Generative Video Response Filtering for Relevant Vision Data

Resolve Bottlenecks,
Find Innovative SolutionsGenerate Solutions

Solution Overview

Problem

Existing generative models (GMs) face inefficiencies in processing vast amounts of visual data, leading to wastage of computational and network resources due to extraneous information, which affects task performance and resource utilization.

Innovation Solution

Client devices filter vision data based on spoken utterances to determine a relevant subset for processing, using GMs to generate responsive content, thereby reducing unnecessary data processing and transmission.

Engineering Contradictions & Design Principles

VSEngineering Contradiction Analysis

1Reliability

If vision data is processed using generative models, then task performance is improved, but computational resources are wasted due to extraneous data

Engineering Contradiction:

Improvetask performanceVSAvoidcomputational resources

Core Design Contradiction:

ReliabilityVSUse of energy by moving object

Solution Approach 1:

The patent extracts only the relevant portions of vision data by filtering out extraneous frames and regions before processing. The system identifies and transmits only the subset of vision data corresponding to the subject matter, removing unnecessary background and irrelevant visual information to reduce computational resource consumption while maintaining task performance.

Inventive Principle:

Principle #2Taking out (Extraction)

Solution Approach 2:

The patent segments the vision data stream into relevant and extraneous portions. By dividing the continuous video stream into discrete frames and identifying only those frames containing relevant information, the system processes only the necessary segments of data, thereby reducing overall computational load while preserving task effectiveness.

Inventive Principle:

Principle #1Segmentation

2Productivity

If vision data is transmitted to remote servers, then generative content can be generated, but network resources are wasted transmitting extraneous data

Engineering Contradiction:

Improvegenerative content generationVSAvoidnetwork resources

Core Design Contradiction:

ProductivityVSLoss of energy

Solution Approach 1:

The patent performs preliminary filtering and processing of vision data locally before transmission to remote servers. By pre-processing the data to identify and isolate only the relevant portions, the system reduces the amount of data that needs to be transmitted over the network, thereby conserving network resources while still enabling effective generative content generation.

Inventive Principle:

Principle #10Preliminary action

Solution Approach 2:

The patent extracts and transmits only the essential vision data portions to remote servers for processing. By removing extraneous frames and visual information before transmission, the system minimizes network resource consumption while ensuring that the transmitted data is sufficient for generating accurate responsive content.

Inventive Principle:

Principle #2Taking out (Extraction)

3Loss of information

If all vision data is processed, then comprehensive information is available, but extraneous data dilutes relevant information

Engineering Contradiction:

Improveinformation completenessVSAvoidtask performance

Core Design Contradiction:

Loss of informationVSReliability

Solution Approach 1:

The patent extracts and isolates only the relevant information from the vision data stream. By identifying and separating the subject matter from background and extraneous elements, the system preserves the completeness of relevant information while eliminating diluting factors that would reduce task performance.

Inventive Principle:

Principle #2Taking out (Extraction)

Solution Approach 2:

The patent applies different processing quality to different parts of the vision data. By focusing computational attention only on the relevant portions of the data rather than uniformly processing all data, the system maintains high task performance while reducing the diluting effect of extraneous information.

Inventive Principle:

Principle #3Local quality

Data Source

PatentUS20250342205A1System(s) and method(s) for utilizing generative model(s) to generate content responsive to video data

Publication Date: 2025.11.06 GOOGLE LLC

AI summary

Some implementations relate to receiving a stream of vision data and a representation of a spoken utterance; processing, using a generative model (GM), first GM input to generate corresponding first GM output, the first GM input including at least the stream of vision data and the representation of the spoken utterance; determining, based on the corresponding first GM output, a subset of the stream of vision data; processing, using the GM, second GM input to generate corresponding second GM output, the second GM input including at least the subset of the stream of vision data and the representation of the spoken utterance; determining, based on the corresponding second GM output, responsive content, wherein the responsive content is responsive to the spoken utterance and the stream of vision data; and causing the responsive content to be rendered at the client device.