Method, System, and Computer-Readable Medium for Training an AI Model to Generate Video Content by Automating Captioning and Understanding of Cinematic Components
AssigneeInterPositive, LLC
InventorBenjamin Geza Affleck-Boldt
Life of this patent
Estimating the term…
Abstract
A method trains an AI model to recognize cinematic components by providing video data with descriptions, processing through a captioner model, and training the AI model with the captioned data. A computing system trains an AI model to generate new video content by providing video data, processing it for captioning, and training the AI model with specific cinematic understanding. A computer-readable medium stores instructions for training an AI model to accurately generate video content by providing video data, automating captioning, and training the AI model with an understanding of cinematic components.
Claims34
A computer-implemented method for training an artificial intelligence (AI) model, the method comprising:
- providing a large volume of video data, each video data being paired with a textual description;
- processing the video data through a captioner model to automate the captioning and labeling of the video data;
- and
- training the AI model using the captioned and labeled video data,
- wherein the captioner model encodes geographic location data numerically to represent spatial relationships within a scene, including interpreting camera position in three-dimensional space relative to a fixed origin point and the position of objects or subjects within the scene using a numerical coordinate system,
- wherein the AI model is configured to recognize and understand specific cinematic components including at least one of focal length, depth maps, movement, occlusion, parallax, or camera angles, to accurately generate new video content based on the training.
(canceled)
(canceled)
11-12. (canceled)
A computing system for training an artificial intelligence (AI) model, the system comprising:
- one or more processors;
- and
- one or more memories, having stored thereon instructions that, when executed, cause the computing system to:
- provide a large volume of video data, each video data being paired with a textual description;
- process the video data through a captioner model to automate the captioning and labeling of the video data,
- wherein the captioner model encodes geographic location data numerically to represent spatial relationships within a scene, including interpreting camera position in three-dimensional space relative to a fixed origin point and the position of objects or subjects within the scene using a numerical coordinate system;
- and
- train the AI model using the captioned and labeled video data,
- wherein the AI model is configured to recognize and understand specific cinematic components including at least one of focal length, depth maps, movement, occlusion, parallax, or camera angles, to accurately generate new video content based on the training.
(canceled)
24-26. (canceled)
A non-transitory computer-readable medium having stored thereon a set of non-transitory instructions that, when executed, cause a computer to:
- provide a large volume of video data, each video data being paired with a textual description;
- process the video data through a captioner model to automate the captioning and labeling of the video data,
- wherein the captioner model encodes geographic location data numerically to represent spatial relationships within a scene, including interpreting camera position in three-dimensional space relative to a fixed origin point and the position of objects or subjects within the scene using a numerical coordinate system;
- and
- train an artificial intelligence (AI) model using the captioned and labeled video data,
- wherein the AI model is configured to recognize and understand specific cinematic components including at least one of focal length, depth maps, movement, occlusion, parallax, or camera angles, to accurately generate new video content based on the training.
(canceled)
37-39. (canceled)
Dependencies are inferred from the claim text.