Raspberry Pi Gains a New 12.3-Megapixel Camera and Interchangeable Lenses (betanews.com) 66
The Raspberry Pi can be used for all sorts of maker projects, and the foundation has offered camera modules for it since 2013, adding vision-related functionality. The first module was a modest 5-megapixel affair that was eventually replaced by an 8-megapixel Sony sensor four years ago. Today, sees the arrival of a new much higher 12.3 megapixel quality camera, and a range of interchangeable lenses. From a report: The new camera is compatible with all Raspberry Pi models -- from Pi 1 Model B onwards -- with the exception of early Pi Zero boards. The camera is available to buy from today for $50.
Finally... (Score:2)
Re:Finally... (Score:4, Informative)
Use the new camera in place of the old in this sophisticated yet achievable microscope project created by Yuksel Temiz of IBM Research - Europe in Zurich, Switzerland. Links for exploration and learning (a good read even if you don't plan to build the project):
Yuksel Temiz' article "Build A Sophisticated Microscope Using Lego, 3D Printing, Arduino's, and a Raspberry Pi" published on the IEEE Spectrum website (includes video overview of construction and operation) https://spectrum.ieee.org/geek-life/hands-on/build-a-sophisticated-microscope-using-lego-3d-printing-arduinos-and-a-raspberry-pi [ieee.org]
GitHub MicroscoPy project resources https://github.com/IBM/MicroscoPy [github.com]
IBM Research article "IBM open sources $300 fully-functional LEGO microscope design" published on Medium.com https://medium.com/@IBMResearch/ibm-open-sources-300-fully-functional-lego-microscope-design-248a6cdc81bf [medium.com]
Consumer sensor still... (Score:2)
Re:Consumer sensor still... (Score:5, Interesting)
Being that the Pi is a $30 computer. I wouldn't be so picky.
What I feel we are really missing right now, it the ability for us to make "White Box" Carrier supported Cell phones.
One of the unsung hero of the 1990's tech boom, was the white box PC. Basically a custom made PC with off the shelf component. So a lot of people made their own PC's. But also a lot of small companies at the time were able to make a good living selling PC's to the market. Because all they need are the parts and they assemble them in a small building to order and mark up the prices by 20% and you had a good business model.
With Laptops and Cell phones, they are no real white box equivalent as all the parts seem to be custom made for every model. If Raspberry Pi provided cell phone kits with say Android OS, that is supported by major carrier I could a resurgence in small electronics companies again. Selling custom phones for businesses, say with that crazy port that hooks up to machinery, replaceable batteries, or without features such as cameras or texting. So they serve their purpose.
Re:Consumer sensor still... (Score:5, Insightful)
Re: (Score:2)
I've thought of putting together a "laptop" from a portable screen, keyboard, mouse, and a raspberry pi. with a USB battery pack. It would be more bulky than a standard laptop, but would have the advantage that you could upgrade each individual component as necessary when new things come out. I don't think a raspberry pi is quite powerful enough for a full laptop replacement, and a lot of the more powerful X86 SBC options couldn't run of a generic USB Battery. I'm confident that in the future this will be
Re: (Score:2)
https://www.theverge.com/circu... [theverge.com]
https://hackernoon.com/raspad-... [hackernoon.com]
Re: (Score:2)
I'm aware of projects like this. I was just looking for something less integrated, so any part can be replaced with another part depending on the task. Also, the Raspberry Pi isn't really powerful enough yet to fullfill a real laptop role yet. Maybe another couple generations and it will be there.
Re: (Score:2)
You can do anything you like, from designing and soldering your own board to buying one of those things pre-made. There are lots of different parts available, the plans are open source if you want to use them, and you can 3d print the case. If you want more power, there are lots of system on a chip ICs and single board computers you could use. This one has a Core M3 and comes with Windows 10 preinstalled: https://www.lattepanda.com/pro... [lattepanda.com]
Very few people will be happy with a build-your-own mobile device thou
Re: (Score:2)
Go ahead. There are lots of cellular modules you can cobble together into a cell phone if you want. Here are some that are already set up for arduinos:
https://www.arduino.cc/en/Guid... [arduino.cc]
https://create.arduino.cc/proj... [arduino.cc]
Re: (Score:3)
Being that the Pi is a $30 computer. I wouldn't be so picky.
You know this has been the excuse since the Pi was released but it's worth remembering that in 2020 the Pi is a $30 quad core computer with up to 4GB DDR4 RAM, 5GHz wifi, BLE 5.0, USB 3.0, Gigabit Ethernet, a GPU with multi-monitor support (dual 4kp60), and hardware H.264/H.265 CODECs.
What was once a toy is now actually something that is incredibly capable in many situations and quite capable of being used in some higher end applications. As it is it outperforms most embedded PCs on the market.
Re: (Score:2)
Re: (Score:2)
Re: (Score:1)
It is $100 camera added onto $30 'puter. Fun, but steep one.
Takes C and CS mount lenses (Score:2)
Pretty interesting, I had never heard of this mount before that apparently there are quite a few lenses for [bhphotovideo.com].
Re: (Score:1)
It is the standard mount on security cameras and other embedded systems. You won't find a DSLR that uses them natively.
Re: (Score:1)
Re: Takes C and CS mount lenses (Score:2)
Re: (Score:2)
I own three Micro 4/3 cameras and a dozen lenses. Image quality is inferior to current new DSLRs and has an overlapping range with APS-C (better than some, but the top end of APS-C is superior).
It is however better image quality than the DSLRs you could buy 10 years ago, and indeed is perfectly usable for almost any print situation. I wouldn't go more than 6 feet wide on anything intended to be looked at close up.
What makes M43 continue to be a popular niche is small light cameras (although the larger M43 c
Re: (Score:1)
No, it isn't.
C and CS mount is a screw mount. MFT has a bayonet mount. It is possible to get an adapter to fit C mount lenses to a MFT camera. It is also possible that a 4/3 sensor is used in a camera that uses a C mount, but that isn't a MFT camera.
Re: (Score:2)
Already out of stock (Score:1)
Re: (Score:2)
Can't really say I'm surprised they're sold out.
Where I can see a real market for this is in Third World security systems. Most companies in the developing nations can't afford a system with a reader module that costs $400 and cameras that start at $500 that need a $1500 back end so they have installed the worst Chinese piece of shit systems you can imagine, which break almost immediately and have no support. Recently there have been Pi-based reader modules coming on the market for under $100, with an ade
What do megapixels have to do with quality? (Score:5, Interesting)
> 12.3 megapixel quality camera
What do megapixels have to do with quality? Is there a comma missing there? A 12.3 megapixel, quality camera is a lovely thing. But in general, at a given price point more megapixels = lower quality due to smaller individual pixel sensors that are more sensitive to noise. You can resolve finer details, but thanks to the noise even if you resample to a much lower resolution the quality will be far lower than if you had just taken the photo with a lower resolution camera.
Re: (Score:1)
Don't be a dick. Everyone else here understood that sentence just fine.
Re: What do megapixels have to do with quality? (Score:2)
Not at all. I came here to comment exactly the same thing. Mod grandparent up, please!
Re: What do megapixels have to do with quality? (Score:1)
Sure the spec wars are dumb, but thinking that megapixels are unrelated to quality is equally stupid.
Re: (Score:2)
I'm new to the Raspberry universe.
I'm trying to think of a use case for this 12.3 megapixel camera. I'll want a built-in camera for video chatting, conference calls, etc. But I don't see why I would need one that would show my every facial blimmer and pimple to the world. That's just TMI; more information about me than I really care to share.
More important to me is making it as easy as possible to transfer images from my stand-alone cameras to my RPi4. My EDC camera is now a Fuji XP (mostly because it is
Re:What do megapixels have to do with quality? (Score:5, Insightful)
There are various projects that work with image recognition, like license plate scanners and so forth, microscope projects, and so forth. There are also more mundane uses, such as home security, that nevertheless could benefit from a selection of lenses.
A good quality camera at a low price point and especially with a variety of supported lenses should make the Pi much more proficient in a wide spread of specialized tasks that people are already putting it to.
ISO support is much better (Score:1)
I'm trying to think of a use case for this 12.3 megapixel camera.
It's not just the resolution, the whole sensor architecture is just way better - so even if you didn't use that full resolution, the image quality would be much better.
That's especially true in low light, which is something you really need in a video conferencing camera - used mostly indoors in poor lighting.
Re: (Score:2)
> I'll want a built-in camera for video chatting, conference calls, etc. But I don't see why I would need one that would show my every facial blimmer and pimple to the world.
So you're not into machine learning. With high-res and low-light you can measure heartrate, bp, BAC, detect cancers, and more.
With $120 in parts and FDA rent-seeking licenses, medical device manufacturers can sell these for $20K. Open source people can have them for $120.
In aerospace you can inspect welds well enough to avoid the u
Re: (Score:2)
I thought the point of ultrasound was that you can't see internal welding voids from the surface, how would a better camera help?
Re: (Score:3)
But I don't see why I would need one that would show my every facial blimmer and pimple to the world.
The benefits of higher pixels extend beyond resolution.
- Pixel binning for better colour with a bayer matrix causes your resolution to drop 4x.
- Noise reduction works better at higher resolution, and downsampling the end result produces an amazing quality image.
- Wide angle lenses with projection unwrap algorithm and selection algorithm can make for a "follow cam".
- Maybe your pi watching a birds nest far away.
Am I missing something?
When talking about a Raspberry Pi everyone is missing something. Even I am. I'm sure someone will
Re: (Score:2)
The sensor is larger, and the individual pixels are larger than the previous camera.
Re: (Score:2)
You can resolve finer details, but thanks to the noise even if you resample to a much lower resolution the quality will be far lower than if you had just taken the photo with a lower resolution camera.
Really? How do you get that? I'd assume that the same number of photons are falling on the sensor lens, and the same number of photons are being picked up; they're just getting divided into smaller buckets. Therefore a resampling algorithm should give *exactly the same* answer as just taking the photo with a lower resolution camera. You obviously believe that information is being lost somewhere. Could you spell out precisely where? -- do you think that higher resolution sensors discard more photons? do you
Re: (Score:2)
By reading:
1.55m × 1.55m pixel size -- double the pixel area of IMX219
Re: (Score:2)
> the same number of photons are being picked up; they're just getting divided into smaller buckets
Exactly. Fewer photons per bucket, but *not* less electrical noise per bucket. The smaller the bucket, the more noise you get per photon collected.
Or to put it another way - the more buckets you have, the more total noise you collect, while the signal (total amount of photons) stays roughly the same.
Re: (Score:2)
The smaller the bucket, the more noise you get per photon collected. Or to put it another way - the more buckets you have, the more total noise you collect, while the signal (total amount of photons) stays roughly the same.
Presumably the noise is random, though, and hence would be reduced by averaging over buckets? Or are you also indicating that the noise is correlated between buckets?
Re: (Score:2)
Noise can (on average) be reduced by averaging over buckets, but not nearly as much as by just using one bigger bucket to start with. (and the nature of random noise means that averaging it doesn't necessarily reduce it.)
As a somewhat extreme example, think of it like this:
One big bucket might on average be filled 10% by noise, leaving 90% of their capacity for signal ( signal-to-noise ratio of at most 9 )
While 4 small buckets will be filled by roughly the same amount of noise, but their total capacity is 1
Re: (Score:2)
If I understand right, you're saying that buckets might "max out" due to noise, and this will naturally lose information, and naturally it's more likely that a small bucket will max out due to random noise than a big bucket. That makes sense. Thanks!
Re: (Score:2)
Not quite. Generally speaking you never want the buckets to "max out" - that saturates the sensor and loses data, so you adjust your lighting or exposure to avoid it.
It might help to think of each physical sub-pixel-bucket as two separate buckets that can only be measured by weighing them both together.
Bucket S holds signal - that could be anywhere from 0% full (for a perfectly black pixel) to 100% full (for a perfectly bright white pixel at maximum exposure)
Bucket N holds noise - that could also be anywhe
Re: (Score:2)
In general, the size of the noise bucket remains roughly constant (based on manufacturing quality), while the size of the signal bucket scales with the size of the pixel sensor.
The noise from the smaller buckets signals will tend to average out some, but it doesn't cancel itself out (how could it, it's random). It does focus the probability distribution more towards a gaussian distribution than the original uniform random distribution, but that is offset by the fact that the distribution now spans 4x the original range.
I see. Thanks for your patience in explaining this to an internet stranger. I understand "size of the noise bucket remains roughly constant", so ten small pickets will have 10x the amount of noise as a single large pixel. That sounds like the key explainer for me. Your sentence "your distribution now spans 4x the original range" is saying that the stdev of the random variable is larger.
Re: (Score:2)
You're welcome.
>"your distribution now spans 4x the original range" is saying that the stdev of the random variable is larger.
Pretty much, though I was thinking of min-max. Keep in mind that stddev *only* applies to Gaussian distributions. Though now that I think of it a Gaussian distribution is probably a much better approximation for noise level than the uniform one I assumed.
Re: (Score:2)
But in general, at a given price point more megapixels = lower quality due to smaller individual pixel sensors that are more sensitive to noise.
Allow me to pick nits.
I'm far from an expert but expect total "quality" correlates more directly with sensor size than price. If I'm gathering 1 cm^2 of light, I can break that into 1 million very sensitive pixels or 10 million less sensitive pixels. Double the sensor size and both the 1m and 10m pixels will be more sensitive.
Anyway, depending on what you're trying to do, more resolution might be what you need, not higher signal:noise. I mean, take it to the limit: a one pixel sensor. Probably really sensit
Re: (Score:2)
I agree size is probably the biggest factor - specifically the size of each individual pixel detector - the bigger the pixel, the better the signal-to-noise ratio.
But price tends to correlate pretty well with size, though I'll admit "equals" was overstating things.
Re: (Score:2)
Thanks. Fair enough observation. I was going to also mention that price changes over time but felt that was being too picky.
So, does price not account for number of pixels, or does it really not matter? I would have thought for the same area, 2x pixels would cost more than 1x if nothing else than because of reduced yield. But I haven't shopped CMOS sensors ever.
Re: (Score:2)
I think that there is a cost premium for using cutting-edge lithography techniques for engraving extremely small details (e.g. New 5 nm Process!) that can only be produced in new facilities that are under high demand (and usually higher failure rates).
However,(don't quote me on this), once you get away from cutting edge lithography I believe pretty much all other costs scale primarily with area. I believe the lithography process itself is akin to using a high-power transparency projector to burn details in
Re: (Score:1)
> 12.3 megapixel quality camera
What do megapixels have to do with quality? Is there a comma missing there? A 12.3 megapixel, quality camera is a lovely thing. But in general, at a given price point more megapixels = lower quality due to smaller individual pixel sensors that are more sensitive to noise. You can resolve finer details, but thanks to the noise even if you resample to a much lower resolution the quality will be far lower than if you had just taken the photo with a lower resolution camera.
Mod parent down because it is stupid.
It reaches a mistaken conclusion because it assess utility of higher sensor pixel counts solely under the condition of an unrepresentative case of poor usage. In fact, overcoming shot noise using higher light levels or temporal averaging preserves spacial resolution! At the same prince point. If you solve your image noise problems by turning the lights up, opening up the camera aperture, or using a longer exposure time/frame, then more pixels really do give you higher
Re: (Score:2)
Yes, there are lots of ways of overcoming poor quality sensors - assuming of course that they're actually an option in a specific use case (e.g. better lighting isn't an option if you specifically want low-light photography, longer exposures aren't an option for moving scenes, etc.)
Also, please not that I said nothing about the *utility* of the sensor, only it's *quality*, which generally refers specifically to the signal-to-noise ratio (and possibly color accuracy, assuming it's not monochrome, since the s
Re: (Score:2)
A 12.3 megapixel, quality camera is a lovely thing.
English may be my second language but I find it much harder to parse what you wrote even though I understood TFS to mean what you are trying to say.
Re: (Score:2)
I agree, and if their intent was (as it seems to be from the article) to say that it was a quality camera with 12.3 megapixels, a better phrasing would probably be
A quality12.3 megapixel camera
with or without the comma.
However, in the original order, and without the comma, the phrasing implies that 12.3 megapixels is a statement about the quality of the camera. And since it's a common misconception that more megapixels = better quality, I felt it was worth drawing attention to the mistake. Especially since the opposite is generally tr
Re: (Score:2)
But in general, at a given price point more megapixels = lower quality due to smaller individual pixel sensors that are more sensitive to noise. You can resolve finer details, but thanks to the noise even if you resample to a much lower resolution the quality will be far lower than if you had just taken the photo with a lower resolution camera.
That has been proven false for many years for anything other than recording raw data. Pushing higher megapixels allows the image algorithms to work in a more ideal scenario, with the most ideal scenario is to have 4x more pixels than the smallest resolvable detail of the lens. In that scenario noise reduction and processing algorithms produce a better image than a large sensor was ever capable of and without affecting the image quality negatively. There's a reason why high-end camera manufacturers are pushi
Re: (Score:2)
The situation has improved, but it doesn't change the basic fact that more pixels is only better if you can keep the signal-to-noise ratio low enough that there's a net increase in accurate information. For a high-quality (probably quite large) sensor, that's absolutely the point - the raw signal to noise ratio is so good that "noise" from the optics themselves rivals the electrical noise from the sensor.
But for a given sensor size and production quality, doubling the number of pixels roughly halves the si
The sensor specs are awsome (Score:5, Informative)
Full resolution @60 frame/s (Normal), 4K2K @60 frame/s (Normal), 1080p @240 frame/s
https://www.sony-semicon.co.jp... [sony-semicon.co.jp]
Re: (Score:1)
1080p240 is game changing in the field of low cost computer vision...
Re: (Score:2)
Wow, I would be happy with D1 at 15 fps (which is better than a lot of security cameras costing hundreds of dollars).
IR cut filter integrated (Score:2)
It would be even nicer if they made a version of this without the IR cut filter. The existing Raspberry PI NOIR camera is one of the cheapest infrared cameras on the market, but it's limited by the absence of a lens mount.
https://www.raspberrypi.org/pr... [raspberrypi.org]
Re: (Score:3)
Specs: 4056(H) x 3040(V) (Score:2)
Resolution: 12.3 megapixel = 4056(H) x 3040(V)
Capture rates:
60fps@12Mpix 10bit
60fps@4K2K 10bit
240fps@1080P 10bit
Re: (Score:1)
Sensor Size (Score:2)
Re: (Score:2)
7.9 mm. It's right at the top of the Raspberry Pi announcement page, which is the first link in the article. Or type Pi High Quality Camera into Google.
Stereo vision, simply? (Score:2)
I want to do stereo vision with synchronized images. I know the cm3 (compute module) can take two cameras at the same time, but I don't want to design my own circuit around it. Does anyone have experience with doing it any other way, e.g. having two RasPis properly syncrhonized somehow (GPIO or NTP or whatever) so that they trigger their camera captures at exactly the same time?
home security camera ? (Score:3)
I wonder what it'd take to get a home security camera built now. Seems all the modules are there. How about electricity? I can run a wire to the RasPi, but can't put the usual pretty large power adapter near. Is there another way to power the thing?
Re: (Score:3)
At Wall-Plug Insert -> Power-Over-Ethernet -> At Pi extract Power from Ethernet
There are shields for the PI on the market - not expensive.
Re: (Score:3)
True, that's a neat idea, and I don't need a WLAN interface for the RasPi either in that case. Sweet.