| ipisoft.com https://forum.ipisoft.com/ |
|
| Calibration Camera Position Problem https://forum.ipisoft.com/viewtopic.php?f=5&t=9486 |
Page 1 of 1 |
| Author: | tommy4811 [ Wed Dec 16, 2015 6:50 am ] |
| Post subject: | Calibration Camera Position Problem |
Hey all. I have been using iPi for a while now and everything has been working great. But the other day I recorded a video indented to be a calibration project and when I imported it into iPi Studio the cameras were not positioning themselves in the correct place but instead they were just on top of each other as seen below. http://i.imgur.com/zGQytjE.jpg Camera 1: http://i.imgur.com/JqnHZEc.jpg Camera 2: http://i.imgur.com/vobcaUe.jpg I thought this might have been a bad recording or something so I went back and recorded a few more but still got the same result. I did a recording the previous day and that worked fine: http://i.imgur.com/J2z1sBJ.jpg So I loaded in the raw footage from the previous day that I know worked and now it acts the same as the new recordings with the cameras on top of each other. Let me know what else you need from me to help with this. Thanks. |
|
| Author: | vmaslov [ Wed Dec 16, 2015 10:28 am ] |
| Post subject: | Re: Calibration Camera Position Problem |
Hi Looks like a video card issue. Does tracking work in older projects? Have you changed your hardware recently? Have you (or OS) installed a new video driver recently? What is your video card? |
|
| Author: | Snapz [ Thu Dec 17, 2015 11:40 am ] |
| Post subject: | Re: Calibration Camera Position Problem |
... You can check what Maslov stated, but much has been stated about the Kv2s and the cardboard plane calibration technique not being the best way to calibrate the the dual Kv2s, but it does work... sometimes ok. I had the same issues when doing it that way, although it would work fine sometimes, others it wouldn't, but I just ended up re-clicking the calibrate button a few times and let it re-cal and after a few runs it would reset the cams to correct position and give a PERFECT reading. You can attempt the light marker calibration and see if that helps, it is stated in the Docs as the best method for Kv2 cams calibration. Good Luck! ... |
|
| Author: | Greenlaw [ Thu Dec 17, 2015 12:36 pm ] |
| Post subject: | Re: Calibration Camera Position Problem |
Just curious but have you tried recording in the evening? The space I shoot in gets too much sunlight (i.e., IR pollution) during the day which interferes with the Kinect sensors. Even a MagLite doesn't work well because the sunlight tends to make it less visible in the room. When I record in the evening with either board, PS Move or Maglite method, I get perfect results every time. (I prefer the board myself since it doesn't require the RGB stream.) Alternatively, I've covered the windows with blackout material during the day but it's kind of a hassle, so I just schedule to shoot in the evenings now. The above has been true for either v1 or v2 sensors. G. |
|
| Author: | tommy4811 [ Thu Dec 17, 2015 1:35 pm ] |
| Post subject: | Re: Calibration Camera Position Problem |
vmaslov wrote: Hi Looks like a video card issue. Does tracking work in older projects? Have you changed your hardware recently? Have you (or OS) installed a new video driver recently? What is your video card? Does tracking work in older projects? I can load in an old calibration file and it works fine. But using the recording I made that calibration from now acts the same way as the new recordings with both cameras on top of each other. Have you changed your hardware recently? No. Have you (or OS) installed a new video driver recently? I installed new Nvidia drivers recently but the calibration was still working after that. What is your video card? 2x Titan X in SLI Snapz wrote: ... You can check what Maslov stated, but much has been stated about the Kv2s and the cardboard plane calibration technique not being the best way to calibrate the the dual Kv2s, but it does work... sometimes ok. I had the same issues when doing it that way, although it would work fine sometimes, others it wouldn't, but I just ended up re-clicking the calibrate button a few times and let it re-cal and after a few runs it would reset the cams to correct position and give a PERFECT reading. You can attempt the light marker calibration and see if that helps, it is stated in the Docs as the best method for Kv2 cams calibration. Good Luck! ... I will give that a try. I have always gotten a PERFECT calibration with the new xbox one kinect so I never thought anything of it. Greenlaw wrote: Just curious but have you tried recording in the evening? The space I shoot in gets too much sunlight (i.e., IR pollution) during the day which interferes with the Kinect sensors. Even a MagLite doesn't work well because the sunlight tends to make it less visible in the room. When I record in the evening with either board, PS Move or Maglite method, I get perfect results every time. (I prefer the board myself since it doesn't require the RGB stream.) Alternatively, I've covered the windows with blackout material during the day but it's kind of a hassle, so I just schedule to shoot in the evenings now. The above has been true for either v1 or v2 sensors. G. I was wondering the same thing but when I used a recording I know was good and made a previous calibration with, I started to rule that out. I will still try that out though just to be sure. Thanks for quick feedback guys. I will keep messing around with it. |
|
| Author: | Snapz [ Thu Dec 17, 2015 2:20 pm ] |
| Post subject: | Re: Calibration Camera Position Problem |
... Wondering how your Kv2 tests are coming along G? You got the calibration issues you had corrected I imagine. ... |
|
| Author: | Greenlaw [ Thu Dec 17, 2015 2:46 pm ] |
| Post subject: | Re: Calibration Camera Position Problem |
Yes, it was just a matter of covering a greater z-distance with the board to make it absolutely clear to Mocap Studio that the sensors were facing each other. If the sensors are on the same side of the room (i.e, 90 degree setup), you only need to move the board about a meter but for a near-180 setup, you should move it at least a couple of meters distance. I've just been walking from one sensor to the other, holding the board out to the side, while giving it just a little random side-to-side motion (positional not rotational) at the same time. Rotating the board doesn't help as much because this motion essentially reads the same from either side of the board. G. |
|
| Author: | vmaslov [ Fri Dec 18, 2015 1:43 am ] |
| Post subject: | Re: Calibration Camera Position Problem |
tommy4811 wrote: I can load in an old calibration file and it works fine. But using the recording I made that calibration from now acts the same way as the new recordings with both cameras on top of each other. I saw you have mentioned that before. With my question, I meant tracking actor's motions. Just to check whether the issue affects the calibration only, or other operations as well. tommy4811 wrote: 2x Titan X in SLI Try to disable SLI or uninstall one of the cards completely. Does that change anything? |
|
| Author: | Greenlaw [ Fri Dec 18, 2015 9:32 am ] |
| Post subject: | Re: Calibration Camera Position Problem |
Greenlaw wrote: Rotating the board doesn't help as much because this motion essentially reads the same from either side of the board. i should clarify that rotating the board doesn't help for 180 degree calibration but it may be a little helpful for 90 degree calibration. Try to visualize how the sensors perceive the scene and what it needs to see to recreate the environment. |
|
| Page 1 of 1 | All times are UTC - 8 hours [ DST ] |
| Powered by phpBB® Forum Software © phpBB Group http://www.phpbb.com/ |
|
