PhotoModeling

that pic shows a maya interface. the image planes are most likely aligned by eye and then the spline cage has to be created by hand, based on the aligned cameras.
so, nothing special here
 

Falango

Active Member
Hard to tell how those images were lined up without more information. Manually just dropping them into world space in maya, then building a spline cage, and constantly moving the image planes would take forever, literally, and it'll never be accurate enough for modeling. A third element is needed to verify the lineups, whether it's a 3d scan, or something like IM.
 
I contacted Kris Eggleton and he said what he did in maya was
Locked the image aspect ratio and loaded the reference photos as image planes in multiple cameras, then I used a scale analogue to match perspective and place the cameras.
 
lol no i wish but if you look here WIP - AlfaRomeo 8c Competizione you can see sort of how he does it. ive just started messing with maya so not sure about the camera stuff. looks like what he does is uses a couple box and align the pics with the cars im thinking thats the analogue.

then you rotate the cam and image planes to match the boxes.
 

Jpaang

Active Member
I noticed something in IM. If you increase the high pixel residual in the preferences menu you will get more green locators. My high residual is at 5 pixels. Not sure if I should go any higher, i'm not sure if it will affect the calibration of my freelander 2 in a bad way. VitaminD, do you know anything about this?
 
Residuals are like treshold for "greenness" of Markers :)
Every marker has its own residual (on each picture) wich defines difference in pixels between manually set position of locator on this picture(marker) and calculated during calibration.

If this distance is lower then "low residual" - marker is Green. It means that difference is lesser then low threshold, you've set in preferences.
On the other hand - if it is higher then High residual - then it turns red, and it means that marker position is too far from calculated.
Yellow - "so so" :)

So changing residuals you setup what preciseness you need for your work. It doesn't influence on calibration or preciseness of calibration itself, only visual representation for estimation of quality.

For images about 1200-1600 pix with cars 2-4/3-5 residuals are good enough for work.

It all depends on resolution of images and dimensions of object (size of "pixel" of image in real scale")
In car on image ~1600px it 1px ~ 1mm so 3-5 mm of threshold is okay

Couple of weeks ago i've made helicopter using photos. Photos were about 4000px wide, but taking into account, that heli was almost 9 meters long, ratio px/mm was almost the same as with cars, so i worked with 2-4 or 3-5 too.