Adaptive Data Path for Computer Vision Processing

Resolve Bottlenecks,
Find Innovative Solutions
Generate Solutions

Solution Overview

Problem

Traditional mobile devices have a fixed data path optimized for imaging applications, which results in excessive processing overhead and reduced battery life when executing computer-vision applications, as they lack the necessary filtering capabilities and adaptability to switch between high and low-processing states efficiently.

Innovation Solution

An adaptive data path is implemented, utilizing a computer-vision processing unit with specialized instruction-based filters that can select and apply relevant image processing functions based on the specific computer-vision application, offloading processing from the application processing unit and optimizing data delivery to reduce power consumption and processing demands.

Engineering Contradictions & Design Principles

VSEngineering Contradiction Analysis

1Productivity

If a fixed data path optimized for imaging applications is used, then imaging application performance is improved, but computer-vision application processing overhead increases and battery life decreases

Engineering Contradiction:
Improveimaging application performanceVSAvoidbattery life
Core Design Contradiction:
ProductivityVSUse of energy by moving object

Solution Approach 1:

The data path is transformed from a fixed configuration to a dynamic, reconfigurable system that can adapt its processing pipeline based on the type of application being executed. The processing unit can dynamically select and activate specific filtering functions and data flow paths depending on whether an imaging or computer-vision application is running, thereby optimizing performance for the current task while minimizing energy consumption.

Inventive Principle:
Principle #15Dynamics

Solution Approach 2:

The system changes operational parameters of the data path including which filtering functions are active, the level of processing detail, and the routing of data flows. By adjusting these parameters based on application type, the system achieves high performance for imaging applications while reducing processing overhead and power consumption for computer-vision applications.

Inventive Principle:
Principle #35Parameter changes

2Productivity

If a fixed data path optimized for imaging applications is used, then imaging application performance is improved, but computer-vision application processing overhead increases

Engineering Contradiction:
Improveimaging application performanceVSAvoidprocessing overhead
Core Design Contradiction:
ProductivityVSDevice complexity

Solution Approach 1:

The data path configuration changes dynamically based on application requirements. For computer-vision applications, the system can deactivate unnecessary imaging-optimized processing stages and filtering functions, simplifying the processing pipeline and reducing overhead while maintaining the capability to provide full functionality when imaging applications are running.

Inventive Principle:
Principle #15Dynamics

3Reliability

If full image processing functions are applied to all images, then processing completeness is improved, but power consumption increases

Engineering Contradiction:
Improveprocessing completenessVSAvoidpower consumption
Core Design Contradiction:
ReliabilityVSUse of energy by moving object

Solution Approach 1:

The system extracts and applies only the necessary subset of processing functions required for each specific application. For computer-vision applications, unnecessary imaging processing functions are removed from the active pipeline, reducing power consumption while maintaining sufficient processing completeness for the specific task at hand.

Inventive Principle:
Principle #2Taking out (Extraction)

Solution Approach 2:

Instead of applying the full set of processing functions to all images regardless of application needs, the system applies partial processing - only the specific functions required for the current application. This avoids excessive processing that would waste energy while ensuring sufficient processing for reliable results.

Inventive Principle:
Principle #16Partial or excessive action

Data Source

PatentEP2973386B1Adaptive data path for computer-vision applications
Publication Date: 2018.12.05 QUALCOMM INC
  • EP2973386B1 patent drawingFigure 1
  • EP2973386B1 patent drawingFigure 2
  • EP2973386B1 patent drawingFigure 3

AI summary

Embodiments of the present invention provide an adaptive data path for computer-vision applications. Utilizing techniques provided herein, the data path can adapt to the needs of a computer-vision application to provide the needed data. The data path can be adapted by applying one or more filters to image data from one or more sensors. Some embodiments may utilize a computer-vision processing unit comprising a specialized instruction-based, in-line processor capable of interpreting commands from a computer-vision application.