RFC: adding a flag to indicate a webcam sensor is installed upside down

RFC: adding a flag to indicate a webcam sensor is installed upside down

[Hans de Goede]

Hi all,

I have this Philips SPC 200NC webcam, which has its sensor installed upside 
down and the sensor does not seem to support flipping the image. So I believe 
the windows drivers fix this little problem in software.

I would like to add a flag somewhere to indicate this to userspace (and then 
add flipping code to libv4l).

I think the best place for this would the flags field of the v4l2_fmtdesc 
struct. Any other ideas / objections to this?

Regards,

Hans

--
video4linux-list mailing list
Unsubscribe mailto:video4linux-list-request@redhat.com?subject=unsubscribe
https://www.redhat.com/mailman/listinfo/video4linux-list

Re: RFC: adding a flag to indicate a webcam sensor is installed upside down

[Laurent Pinchart]

On Thursday 07 August 2008, Hans de Goede wrote:
> Hi all,
> 
> I have this Philips SPC 200NC webcam, which has its sensor installed upside 
> down and the sensor does not seem to support flipping the image. So I
> believe the windows drivers fix this little problem in software.
> 
> I would like to add a flag somewhere to indicate this to userspace (and then 
> add flipping code to libv4l).
> 
> I think the best place for this would the flags field of the v4l2_fmtdesc 
> struct. Any other ideas / objections to this?

More often than sensors being mounted upside down in a webcam, what I've noticed frequently is webcam modules being mounted upside down in a laptop screen. There is no way that I'm aware of to check the module orientation based on the USB descriptors only. We will need a pure userspace solution.

Best regards,

Laurent Pinchart

--
video4linux-list mailing list
Unsubscribe mailto:video4linux-list-request@redhat.com?subject=unsubscribe
https://www.redhat.com/mailman/listinfo/video4linux-list

Re: RFC: adding a flag to indicate a webcam sensor is installed upside down

[Hans de Goede]

Laurent Pinchart wrote:
> On Thursday 07 August 2008, Hans de Goede wrote:
>> Hi all,
>>
>> I have this Philips SPC 200NC webcam, which has its sensor installed upside 
>> down and the sensor does not seem to support flipping the image. So I
>> believe the windows drivers fix this little problem in software.
>>
>> I would like to add a flag somewhere to indicate this to userspace (and then 
>> add flipping code to libv4l).
>>
>> I think the best place for this would the flags field of the v4l2_fmtdesc 
>> struct. Any other ideas / objections to this?
> 
> More often than sensors being mounted upside down in a webcam, what I've noticed frequently is webcam modules being mounted upside down in a laptop screen. There is no way that I'm aware of to check the module orientation based on the USB descriptors only. We will need a pure userspace solution.
> 

Interesting, still in my case it can be told from just the usb id (philips 
luckily uses its own id's instead of generic id's). So I think in cases were we 
can tell it at the kernel level we should set a flag somewhere (and I believe 
the flags field of the v4l2_fmtdesc is the best place) to share this knowledge 
with userspace.

Then for laptops we will need a detection mechanism probably based on DMI 
strings (I feel hal is going to play a role here) and then check for flipx and 
y v4l2_ctrl's and if those are not present use software flipping (which I'm 
going to implement today for my philips cam).

Hmm, thinking more about the laptop problem, ideally hal would just do the 
v4l-ctrl calls itself, so that this happens only once (on plugin) and users can 
later override this. And then we would need some kinda hal boolean which libv4l 
can query called something like "needs software flip" for the other cases.

Regards,

Hans


p.s.

If someone can ship me a laptop which an upside down mounted cam I'll happily 
work on this issue :)

--
video4linux-list mailing list
Unsubscribe mailto:video4linux-list-request@redhat.com?subject=unsubscribe
https://www.redhat.com/mailman/listinfo/video4linux-list

Re: RFC: adding a flag to indicate a webcam sensor is installed upside down

[Jean-Francois Moine]

On Thu, 2008-08-07 at 12:37 +0200, Laurent Pinchart wrote:
> On Thursday 07 August 2008, Hans de Goede wrote:
> > Hi all,
> > 
> > I have this Philips SPC 200NC webcam, which has its sensor installed upside 
> > down and the sensor does not seem to support flipping the image. So I
> > believe the windows drivers fix this little problem in software.
> > 
> > I would like to add a flag somewhere to indicate this to userspace (and then 
> > add flipping code to libv4l).
> > 
> > I think the best place for this would the flags field of the v4l2_fmtdesc 
> > struct. Any other ideas / objections to this?
> 
> More often than sensors being mounted upside down in a webcam, what I've noticed frequently is webcam modules being mounted upside down in a laptop screen. There is no way that I'm aware of to check the module orientation based on the USB descriptors only. We will need a pure userspace solution.

Agree.

Having a horizontal or vertical inversion may exist in any webcam (and
tuner?), and may be the fact of wire inversion when assembling the
device. So, having a flag in the driver could not work with hardware
error.

If the sensor (or bridge?) may do H or V flip, we are done. If not, this
may be done by software, but in the userspace, i.e. at decoding time.

What I propose is to add a check of the control commands V4L2_CID_HFLIP
and V4L2_CID_VFLIP in the v4l library: if the driver does not have these
controls, the library silenctly adds them and handles their requests.
Then, at frame decoding time, the hflip and vflip may be given to the
decoding functions.

How do you feel it, Hans?

-- 
Ken ar c'hentañ |             ** Breizh ha Linux atav! **
Jef             |               http://moinejf.free.fr/


--
video4linux-list mailing list
Unsubscribe mailto:video4linux-list-request@redhat.com?subject=unsubscribe
https://www.redhat.com/mailman/listinfo/video4linux-list

Re: RFC: adding a flag to indicate a webcam sensor is installed upside down

[Hans Verkuil]

On Friday 08 August 2008 11:10:36 Jean-Francois Moine wrote:
> On Thu, 2008-08-07 at 12:37 +0200, Laurent Pinchart wrote:
> > On Thursday 07 August 2008, Hans de Goede wrote:
> > > Hi all,
> > >
> > > I have this Philips SPC 200NC webcam, which has its sensor
> > > installed upside down and the sensor does not seem to support
> > > flipping the image. So I believe the windows drivers fix this
> > > little problem in software.
> > >
> > > I would like to add a flag somewhere to indicate this to
> > > userspace (and then add flipping code to libv4l).
> > >
> > > I think the best place for this would the flags field of the
> > > v4l2_fmtdesc struct. Any other ideas / objections to this?
> >
> > More often than sensors being mounted upside down in a webcam, what
> > I've noticed frequently is webcam modules being mounted upside down
> > in a laptop screen. There is no way that I'm aware of to check the
> > module orientation based on the USB descriptors only. We will need
> > a pure userspace solution.
>
> Agree.
>
> Having a horizontal or vertical inversion may exist in any webcam
> (and tuner?), and may be the fact of wire inversion when assembling
> the device. So, having a flag in the driver could not work with
> hardware error.
>
> If the sensor (or bridge?) may do H or V flip, we are done. If not,
> this may be done by software, but in the userspace, i.e. at decoding
> time.
>
> What I propose is to add a check of the control commands
> V4L2_CID_HFLIP and V4L2_CID_VFLIP in the v4l library: if the driver
> does not have these controls, the library silenctly adds them and
> handles their requests. Then, at frame decoding time, the hflip and
> vflip may be given to the decoding functions.
>
> How do you feel it, Hans?

I'm a different Hans :-), but I was thinking along the same lines. 
Having the v4l lib implement HFLIP/VFLIP if the driver doesn't seems 
like a good idea.

I also think that the UPSIDEDOWN flag doesn't belong to v4l2_fmtdesc 
since this is a global property, not specific to a format. I would 
suggest adding it to v4l2_capability instead: 
V4L2_CAP_SENSOR_UPSIDE_DOWN. It should only be set if you know that the 
sensor is always upside down for that specific device and if there is 
no VFLIP capability (because if there is, then you can VFLIP by default 
for that device).

Regards,

	Hans

--
video4linux-list mailing list
Unsubscribe mailto:video4linux-list-request@redhat.com?subject=unsubscribe
https://www.redhat.com/mailman/listinfo/video4linux-list

Re: RFC: adding a flag to indicate a webcam sensor is installed upside down

[Hans de Goede]

Hans Verkuil wrote:
> On Friday 08 August 2008 11:10:36 Jean-Francois Moine wrote:
>> On Thu, 2008-08-07 at 12:37 +0200, Laurent Pinchart wrote:
>>> On Thursday 07 August 2008, Hans de Goede wrote:
>>>> Hi all,
>>>>
>>>> I have this Philips SPC 200NC webcam, which has its sensor
>>>> installed upside down and the sensor does not seem to support
>>>> flipping the image. So I believe the windows drivers fix this
>>>> little problem in software.
>>>>
>>>> I would like to add a flag somewhere to indicate this to
>>>> userspace (and then add flipping code to libv4l).
>>>>
>>>> I think the best place for this would the flags field of the
>>>> v4l2_fmtdesc struct. Any other ideas / objections to this?
>>> More often than sensors being mounted upside down in a webcam, what
>>> I've noticed frequently is webcam modules being mounted upside down
>>> in a laptop screen. There is no way that I'm aware of to check the
>>> module orientation based on the USB descriptors only. We will need
>>> a pure userspace solution.
>> Agree.
>>
>> Having a horizontal or vertical inversion may exist in any webcam
>> (and tuner?), and may be the fact of wire inversion when assembling
>> the device. So, having a flag in the driver could not work with
>> hardware error.
>>
>> If the sensor (or bridge?) may do H or V flip, we are done. If not,
>> this may be done by software, but in the userspace, i.e. at decoding
>> time.
>>
>> What I propose is to add a check of the control commands
>> V4L2_CID_HFLIP and V4L2_CID_VFLIP in the v4l library: if the driver
>> does not have these controls, the library silenctly adds them and
>> handles their requests. Then, at frame decoding time, the hflip and
>> vflip may be given to the decoding functions.
>>
>> How do you feel it, Hans?
> 
> I'm a different Hans :-), but I was thinking along the same lines. 
> Having the v4l lib implement HFLIP/VFLIP if the driver doesn't seems 
> like a good idea.
> 
> I also think that the UPSIDEDOWN flag doesn't belong to v4l2_fmtdesc 
> since this is a global property, not specific to a format. I would 
> suggest adding it to v4l2_capability instead: 
> V4L2_CAP_SENSOR_UPSIDE_DOWN. It should only be set if you know that the 
> sensor is always upside down for that specific device and if there is 
> no VFLIP capability (because if there is, then you can VFLIP by default 
> for that device).
> 


I agree that this is a per device thing, and that the flag thus should 
be moved to v4l2_capability. I'll redo a new patch against the gspca 
zc3xx driver with the flag moved to v4l2_capability.


About adding fake controls for HFLIP VFLIP to libv4l for devices which 
don't do this natively. This might be good to have, but my current patch 
for the upside down issue only supports. 180 degree rotation so, both 
vflip and hflip in one go.

Adding support for doing just vflip or just hflip is possible, but will 
take some time.

Also faking v4l2 controls currently is not done yet in libv4l, so doing 
this requires adding some infrastructure. All doable, but not at the top 
of my todo list at the moment.


As I see it there are 3 kind of upsidedown problems:

1) The sensor is upside down, and we can easily detect in the driver (by
usb id for example) and the driver does not support vflip and/or
hflip in the hardware. Solution: set a capability flag indicating this 
(and then in libv4l flip the image back in software, I've already 
written support for this).
(Note when h and vflip are supported the driver should invert the 
meaning of the ctrl's so that the image ends up the right way up).

2) The sensor is upside down, and we can not easily detect this in the 
driver, but we can detect it in userspace. For example a sensor mounted 
upside down in the screen bevel of a laptop, which we can detect be 
determining the model of laptop from dmi strings as hal already does for 
things like setting up bindings of special keyb keys and for enabling 
suspend / resume problem workarounds.

Proposed solution: Give userspace (hal) a way to tell the kernel that 
the sensor is upside down, I'm thinking about a sysfs file for each 
video device here myself. and then let the kernel behave as scenario 1.

3) There is no way to tell the sensor is upside down. Proposed solution 
write a videodev preferences gui, where this can be indicated and then 
use the sysfs file mentioned in 2) to tell the kernel about the sensor 
being upside down. Note that currently no practical examples of this 
version of the problem are known.


I'm currently focussing on 1)  as I actually have a cam which has its 
sensor upside down and has a unique device id. I'll submit a new patch 
against gspca adding a capability flag to indicate the image is upside 
down and adjust my libv4l patches to check for this flag and do flipping 
in sw when its set.

When this is in place adding a sysfs entry which allows userspace to set 
this flag should be easy.

Regards,

Hans



--
video4linux-list mailing list
Unsubscribe mailto:video4linux-list-request@redhat.com?subject=unsubscribe
https://www.redhat.com/mailman/listinfo/video4linux-list

Re: RFC: adding a flag to indicate a webcam sensor is installed upside down

[Pádraig Brady]

Hans de Goede wrote:
> 
> I agree that this is a per device thing, and that the flag thus should
> be moved to v4l2_capability. I'll redo a new patch against the gspca
> zc3xx driver with the flag moved to v4l2_capability.
> 
> 
> About adding fake controls for HFLIP VFLIP to libv4l for devices which
> don't do this natively. This might be good to have, but my current patch
> for the upside down issue only supports. 180 degree rotation so, both
> vflip and hflip in one go.

you're referring flipping in hardware above right?

> Adding support for doing just vflip or just hflip is possible, but will
> take some time.

Any update on this.

Personally I'm using a 0402:5606 webcam with the latest uvcvideo.ko
There is a firmware update available for the webcam to flip the image,
so I'm presuming that is just changing the appropriate setting
to tell the hardware to flip the image.

There is no mention of *FLIP* capability in the uvcvideo driver at all.
Is there a standard way to get/set that in the camera?

> Also faking v4l2 controls currently is not done yet in libv4l, so doing
> this requires adding some infrastructure. All doable, but not at the top
> of my todo list at the moment.

So there is no software support yet.

cheers,
Pádraig.


--
video4linux-list mailing list
Unsubscribe mailto:video4linux-list-request@redhat.com?subject=unsubscribe
https://www.redhat.com/mailman/listinfo/video4linux-list

Re: RFC: adding a flag to indicate a webcam sensor is installed upside down

[Laurent Pinchart]

Hi,

On Tuesday 21 October 2008, Pádraig Brady wrote:
> Hans de Goede wrote:
> > I agree that this is a per device thing, and that the flag thus should
> > be moved to v4l2_capability. I'll redo a new patch against the gspca
> > zc3xx driver with the flag moved to v4l2_capability.
> >
> >
> > About adding fake controls for HFLIP VFLIP to libv4l for devices which
> > don't do this natively. This might be good to have, but my current patch
> > for the upside down issue only supports. 180 degree rotation so, both
> > vflip and hflip in one go.
>
> you're referring flipping in hardware above right?
>
> > Adding support for doing just vflip or just hflip is possible, but will
> > take some time.
>
> Any update on this.
>
> Personally I'm using a 0402:5606 webcam with the latest uvcvideo.ko
> There is a firmware update available for the webcam to flip the image,
> so I'm presuming that is just changing the appropriate setting
> to tell the hardware to flip the image.

Most image sensors support vertical and horizontal flipping (or at least image 
rotation). So it's really a matter of setting the image sensor parameters 
correctly.

> There is no mention of *FLIP* capability in the uvcvideo driver at all.
> Is there a standard way to get/set that in the camera?

Unfortunately the upside-down cameras don't export any hflip/vflip control, so 
there's no (known) way to control image rotation from the host.

> > Also faking v4l2 controls currently is not done yet in libv4l, so doing
> > this requires adding some infrastructure. All doable, but not at the top
> > of my todo list at the moment.
>
> So there is no software support yet.

Best regards,

Laurent Pinchart

--
video4linux-list mailing list
Unsubscribe mailto:video4linux-list-request@redhat.com?subject=unsubscribe
https://www.redhat.com/mailman/listinfo/video4linux-list