I see. So you need the lookAt in reference to the world, and not to the camera. You have already figured that out, but all you need then is to take your lookAt direction vector and add the camera position to it:
To avoid the slicing of the original matrix, you could do the following: #LOOK AT POSITION lookAt = camMatrix.vtransform(nuke.math.Vector3(0,0,-1) ) # This is just the "direction vector" lookAt += camPos or just this, which should be the same: lookAt = camMatrix.transform(nuke.math.Vector3(0,0,-1) ) Basically, what you were doing with: lookAt = nuke.math.Vector3( camMatrix[8], camMatrix[9], camMatrix[10] ) lookAt.negate() should always give you a (0,0,-1) vector when the camera is not rotated. So the above should be equivalent, I think. Makes sense? On Mon, Mar 7, 2011 at 5:06 PM, Frank Rueter <[email protected]> wrote: > looks like I got confused on my side. I realised I don't need a lookAt > vector but a position in space that the camera is "looking at". > > a friend helped out and gave me this: > > camMatrix = getWorldMatrix( camNode ) #USING IVANS FUNCTION > > #POSITION camPos = camMatrix.transform( nuke.math.Vector3(0,0,0) ) > > #UP VECTOR > upVect = camMatrix.yAxis() > #LOOK AT POSITION > lookAt = nuke.math.Vector3( camMatrix[8], camMatrix[9], camMatrix[10] ) > lookAt.negate() lookAt += camVect > > > This seems to work like a charm. I do wonder if there is a more elegant way > to get the initial lookAt though without slicing? > > Cheers, > frank > > > On Mar 8, 2011, at 12:45 PM, Frank Rueter wrote: > > hm, I'm having trouble getting the right lookAt vector. > this is what I'm doing: > > def getWorldMatrix( node ): > m = nuke.math.Matrix4() > for i in range(0,16): > m[i] = node['world_matrix'].value(i%4, i/4) > return m > > camNode = nuke.createNode( 'Camera2' ) > camNode['translate'].setValue( (1,2,3) ) > > camMatrix = getWorldMatrix( camNode ) > worldLookAt = nuke.math.Vector3(0,0,-1) > localLookAt = camMatrix.vtransform( worldLookAt ) > localLookAt = camMatrix.vtransform( worldLookAt ) > print localLookAt > > > > > > The result for the lookAt vector stays at { 0,0,-1} even though the > camera's position is {1,2,3}. > What am I missing? > > Cheers, > frank > > > > > On Mar 7, 2011, at 3:42 PM, Ivan Busquets wrote: > > Hallo, Frank. > Despite writing that tutorial, I am by no means an expert in matrix and > maths. :) > > But I've learnt a thing or two about how they apply to things like camera > transforms, etc., so hopefully I'll be able to help with this one. > > I think the easiest in your case (easiest to code, not necessarily to > understand :p) is to start with the camera's transformation matrix instead > of the Euler rotations. If you want to use Euler angles as a starting point, > then you'll need to check for rotation order and such. > > To get the camera's transformation matrix, i usually do one on the > following: > > - If your camera doesn't have any other Cameras or Axis driving its final > transformation, you can simply do: > > cam_transform = nuke.toNode('camera_name')['transform'].value() > > - If you're interested in getting the final transformation matrix of a > concatenated set of transforms, I usually use the following to wrap the > contents of the 'world_matrix' knob into a nuke.math.Matrix4 object: > > ######### > def getCameraTransformMatrix(cameraNode): > m = nuke.math.Matrix4() > try: > for i in range(0,16): > m[i] = cameraNode['world_matrix'].value(i%4, i/4) > except: > print "couldn't get all values for a 4x4 Matrix. Identity matrix returned > instead" > m.makeIdentity() > return m > ############## > > > Finally, once you have your transformation matrix in a nuke.math.Matrix4 > object, getting the lookAt and the Up vectors should be a matter of setting > the initial direction of those two vectors, and then transform them using > the matrix. > > I.E. > > # For a single, non concatenated camera. Otherwise use the helper funcion > above. > cam_matrix = nuke.toNode('camera_name')['transform'].value() > > # set up initial lookAt vector > lookAt = nuke.math.Vector3(0,0,-1) # because Nuke's default cam looks at -Z > > # Initial up vector > up = nuke.math.Vector3(0,1,0) > > #Final lookAt vector > lookAt = cam_matrix.vtransform(lookAt) > > #Final up vector > up = cam_matrix.vtransform(up) > > # Note that the key here is to use the "vtransform" method instead of > "transform", since you want to apply the transformation to a direction > vector, but ignore any scaling or translation. > > I think this should get you the right values to go from a Nuke camera to, > say, a Camera + Aim + Up in Maya (don't have Maya here to check, though). > > Hope that helps! > > Cheers, > Ivan > > > On Sun, Mar 6, 2011 at 3:06 PM, Frank Rueter <[email protected]>wrote: > >> Hola, >> >> has anybody converted a nuke camera (Euler rotations) to LookAt and Up >> vectors? as used by some other applications >> I am just starting to read Ivan's tutorial on Nukepedia hoping it will >> shed some light on this for somebody with very limited understanding of 3D >> rotational maths (aka "me") but thought I'd spread my feelers in here at the >> same time. >> >> Cheers, >> frank_______________________________________________ >> Nuke-python mailing list >> [email protected] >> http://support.thefoundry.co.uk/cgi-bin/mailman/listinfo/nuke-python >> > > _______________________________________________ > Nuke-python mailing list > [email protected] > http://support.thefoundry.co.uk/cgi-bin/mailman/listinfo/nuke-python > > > _______________________________________________ > Nuke-python mailing list > [email protected] > http://support.thefoundry.co.uk/cgi-bin/mailman/listinfo/nuke-python > > > > _______________________________________________ > Nuke-python mailing list > [email protected] > http://support.thefoundry.co.uk/cgi-bin/mailman/listinfo/nuke-python > >
_______________________________________________ Nuke-python mailing list [email protected] http://support.thefoundry.co.uk/cgi-bin/mailman/listinfo/nuke-python
