AVM Object Detection With Confidence-Guided Re-Detection

Resolve Bottlenecks,
Find Innovative Solutions
Generate Solutions

Solution Overview

Problem

Autonomous driving technologies face challenges in reducing computing power consumption without compromising performance, leading to potential safety issues due to increased errors.

Innovation Solution

A method utilizing confidence scores generated in parallel with the object detection process to selectively re-detect objects in regions with low confidence scores, optimizing resource allocation and improving detection accuracy while saving computing power.

Engineering Contradictions & Design Principles

VSEngineering Contradiction Analysis

1Use of energy by moving object

If computing powers are reduced to save energy, then energy consumption is reduced, but detection accuracy and safety are worsened

Engineering Contradiction:
Improvecomputing power consumptionVSAvoidautonomous driving safety
Core Design Contradiction:
Use of energy by moving objectVSReliability

Solution Approach 1:

The patent divides the autonomous driving system into multiple processing units or modules that can independently handle different detection tasks. By segmenting the computing workload, the system can allocate resources more efficiently, performing comprehensive safety-critical detections only where needed while using simplified processing for less critical functions, thus reducing overall computing power consumption without compromising safety

Inventive Principle:
Principle #1Segmentation

Solution Approach 2:

The patent applies different processing qualities to different regions or aspects of the autonomous driving system. Critical safety-related detection tasks receive full computational resources and high processing quality, while non-critical tasks use reduced computational power. This local differentiation allows the system to maintain safety where it matters most while reducing energy consumption in other areas

Inventive Principle:
Principle #3Local quality

2Reliability

If comprehensive object detection is performed to ensure safety, then detection accuracy is improved, but computing power consumption increases

Engineering Contradiction:
Improveobject detection accuracyVSAvoidcomputing power consumption
Core Design Contradiction:
ReliabilityVSUse of energy by moving object

Solution Approach 1:

The patent implements a multi-stage detection approach where not all detection tasks are performed at full computational intensity simultaneously. Instead, the system performs partial detections in early stages with lower computational requirements, then applies more intensive processing only to regions or objects that require higher accuracy, avoiding excessive computing power consumption while maintaining necessary detection accuracy

Inventive Principle:
Principle #16Partial or excessive action

Solution Approach 2:

The patent performs preliminary object detection and filtering before applying more computationally intensive analysis. By first identifying potential objects of interest using lighter processing, the system can then focus comprehensive detection resources only on relevant targets, reducing overall computing power consumption while ensuring accurate detection when needed

Inventive Principle:
Principle #10Preliminary action

Data Source

PatentEP3690735B1Method and device for attention-driven resource allocation by using AVM to thereby achieve safety of autonomous driving
Publication Date: 2023.10.11 STRADVISION
  • EP3690735B1 patent drawingFigure 1
  • EP3690735B1 patent drawingFigure 2
  • EP3690735B1 patent drawingFigure 3

AI summary

A method for achieving better performance in an autonomous driving while saving computing powers, by using confidence scores representing a credibility of an object detection which is generated in parallel with an object detection process is provided. And the method includes steps of: (a) a computing device acquiring at least one circumstance image on surroundings of a subject vehicle, through at least one panorama view sensor installed on the subject vehicle; (b) the computing device instructing a Convolutional Neural Network(CNN) to apply at least one CNN operation to the circumstance image, to thereby generate initial object information and initial confidence information on the circumstance image; and (c) the computing device generating final object information on the circumstance image by referring to the initial object information and the initial confidence information.