Disney patents method to train AI for separating video objects
Disney describes a technique for creating training data to teach computer models how to separate foreground objects from backgrounds in video. The process involves capturing real objects or generating synthetic ones, combining them with simulated backgrounds, and labeling the results to help the model learn to identify specific items.
In the filing’s own words
Embodiments provide techniques for training machine learning models for rotoscoping and matting in VFX workflows. The method includes generating one or more foreground objects, comprising capturing real-world objects against one or more backgrounds using a color triangulation setup, or generating synthetic objects using multi-dimensional rendering models, generating one or more composite training samples by combining at least one of the one or more foreground objects with one or more simulated backgrounds, generating a first training dataset for a segmentation model, comprising annotating at least one of the one or more composite training samples to create a corresponding segmentation mask, and training the segmentation model to predict a segmentation mask for an input sample using the first training dataset.
What exactly they claim (claim 1)
1. A method, comprising: generating one or more foreground objects, comprising at least one of: capturing real-world objects against one or more backgrounds using a color triangulation setup, or generating synthetic objects using multi-dimensional rendering models; generating one or more composite training samples by combining at least one of the one or more foreground objects with one or more simulated backgrounds; generating a first training dataset for a segmentation model, comprising: annotating at least one of the one or more composite training samples to create a corresponding segmentation mask; and training the segmentation model to predict a segmentation mask for an input sample using the first training dataset.
Facts
- Company
- Disney (Disney Enterprises, Inc.,)
- Official title
- BOOTSTRAPPED ROTOSCOPE-MATTING TRAINING
- Publication number
- US20260278805A1
- Application number
- 19077203
- Filed
- Mar 12, 2025
- Published
- Sep 17, 2026
- Inventors
- 1
- Claims
- 20
Classified under
G06T 7/194 Physics & computingG06T 5/50 Physics & computingG06V 10/44 Physics & computingG06V 20/70 Physics & computingG06T 2207/20081 Physics & computingG06T 2207/20221 Physics & computingMore from Disney in AI & software
- Disney patents AI character that creates art from your shared memories Sep 8, 2026
- Disney patents video compression using generative models to cut file sizes Sep 3, 2026
- Disney patents AI characters that create art from shared memories Oct 22, 2026
- Disney patents tech to sync translated speech with body gestures Sep 17, 2026
- Disney patents AI characters that pause and react when you interrupt them Sep 24, 2026
Summaries are simplified from the filing’s own abstract and first claim. A patent filing is not a product announcement.
Patent data and drawings: U.S. Patent and Trademark Office (USPTO). Nebula is not affiliated with or endorsed by the USPTO.