OPTIMIZED VIDEO PROCESSING THROUGH SOURCE-SIDE TAGGING FOR GENERATIVE ARTIFICIAL INTELLIGENCE SYSTEMS
DE102025135576A1Pending Publication Date: 2026-03-05NVIDIA CORP
View PDF 0 Cites 0 Cited by
Patent Information
- Application Number
- DE102025135576
- Authority / Receiving Office
- DE · DE
- Patent Type
- Applications
- Current Assignee / Owner
- Priority Date
- 2024-09-05
- Filing Date
- 2025-09-04
- Publication Date
- 2026-03-05
AI Technical Summary
Technical Problem
Conventional video language models (VLMs) require significant computational resources and fail to capture all relevant information due to impractical processing of entire video frames, missing critical details in unsampled frames.
Method used
Implementing local event detection on the device capturing the video stream to tag frames with attributes like motion, objects, or temporal activity, generating an encoded bitstream with markers for relevant frames to be selectively provided to the VLM.
Benefits of technology
Reduces computational requirements by providing only relevant frames to the VLM, avoiding costly processing operations and ensuring critical details are captured.
✦ Generated by Eureka AI based on patent content.
Smart Images
Figure 00000000_0000_ABST
Abstract
Various examples disclose systems and methods for generating video streams for generative artificial intelligence models. A system can receive a multitude of frames from a capture device that captures a video stream. The system can determine that at least one frame from the multitude of frames should be provided as input to a machine learning model. The system can generate a statement indicating that the at least one frame should be provided as input to the machine learning model. The system can generate an encoded bitstream for the video stream. The encoded bitstream can contain encoded data for the multitude of frames and the statement.
Need to check novelty before this filing date? Find Prior Art