STICKY: Foundation (old) Camera Board information

[jamesh]

Quick status report.

I've been working on command line demo app for the new module, and have finally got captures working OK. Video is also working but only at lo-res - (640x480). Hope to sort that tomorrow. The app allows you to change such things as AWB, ISO, resolution, JPQ quality, effects, metering mode, exposure mode etc. Looking pretty good. We still have some tuning work to do on the image quality - it's not as good as it should be.

But progress is now getting on nicely.

[tonyhughes]

Drool.

I haven't used any add-ons except an off switch hooked to a GPIO.

But the camera... well... that will be a bit of me!

[mk9900]

Not sure if this has been asked before, but will the cam be compatible with Motion (http://www.lavrsen.dk/foswiki/bin/view/Motion/WebHome) ?

[jamesh]

Not sure if this has been asked before, but will the cam be compatible with Motion (http://www.lavrsen.dk/foswiki/bin/view/Motion/WebHome) ?

At the moment we just have a command line app that grabs stills or captures video. We do no yet support /dev/video which is how Motion connects to the camera. So not at first. I expect someone will sort out a /dev/video driver at some point, I have no idea how to do that (but will look at it at some point).

[aTao]

Quick status report.

I've been working on command line demo app for the new module, and have finally got captures working OK.

samples?, samples, samples!, SAMPLES.

We still have some tuning work to do on the image quality - it's not as good as it should be.

But progress is now getting on nicely.

Ohh, not quite yet then

[jbeale]

"Pretty sure will work with the OpenMAX API, but there is also a custom API on the Arm which is easier to use."
-from http://www.raspberrypi.org/phpBB3/viewt ... 25#p220439

Is your command line demo app written in C, or some other language? I presume you are using the custom API?
Would it be useful to publish the API ahead of the hardware release, to help people get their infrastructure ready to use the camera when it comes? Or is it still in flux?

[jamesh]

The apps are in C and uses the 'mmal' shim over the top of OpenMAX. It's basically an effort by Broadcom to make the OpenMAX component system a bit easier to use. There is currently no API documentation on MMAL but it is fully doxygen commented and is supplied in the userland git repo, so you can make your own docs.

I'm seriously considering written a book/ebook on this subject though - MultiMedia programming on the Raspberry Pi or somehting like that.

[texy]

Out of interest, once the camera is available, how will they be distributed? Via RS/Farnell?

Texy

[jamesh]

Out of interest, once the camera is available, how will they be distributed? Via RS/Farnell?

Texy

I don't know - I knew there was something I meant to ask Liz at the birthday get together.

[gcrispin]

I've been following this for months and months... when will you take my money?

[ukscone]

Out of interest, once the camera is available, how will they be distributed? Via RS/Farnell?

Texy

Eben will stand on the top of Great Orme and scatter them in the winds to the 4 corners of the earth

[texy]

I had to look that up
Texy

[jamesh]

I've been following this for months and months... when will you take my money?

When it's ready. Or you can buy one that doesn't work. But I'm guessing that wouldn't be any good?

(soon btw)

[jbeale]

I have to admire people who can pronounce something like "Cyngreawdr Fynydd".

I can imagine some beta-tester interest in a camera that is only partly and/or sometimes working, especially if there is hope of a future downloadable update that would bring it to fully operational. But it sounds like the camera isn't quite there yet, so we'll have to be patient!

[jamesh]

I have to admire people who can pronounce something like "Cyngreawdr Fynydd".

I can imagine some beta-tester interest in a camera that is only partly and/or sometimes working, especially if there is hope of a future downloadable update that would bring it to fully operational. But it sounds like the camera isn't quite there yet, so we'll have to be patient!

We have a bunch of camera going out to testers (internal only, so don't ask) soon I believe. And it won't be long before release anyway, it's mostly there; the board has passed testing, I'm working on a easy (!) to understand set of apps, and we have a top notch camera guy doing some more tuning. He's already sorted a problems with denoise, and flickering AWB, and is looking at other stuff now. It's starting to look pretty good tbh.

[rjphipps]

Will the camera/drivers also support either RAW (i.e. non-demosaiced - e.g. DNG) output, or at least noncompressed (i.e. non-JPEG - e.g.TIF) output options? Also, what colorspace(s) will be supported (e.g. only sRGB, or a larger possibly native color profile)?

RJP

[redhawk]

What kind of software is coming with this camera would this enable video streaming over a web page served by apache??

Richard S.

[jamesh]

Will the camera/drivers also support either RAW (i.e. non-demosaiced - e.g. DNG) output, or at least noncompressed (i.e. non-JPEG - e.g.TIF) output options? Also, what colorspace(s) will be supported (e.g. only sRGB, or a larger possibly native color profile)?

RJP

Yes, the example code I've written can output the RAW Bayer data as JPEG metadata. Not sure of the exact Bayer format though, I'll need to check. I've also written a quick app that dumps the image as uncompressed YUV420 rather than JPEG.

[jamesh]

What kind of software is coming with this camera would this enable video streaming over a web page served by apache??

Richard S.

See previous post, but I have written three apps, one to take JPEG stills, one to take video and one to take YUV420. And Dom has got some instructions that can use the video recording app to stream over the network. Each app can apply various different effects to the data, and the first two have a demo mode that cycles though various options. When I have time I'll post the options available.

The apps I am sure will also be improved as time goes on, and since they will come with source code I hope lots of people will adapt and modify to their own specifications.

[jbeale]

Anyone working with raw images on Linux probably already knows this. But just in case not, be aware that Dave Coffin's open source program DCRaw http://www.cybercom.net/~dcoffin/dcraw/ can likely do what needs to be done to convert a raw still frame into a usable format. Just need to figure out the parameters for your specific device.

[bantammenace2012]

wrt. The Great Orme, perhaps the camera modules might be geo-cached as a method of distribution.

[JDWraspberryPi]

I would like to suggest command-line software that would open a file, capture video data from the Foundation Camera Board, write the data to the file, and close the file after receiving a termination signal. This method would would allow the user to interactively stop the video capture at any arbitrary time after starting the capture. A user-interactive session might look like the following example:

Code: Select all

$ vcap video.mp4 &
$ kill `ps | grep "grep" --invert-match | grep "vcap.sh video.mp4" | head -n 1 | awk '{print $1}'`;

Here "$ " represents a Linux command prompt, "vcap" is the name of the command, and "video.mp4" is the name of the file. The user would execute the first command to begin the video capture, wait for some time, and then execute the second command to end the video capture. The big deal here is that the software gives the user control over when to stop the video capture. Unfortunately, some command-line, video-capture commands do not provide the user with this capability.

This method would also allow the user to non-interactively stop the video capture after a predetermined time using a shell script like the following:

Code: Select all

#!/bin/bash
vcap $1 &
sleep $2 
kill `ps | grep "grep" --invert-match | grep "vcap.sh video.mp4" | head -n 1 | awk '{print $1}'`;

With this script the user could capture video to a file for a specified time interval by entering the following commands at a Linux prompt:

Code: Select all

$ chmod +x vcap.sh
$ vcap.sh video.mp4 60 &

Here "vcap.sh" is the name of the shell script, "video.mp4" is the name of the file, and "60" is the number of seconds in the specified time interval.

The vcap program would trap the termination signal, close the video-capture file, and exit gracefully. Alternatively, the program could achieve the same functionality by monitoring a file. When the user or a shell script modified that file in a predetermined way then the program would treat that as a termination signal, close the video-capture file, and exit gracefully. Of course the video-capture program could also write to stdout and the user could redirect the output to a file using a command like the following:

Code: Select all

$ vcap > video.mp4 &

This last method would give the user even more control and greater modularity for software developers. For instance the user could pipe the data through a "transcoder" program before saving it to the file.

Code: Select all

$ vcap | transcoder > video.mp4 &

Then any number of people could work on different transcoders without touching the code for the video-capture program. In the last example I am not sure how the transcoder would identify which bits of information were part of the data and which bits of information were part of the wrapper but this method would work well if it were actually possible.

[fastmapper]

I'm very interested in the camera board, but I do have some questions.

I've seen the question asked several times before, but I haven't seen it answered: what is the camera focal length? I need to know in order to determine what resolution and field of view I can expect from photos collected by the camera.

I am also interested to know: will the ARM processor have access to camera timing signals such as vertical sync or horizontal reference?

[poing]

I don't think it's known as a hard fact what focal length is used, but from what I've seen I'd guess it's equivalent to around 28-35mm on a 135 film SLR where the chip sends out 5MP. Keep in mind though that it's a $25 module including the lens so guessing the resolution based on these figures seems pointless, for the actual resolution depends on both the chip and the lens. You'd need actual measurements.

If you look at this time laps video @ 1.10 you can see what I mean (although I have no clue what settings were used to record the images): http://www.recantha.co.uk/blog/?p=3201

If I see that my first reaction is 'how do I get that lens off' (but hey, I'm spoiled). Guess I'll have to buy two and use a hacksaw on one of them

[RaTTuS]

.... Guess I'll have to buy two and use a hacksaw on one of them

Scalpel !