Nuke node for model-based single view object and camera tracking
GeoTracker is a Nuke node for model-based single view object and camera tracking. It’s accurate, easy-to-use and fast. GeoTracker is developed using low level Nuke’s C++ API. That’s why it could be pretty seamlessly integrated into your familiar workflow.
Accurate object tracking
GeoTracker is designed to fill the gap in Nuke’s functionality and provide fast and reliable object tracker right in the compositing software.
PinTool under the hood
GeoTracker is built on top of PinTool and includes all its strengths such as simple and powerful model positioning tool, advanced model drawing, lots of exporting options, focal length estimation, etc.
Scene-based camera tracking
GeoTracker can track camera even in tricky shots with zoom and high motion blur. Scene-based camera tracking gives you ability to use keyframes, which allows localizing and tackling complex part of a sequence. Also resulting camera is already aligned to a scene geometry. Zooming lenses could add a significant complexity for simple camera tracking. Fortunately for scene-based tracking you can track it much easier and more precise.
To obtain more stable and robust object tracking you can import you ordinary 2D tracks from Nuke’s Tracker node into GeoTracker. It will consider them as “gold” tracks and give them additional weight in solver. Usually you’ll not need them but for really complex shots with poor features they can help a lot.
GeoTracker allows you to use 2D mask to eliminate occlusion and 3D surface masks to exclude reflective and transparent parts of a tracked object.
Focal length tracking
Although it is possible to estimate focal length in key frames and use interpolated focal length between them it could be very useful to track camera zoom changes with focal length tracking.
“Magic Keyframe″ allows you to automatically set up key frames. It speed ups tracking process a lot and provides you with consistent and precise key frames.
Rigged geometry tracking
GeoTracker allows tracking of objects with skeleton by calculating bone transformations. It’s also possible to fixate some degrees of freedom to match real object behaviour or to export single bone transformation to attach rigid object to tracked rigged geometry.
Do I need a model to perform tracking?
Yes, you need a model of an object or a scene to perform tracking.
Should the model fit the scene perfectly?
No, it shouldn’t. But the better model you have the better results you obtain from every key frame.
How detailed should the model be?
The model should fit the scene. Minor details aren’t very important and could have negative impact on tracking performance
What if my model has semi-transparent or reflective parts?
You may use 3D surface mask to exclude semi-transparent or reflective parts from tracking process.
What if the object I’m tracking is partially overlapped?
If the overlapped area is relatively small GeoTracker will handle it. Otherwise you may use a 2D mask to exclude overlapping from tracking process.
What’s the output of GeoTracker node?
GeoTracker’s output is the transformed input geometry.
Do I need GeoTracker installed on every machine in my pipeline?
No. As soon as you’ve done tracking you may export the results as a usual Nuke node (TransformGeo or Camera node) to pass them to other machines.
How to export tracking results?
You can export the results as a TransformGeo or a Camera node on the results tab.
How to export tracking results to other software?
You can export the results from the exported TransformGeo or Camera node as a .chan file.
The Foundry Nuke 8, 9, 10, 10.5, 11, 11.1
This link is being provided as a convenience and for informational purposes only; it does not constitute an endorsement or an approval by Toolfarm.com of any of the products, services or opinions of the corporation or organization or individual. Toolfarm.com bears no responsibility for the accuracy, legality or content of the external site or for that of subsequent links. Contact the external site for answers to questions regarding its content.