Attention-Guided Visual Perception for High-Resolution ROI Detection

Resolve Bottlenecks,
Find Innovative SolutionsGenerate Solutions

Solution Overview

Problem

Automatic object detection systems face challenges in processing high-definition visual data efficiently, particularly in varying lighting conditions and diverse visual imagery, while requiring substantial computing resources.

Innovation Solution

A perception system with a controller that includes a subsampling module, an object detection module, and an attention module, which generates a rescaled whole image frame and identifies regions of interest, allowing for the creation of high-resolution images within selected areas of interest, reducing the computational load by processing lower resolution peripheries and higher resolution areas of interest.

Engineering Contradictions & Design Principles

VSEngineering Contradiction Analysis

1Measurement precision

If high-definition visual data is processed to maintain image quality, then measurement precision is improved, but computing resources required increase

Engineering Contradiction:

Improveimage qualityVSAvoidcomputing resources

Core Design Contradiction:

Measurement precisionVSQuantity of substance

Solution Approach 1:

The system applies different quality levels to different regions of the image. Regions of interest are processed at high resolution to maintain measurement precision, while peripheral regions are processed at lower resolution to reduce computing resource requirements. This is achieved through selective subsampling and region-of-interest identification mechanisms.

Inventive Principle:

Principle #3Local quality

Solution Approach 2:

The visual data processing is divided into multiple stages and regions. The system segments the image into regions of interest and peripheral regions, processes them through different processing pipelines, and combines the results. This segmentation allows high-definition processing only where necessary, reducing overall computing resource consumption while maintaining image quality in critical areas.

Inventive Principle:

Principle #1Segmentation

2Measurement precision

If full-resolution images are processed, then measurement precision is improved, but productivity decreases

Engineering Contradiction:

Improvedetection accuracyVSAvoidprocessing speed

Core Design Contradiction:

Measurement precisionVSProductivity

Solution Approach 1:

The system processes only critical regions at full resolution to maintain detection accuracy, while processing peripheral regions at lower resolution to improve processing speed. This selective quality approach ensures that measurement precision is maintained where it matters most while overall productivity is enhanced through reduced computational burden.

Inventive Principle:

Principle #3Local quality

Solution Approach 2:

The system applies full processing power partially - only to regions of interest - rather than uniformly to the entire image. This partial action approach maintains detection accuracy in critical areas while reducing overall processing time and increasing productivity through selective application of high-resolution processing.

Inventive Principle:

Principle #16Partial or excessive action

3Measurement precision

If high-definition sensing is used, then measurement precision is improved, but use of energy increases

Engineering Contradiction:

Improvesensing resolutionVSAvoidcomputational energy

Core Design Contradiction:

Measurement precisionVSUse of energy by moving object

Solution Approach 1:

The system maintains high sensing resolution in regions of interest while using lower resolution for peripheral regions, thereby reducing the total computational energy required to process the visual data. This local quality approach ensures measurement precision is preserved where needed while energy consumption is reduced through selective processing.

Inventive Principle:

Principle #3Local quality

4Quantity of substance

If lower resolution processing is used, then computing resources are reduced, but measurement precision deteriorates

Engineering Contradiction:

Improvecomputing resourcesVSAvoidimage quality

Core Design Contradiction:

Quantity of substanceVSMeasurement precision

Solution Approach 1:

The system strategically applies lower resolution processing only to peripheral regions where measurement precision requirements are reduced, while maintaining high resolution in regions of interest. This local differentiation allows computing resources to be reduced overall while measurement precision is preserved in critical areas through selective high-resolution processing.

Inventive Principle:

Principle #3Local quality

Data Source

PatentUS11989947B2Perception system with attention module for processing visual data

Publication Date: 2024.05.21 GM GLOBAL TECHNOLOGY OPERATIONS LLC

AI summary

A perception system is adapted to receive visual data from a camera and includes a controller having a processor and tangible, non-transitory memory on which instructions are recorded. A subsampling module, an object detection module and an attention module are each selectively executable by the controller. The controller is configured to sample an input image from the visual data to generate a rescaled whole image frame, via the subsampling module. The controller is configured to extract feature data from the rescaled whole image frame, via the object detection module. A region of interest in the rescaled whole image frame is identified, based on an output of the attention module. The controller is configured to generate a first image based on the rescaled whole image frame and a second image based on the region of interest, the second image having a higher resolution than the first image.