Pixels2Position v1.3

Updates 1.3:

Fixed: 'exportXY' won't break if camera has a parent axis.

Updates 1.2:

1-Added support for nuke camera v001

2-Fixed exportXY on linux for nukev11.1 

Usage:

1-sample the viewer and click add point.

2- go to another frame where the feature is still visible.

3-sample the same feature and voila.

 P.S: to get an accurate result, leave at least few frames between the two samples. the bigger the distance between them, the more accurate result. 'Scene Scale' does not need to be changed unless your scene scale is enormous. 

Accuracy close to 0 is best, values under 0.1 is good, and anything between 0.1 and 0.5 is sometimes acceptable, above 0.5 something is wrong with camera track or distortion map.

Method:

This tool shoots two vectors from camera into depth passing through the sampled pixels, then it calculates the intersection points (3d position) between the two vectors, from there you have the option to generate 2d position of that feauture.

Technical info:

Foundry's 'PointsTo3D' and other tools, depend on nuke's 'snap3d' module which doesn't work with cameras that have 'window translate' set to anything but default. This tool calculates the camera's projection matrix correctly and apply it to the result.

Super simple to setup and it produces a very accurate result by reducing human error and depending completely on math calculations .

Known limitation:

'exportXY' will not produce correct result if camera has a look parent.