Hi Stuart, Looking at the PerspectiveCamera source with "hg blame", it looks like it has seen only superficial changes since January 2013. I bet what you're seeing is due to something else changing in the volume rendering interface underneath it. FWIW, the PerspectiveCamera does have tests: https://bitbucket.org/yt_analysis/yt/src/5cd6764df8f3fbb981eefa5f19ee9f98b0e... But the tests only check to make sure the camera works without crashing, and does not verify that the answer is correct. It's entirely possible something has broken in the last two years. -Nathan On Thu Dec 04 2014 at 1:01:51 PM Michael Zingale < michael.zingale@stonybrook.edu> wrote:

Hi Stuart, I've had a similar experience and haven't looked back at it in about 6 mos. I was planning on waiting until the new renderer is in, and then I was going to try again.

Mike

On Thu, Dec 4, 2014 at 2:55 PM, Stuart Levy wrote:

...

I know it's being rewritten, thanks to Sam and Cameron and others, but I'm hoping to use some sort of perspective camera for volume rendering. (What I'll eventually want is the dome camera which I know also exists, but wanted to start with a simpler plane-projection perspective camera.)

What I get: plausible images from the default orthographic Camera, but reliably blank images from PerspectiveCamera.

If I extract the sampler rays from PerspectiveCamera and plot them, they always seem to point in the hemisphere away from the given center point. If I change the sign of Z width, the viewpoint is put on the opposite side of the data region, and the rays still point away from it. Also they're extremely short, like .001 or so of the distance from the viewpoint to the center point - much too short to intersect anything if the viewpoint is outside of the data volume.

If I hack the sampler vectors in PerspectiveCamera to (a) reverse their direction and (b) make them many times longer, I do see something.

I don't know whether I'm passing nonsense parameters to PerspectiveCamera, or whether its current code is really not right...?

A sample program, which contains its own replacement for Camera.snapshot() that allows for tweaking PerspectiveCamera rays, is in:

http://paste.yt-project.org/show/5240/

_______________________________________________ yt-users mailing list yt-users@lists.spacepope.org http://lists.spacepope.org/listinfo.cgi/yt-users-spacepope.org

-- Michael Zingale Associate Professor

Dept. of Physics & Astronomy • Stony Brook University • Stony Brook, NY 11794-3800 *phone*: 631-632-8225 *e-mail*: Michael.Zingale@stonybrook.edu *web*: http://www.astro.sunysb.edu/mzingale _______________________________________________ yt-users mailing list yt-users@lists.spacepope.org http://lists.spacepope.org/listinfo.cgi/yt-users-spacepope.org

On Fri Dec 05 2014 at 9:04:36 AM Suoqing JI wrote:

Hi Stuart,

I’ve just read over the source code of PerspectiveCamera, but do not quite understand it… So I wrote anther version of PerspectiveCamera class and pasted it here: http://paste.yt-project.org/show/5243/ http://paste.yt-project.org/show/WPJ9EYMeYDyb4oPtgC0z/

And here’re the test figures of moving camera through the volume:

http://pbrd.co/1FWWiLE http://pbrd.co/1FWWpH2 http://pbrd.co/1FWWGJT

Also, in the YT-3 dev version there’re SphericalCamera and StereoSphericalCamera, which will generate 360 degree dome views.

PS. I’m not sure whether there’ll be someone who can fix PerspectiveCamera, or I’ll create a PR…

I think this would be a great contribution. Cameron or Sam, are you ok with this updated camera going in while the VR rewrite proceeds?

Best wishes,

-- Suoqing JI Ph.D Student Department of Physics University of California, Santa Barbara CA 93106, USA

On Dec 4, 2014, at 1:08 PM, Nathan Goldbaum wrote:

Hi Stuart,

Looking at the PerspectiveCamera source with "hg blame", it looks like it has seen only superficial changes since January 2013. I bet what you're seeing is due to something else changing in the volume rendering interface underneath it.

FWIW, the PerspectiveCamera does have tests:

https://bitbucket.org/yt_analysis/yt/src/5cd6764df8f3fbb981eefa5f19ee9f98b0e...

But the tests only check to make sure the camera works without crashing, and does not verify that the answer is correct. It's entirely possible something has broken in the last two years.

-Nathan

On Thu Dec 04 2014 at 1:01:51 PM Michael Zingale < michael.zingale@stonybrook.edu> wrote:

...

Hi Stuart, I've had a similar experience and haven't looked back at it in about 6 mos. I was planning on waiting until the new renderer is in, and then I was going to try again.

Mike

On Thu, Dec 4, 2014 at 2:55 PM, Stuart Levy wrote:

...

I know it's being rewritten, thanks to Sam and Cameron and others, but I'm hoping to use some sort of perspective camera for volume rendering. (What I'll eventually want is the dome camera which I know also exists, but wanted to start with a simpler plane-projection perspective camera.)

What I get: plausible images from the default orthographic Camera, but reliably blank images from PerspectiveCamera.

If I extract the sampler rays from PerspectiveCamera and plot them, they always seem to point in the hemisphere away from the given center point. If I change the sign of Z width, the viewpoint is put on the opposite side of the data region, and the rays still point away from it. Also they're extremely short, like .001 or so of the distance from the viewpoint to the center point - much too short to intersect anything if the viewpoint is outside of the data volume.

If I hack the sampler vectors in PerspectiveCamera to (a) reverse their direction and (b) make them many times longer, I do see something.

I don't know whether I'm passing nonsense parameters to PerspectiveCamera, or whether its current code is really not right...?

A sample program, which contains its own replacement for Camera.snapshot() that allows for tweaking PerspectiveCamera rays, is in:

http://paste.yt-project.org/show/5240/

_______________________________________________ yt-users mailing list yt-users@lists.spacepope.org http://lists.spacepope.org/listinfo.cgi/yt-users-spacepope.org

-- Michael Zingale Associate Professor

Dept. of Physics & Astronomy • Stony Brook University • Stony Brook, NY 11794-3800 *phone*: 631-632-8225 *e-mail*: Michael.Zingale@stonybrook.edu *web*: http://www.astro.sunysb.edu/mzingale _______________________________________________ yt-users mailing list yt-users@lists.spacepope.org http://lists.spacepope.org/listinfo.cgi/yt-users-spacepope.org

_______________________________________________ yt-users mailing list yt-users@lists.spacepope.org http://lists.spacepope.org/listinfo.cgi/yt-users-spacepope.org

_______________________________________________ yt-users mailing list yt-users@lists.spacepope.org http://lists.spacepope.org/listinfo.cgi/yt-users-spacepope.org

Hi Suoqing, I agree, this would be great to add, and we can thread it through to the new VR framework once it gets merged. Great job! On Fri Dec 05 2014 at 3:06:15 PM Stuart Levy wrote:

Wow, this is wonderful! I was thinking to just rewrite it, which seemed easier than trying to understand the existing code, but you've done it so quickly - thank you!

I haven't had a chance to try your new PerspectiveCamera yet but will do. It looks like a perfectly sensible computer-graphics camera, with field-of-view equal to 2*atan(0.5*width[0]/width[2]) and 2*atan(0.5*width[1]/width[2]).

thanks again

Stuart

On 12/5/14 1:07 AM, Suoqing JI wrote:

Hi Stuart,

I’ve just read over the source code of PerspectiveCamera, but do not quite understand it… So I wrote anther version of PerspectiveCamera class and pasted it here: http://paste.yt-project.org/show/5243/ http://paste.yt-project.org/show/WPJ9EYMeYDyb4oPtgC0z/

And here’re the test figures of moving camera through the volume:

http://pbrd.co/1FWWiLE http://pbrd.co/1FWWpH2 http://pbrd.co/1FWWGJT

Also, in the YT-3 dev version there’re SphericalCamera and StereoSphericalCamera, which will generate 360 degree dome views.

PS. I’m not sure whether there’ll be someone who can fix PerspectiveCamera, or I’ll create a PR…

Best wishes,

-- Suoqing JI Ph.D Student Department of Physics University of California, Santa Barbara CA 93106, USA

On Dec 4, 2014, at 1:08 PM, Nathan Goldbaum wrote:

Hi Stuart,

Looking at the PerspectiveCamera source with "hg blame", it looks like it has seen only superficial changes since January 2013. I bet what you're seeing is due to something else changing in the volume rendering interface underneath it.

FWIW, the PerspectiveCamera does have tests:

https://bitbucket.org/yt_analysis/yt/src/5cd6764df8f3fbb981eefa5f19ee9f98b0e...

But the tests only check to make sure the camera works without crashing, and does not verify that the answer is correct. It's entirely possible something has broken in the last two years.

-Nathan

On Thu Dec 04 2014 at 1:01:51 PM Michael Zingale < michael.zingale@stonybrook.edu> wrote:

...

Hi Stuart, I've had a similar experience and haven't looked back at it in about 6 mos. I was planning on waiting until the new renderer is in, and then I was going to try again.

Mike

On Thu, Dec 4, 2014 at 2:55 PM, Stuart Levy wrote:

...

I know it's being rewritten, thanks to Sam and Cameron and others, but I'm hoping to use some sort of perspective camera for volume rendering. (What I'll eventually want is the dome camera which I know also exists, but wanted to start with a simpler plane-projection perspective camera.)

What I get: plausible images from the default orthographic Camera, but reliably blank images from PerspectiveCamera.

If I extract the sampler rays from PerspectiveCamera and plot them, they always seem to point in the hemisphere away from the given center point. If I change the sign of Z width, the viewpoint is put on the opposite side of the data region, and the rays still point away from it. Also they're extremely short, like .001 or so of the distance from the viewpoint to the center point - much too short to intersect anything if the viewpoint is outside of the data volume.

If I hack the sampler vectors in PerspectiveCamera to (a) reverse their direction and (b) make them many times longer, I do see something.

I don't know whether I'm passing nonsense parameters to PerspectiveCamera, or whether its current code is really not right...?

A sample program, which contains its own replacement for Camera.snapshot() that allows for tweaking PerspectiveCamera rays, is in:

http://paste.yt-project.org/show/5240/

_______________________________________________ yt-users mailing list yt-users@lists.spacepope.org http://lists.spacepope.org/listinfo.cgi/yt-users-spacepope.org

-- Michael Zingale Associate Professor

Dept. of Physics & Astronomy • Stony Brook University • Stony Brook, NY 11794-3800 *phone*: 631-632-8225 *e-mail*: Michael.Zingale@stonybrook.edu *web*: http://www.astro.sunysb.edu/mzingale _______________________________________________ yt-users mailing list yt-users@lists.spacepope.org http://lists.spacepope.org/listinfo.cgi/yt-users-spacepope.org

_______________________________________________ yt-users mailing list yt-users@lists.spacepope.org http://lists.spacepope.org/listinfo.cgi/yt-users-spacepope.org

_______________________________________________ yt-users mailing listyt-users@lists.spacepope.orghttp://lists.spacepope.org/listinfo.cgi/yt-users-spacepope.org

_______________________________________________ yt-users mailing list yt-users@lists.spacepope.org http://lists.spacepope.org/listinfo.cgi/yt-users-spacepope.org