Neural Network Model Construction for Object Detection

Resolve Bottlenecks,
Find Innovative Solutions
Generate Solutions

Solution Overview

Problem

Current machine learning and deep learning technologies, particularly neural networks, require specialized knowledge and are time-consuming to configure and train for object detection, limiting their application to specific domains and requiring extensive data annotation, which is labor-intensive and often ineffective for non-standardized training data.

Innovation Solution

A computer-implemented method for constructing a neural network model for object detection in unprocessed images using a graphical user interface, allowing users to annotate objects, associate classes, and train collective model variables, enabling object detection with reduced technical expertise and time, leveraging cloud infrastructure and CNNs for scalability across various domains.

Engineering Contradictions & Design Principles

VSEngineering Contradiction Analysis

1Adaptability or versatility

If pre-trained models are used for object detection, then object detection capability is provided, but the models are limited to narrow fields and require extensive data annotation

Engineering Contradiction:
Improveapplication field rangeVSAvoiddata annotation time
Core Design Contradiction:
Adaptability or versatilityVSLoss of time

Solution Approach 1:

The system performs preliminary actions by using the neural network to automatically generate annotations for training data before the actual model training process. This preliminary annotation step eliminates the need for manual data annotation, resolving the contradiction between adaptability and time loss by preparing training data in advance through automated processing.

Inventive Principle:
Principle #10Preliminary action

Solution Approach 2:

The neural network performs self-service by automatically annotating training data without human intervention. The system uses the trained model to generate its own training data annotations, creating a self-sustaining process that eliminates manual annotation work and enables rapid adaptation to new application fields.

Inventive Principle:
Principle #25Self-service

2Measurement precision

If neural networks are configured and trained with specialized knowledge, then accurate object detection is achieved, but the process requires extensive specialist knowledge and time

Engineering Contradiction:
Improveobject detection accuracyVSAvoidconfiguration complexity
Core Design Contradiction:
Measurement precisionVSEase of operation

Solution Approach 1:

The system enables self-service operation where the neural network automatically performs data annotation and model training without requiring specialist knowledge from the user. The automated processes handle the complex configuration and training steps, making the system easy to operate while maintaining high detection accuracy.

Inventive Principle:
Principle #25Self-service

Solution Approach 2:

The system introduces an intermediary automated annotation process that mediates between the user and the complex neural network training process. This intermediary layer handles the specialized knowledge requirements internally, allowing users to operate the system without needing deep understanding of neural network configuration and training.

Inventive Principle:
Principle #24Intermediary (Mediator)

3Ease of manufacture

If standardized training data is used, then model training is simplified, but the training data is limited in size and application fields

Engineering Contradiction:
Improvemodel training simplicityVSAvoidtraining data applicability
Core Design Contradiction:
Ease of manufactureVSAdaptability or versatility

Solution Approach 1:

The system applies dynamics by transforming static standardized training data into dynamic, domain-specific training data through automated annotation. The neural network adapts the training data to match the specific application domain while maintaining the simplicity of the training process, resolving the contradiction between ease of manufacture and adaptability.

Inventive Principle:
Principle #15Dynamics

Solution Approach 2:

The system performs preliminary adaptation of training data to the target domain before model training begins. By pre-processing and annotating data specific to the application field in advance, the system maintains training simplicity while ensuring the training data is highly applicable to the specific domain.

Inventive Principle:
Principle #10Preliminary action

Data Source

PatentEP3452959B1Model construction in a neural network for object detection
Publication Date: 2023.01.25 SCOPITO APS
  • EP3452959B1 patent drawingFigure 1
  • EP3452959B1 patent drawingFigure 2
  • EP3452959B1 patent drawingFigure 3

AI summary

The present invention relates to a computer-implemented method for constructing a model in a neural network for object detection in an unprocessed image, where the construction may be performed based on at least one image training batch. The model is constructed by training one or more collective model variables in the neural net- work to classify the individual annotated objects as a member of an object class. The model in combination with the set of specifications when implemented in a neural network is capable of object detection in an unprocessed image with probability of object detection.