Extracting From A Video - Not working for me

[red]

Hello - When using pictures imported from a video, from a turntable shoot, I am getting distorted/warped cameras/images. Furthermore the cameras in the viewport do not orient like they have (ie around the subject) when I do a turntable shoot w/o using video. Needless to say the results are useless.

Any advice? Is there any relevant video tutorial perhaps? I have never used the video technique and would love to have some success with it. Documentation seems to be all over the place, and I can't find a good step by step.

Thanks.

[cam3d]

Hi red

Can you please share the video so I (or someone else) can analyze and give you feedback? Could be one of many factors but principally it's the same as photos, except extracted from video frames.

Re: The best source of documentation can be found here: https://www.3dflow.net/technology/do...documentation/
Starting a project from a video: https://www.3dflow.net/zephyr-doc/en...romavideo.html

[cam3d]

Typical problems with video are motion blur and low resolution - if you don't have enough light and are compensating with ISO or low F-stop then this can cause issues, but without the video itself it's hard to say what the culprit is

[Andrea Alessi]

the cameras in the viewport do not orient like they have (ie around the subject) when I do a turntable shoot w/o using video. Needless to say the results are useless.

As Cameron pointed out it depends from the video, but given what you say, my suspect is that you haven't masked your photos. If you're using a turntable, you must mask out the background as zephyr has no way to know if you want to reconstruct your object or the room, as they are no longer coherent the moment you spin the turntable. So unless you can recreate a plain background, you need to use masquerade.

If you want to share the video, we'll be happy to take a look at it!

[red]

Thank you cam3d and Andrea Alessi

I looked at the footage again this morning and realize my lighting is pretty bad, and the subject as a bit of shine. I have posted it on Dropbox for you to download.

All the pictures have all been masked using Masquerade.

That said, despite the bad footage, I wouldn't think I would get the strange fisheye effect with the cameras. Furthermore, there is no semblance of a turntable camera set-up in the viewport. See below.

Thank you for your help.

https://www.dropbox.com/s/r2xfpabqxs...otate.mp4?dl=0

[cam3d]

red Thanks for sharing the video. Here are some things to note:


- Resolution is 1080*1920 and of this resolution, the actual subject only takes up a small portion of this so you're working with an incredibly small amount of data.
- The subjects' surface is changing with the light - Metallic surfaces are notorious for this. I would gently dust with talc or something similar to nullify this problem.
- Spinning quickly introduces motion blur - I note that you shot at 59.94fps and it's hard to perceive but even the smallest amount of motion blur compounds with the issues above to further cause problems.

IMO video for SFM/MVS etc... is cool, but it's way harder to get good results than classic photography and I personally stay away from it because it's an uphill struggle, def not easy street.

Something which would help with alignment would be using a feature-rich base instead of the black one, like fine print newspaper or even better, a printout of perlin noise.

As for the fisheye, I'm not sure but I'd guess that it has something to do with that fact the video file doesn't have meta-data alluding to camera settings (Focal/Fstop/etc..). Andrea Might be able to elaborate or correct me on this

[red]

red Thanks for sharing the video. Here are some things to note:


- Resolution is 1080*1920 and of this resolution, the actual subject only takes up a small portion of this so you're working with an incredibly small amount of data.
- The subjects' surface is changing with the light - Metallic surfaces are notorious for this. I would gently dust with talc or something similar to nullify this problem.
- Spinning quickly introduces motion blur - I note that you shot at 59.94fps and it's hard to perceive but even the smallest amount of motion blur compounds with the issues above to further cause problems.

IMO video for SFM/MVS etc... is cool, but it's way harder to get good results than classic photography and I personally stay away from it because it's an uphill struggle, def not easy street.

Something which would help with alignment would be using a feature-rich base instead of the black one, like fine print newspaper or even better, a printout of perlin noise.

As for the fisheye, I'm not sure but I'd guess that it has something to do with that fact the video file doesn't have meta-data alluding to camera settings (Focal/Fstop/etc..). Andrea Might be able to elaborate or correct me on this

I think you nailed it on the head. Thank you very much for taking the time to look. I really did ham fist my lighting/setup. The perlin noise idea is new to me. Great idea. I think you are right regarding the fisheye. I just used the autocalibration, of which I had little faith in.

I will indeed be sticking to classic photography. Thanks again.

[cam3d]

No worries - If you ever get stuck again with any step of the process we are here to help!