depth data? I have been looking it up and have found nothing impressive.

3D stereo image processing: rectification + depth map

 

[StealthGoblin]

Don't be surprised to see Sony have all of the OS navigation (inc. voice commands) for PS4.

 

[JaggedSac]

3D stereo image processing: rectification + depth map

That is pretty decent. Thanks.

 

[onQ123]

That is pretty decent. Thanks.

HD & 60FPS just like the PS4Eye, they used a FPGA which is good for things like Image processing, if Sony is smart they will have chips specially made for image processing that will make lite work of getting the depth map out of the 2 cameras & giving the PS4 a clean 3D image of the players to use for tracking.

 

[backbreaker65]

Question I was reading the pdf file on the PS4eye and found that it will have its own slot and not USB this time around "CONNECTION TYPE PS4 DEDICATED CONNECTOR (AUX CONNECTOR) what is the benefit of this as opposed to going back to USB?

 

[onQ123]

Question I was reading the pdf file on the PS4eye and found that it will have its own slot and not USB this time around "CONNECTION TYPE PS4 DEDICATED CONNECTOR (AUX CONNECTOR) what is the benefit of this as opposed to going back to USB?

that's something we would have to wait to get more info about, it could be just a USB 3.0 port made for the PS4Eye or it could have special hardware at the input that will help process the video as it comes in from the PlayStation4Eye.