Photo/camera match

[danielcg86]

Hi all

 

I'm a third year architecture student and am wishing to insert a 3ds max model into a photograph - ie. exterior perspective. I have read a number of tutorials online discussing how to position the model using known measured points.

 

Is there an alternative approach as I don't know the co-ordinates of any points in the photograph. It doesn't have to blow my tutor's mind but I'd be happy if it looks like it 'belongs'.

 

Any advice would be greatly appreciated. Thanks in advance.

 

Daniel

[andstef]

you can set the photograph as viewport background and position your model approximately, eyeball-it

[Crazy Homeless Guy]

Max sucks at camera matching. It is possible if you have extremely exact measurements on the site, and very specific information on camera settings, but even at that seasoned professionals will struggle with getting a perfect match according to the numbers. You are far better eyeballing it as has been suggested.

 

I went to a recent lecture on using Autodesk Matchmover in conjunction with 3dsMax, and it produced believable results, but you would need video and a decent amount of technical understanding.

[WAcky]

Even with a good site survey and correct camera data (exif data) the results can be iffy. The only way (imo) to get near correct results is to have the camera postion added to the site survey.

 

You may be able to get the exif info from the photo you're using to help you out in eyeballing it...

[2hands]

kinda sad that the best method seems to be to eyeball it

[William Alexander]

If you have Photoshop or CAD you can, depending upon the structures in the image, get some rough perspective lines to help close in on 'rough dimensions'. Generally finding vanish points is very helpful albeit frustrating and usally requiring lens (distortion) Corrections.

 

You can export the image with the perspective lines as the BG image in max and change camera settings and position to match the perspective grid/s.

 

Suggest doing what you can for the camera match and work very hard to make the lighting, GI ect "marry" the CG to the Image.

 

Personally have not really used or seen any application that works stupid well. Image Modeler is/was good at camera matching, however you need multiple views to make it work. Basically a static version of Match Mover, created by the same company as Image Modeler

 

Good Luck

Wm

[Claudio Branch]

I've been an eye-baller for years now. If you ouput your file as a .png, the background will automatically drop out as alpha. This will give you a lot more flexibility in post if you find yourself needing/wanting to push and pull background pixels as well.

[TSuess]

Yeah Eye-balling is the way to go. But be absolutely sure to match the lens length. Placing a 24mm rendered object on a photo taken with a 50 mm lens is an excercise in futility.

[sconlogue]

If you have multiple shots of the subject at hand you could give Autodesk Labs Photofly a try. I have been playing around with it a bit and have found it very useful. Basically it's a cloud computing service that you upload the images to then it process them (depending on the image resolution, quantity and subject this can take some time) then provide you with a number of formats to download the 3D data set it creates. In your case I would download the FBX file. When you load it into max you will see dummy objects representing all the tracking points it derived from your images as well as the cameras that represent cam info for each image you provided. It may not work for what your doing but it's certainly worth a try.

 

(You may want to turn your speakers down as Autodesk rocks out on this one.

 

BTW Photofly exports to Autodesk Image Modeler if you need to actual model an existing structure.