Unified AI Model Training Platform for Diverse Data Sources

Resolve Bottlenecks,
Find Innovative Solutions
Generate Solutions

Solution Overview

Problem

Industrial sectors face challenges in integrating diverse data sources for AI model applications due to the need for specialized knowledge and hardware, making it difficult for non-experts to connect and train AI models, which increases costs and reduces flexibility.

Innovation Solution

A universal platform with a connection module for data sources, a service module for driver management, a labeling module for data annotation, and a training module for AI model training, allowing untrained personnel to easily select, create, and adapt AI models for various applications.

Engineering Contradictions & Design Principles

VSEngineering Contradiction Analysis

1Reliability

If specialized drivers and hardware knowledge are used to integrate data sources, then integration reliability is improved, but device complexity and ease of operation deteriorate

Engineering Contradiction:
Improvedata source integration reliabilityVSAvoidintegration system complexity
Core Design Contradiction:
ReliabilityVSDevice complexity

Solution Approach 1:

The patent introduces a connection module as an intermediary component that sits between diverse data sources and the AI processing system. This connection module provides standardized interfaces and protocols, translating various data source formats into a unified structure. By acting as a mediator, it ensures reliable data integration while shielding users from the underlying complexity of hardware-specific drivers and protocols.

Inventive Principle:
Principle #24Intermediary (Mediator)

Solution Approach 2:

The connection module is designed with universal functionality to handle multiple types of data sources (cameras, sensors, microphones, etc.) through a single standardized interface. This multi-functional design allows the system to integrate diverse hardware without requiring separate specialized components for each data source type, thereby reducing overall system complexity while maintaining integration reliability.

Inventive Principle:
Principle #6Universality (Multi-functionality)

2Measurement precision

If specialized programming knowledge is required to train AI models, then model training precision is improved, but ease of operation deteriorates

Engineering Contradiction:
ImproveAI model training precisionVSAvoidmodel training ease of operation
Core Design Contradiction:
Measurement precisionVSEase of operation

Solution Approach 1:

The system implements self-service capabilities through automated model selection and training orchestration. The connection module automatically identifies suitable AI models for given data sources and training objectives, then configures and executes the training process without requiring users to manually program complex training pipelines. This automation maintains training precision while dramatically improving ease of operation for non-expert users.

Inventive Principle:
Principle #25Self-service

Solution Approach 2:

The system performs preliminary actions by pre-configuring AI model templates and training parameters based on the type of data source and desired application. Before actual training begins, the connection module pre-processes data sources, selects appropriate models, and sets up training environments, thereby reducing the operational burden on users while ensuring precise and optimized model training.

Inventive Principle:
Principle #10Preliminary action

3Adaptability or versatility

If multiple specialized frameworks are used for AI models, then model functionality is improved, but device complexity increases

Engineering Contradiction:
ImproveAI model functionalityVSAvoidframework integration complexity
Core Design Contradiction:
Adaptability or versatilityVSDevice complexity

Solution Approach 1:

The patent merges multiple AI model frameworks and functionalities into a single unified AI processing system. The connection module provides a standardized interface layer that can accommodate various AI frameworks (such as TensorFlow, PyTorch, or other specialized models) while presenting a consistent user experience. This consolidation maintains the versatility and adaptability of multiple frameworks while reducing the complexity of managing and integrating them separately.

Inventive Principle:
Principle #5Merging (Combining)

Data Source

PatentEP3901838A1System for providing trained ai models for different applications
Publication Date: 2021.10.27 VOLKSWAGEN AG
  • EP3901838A1 patent drawingFigure 1
  • EP3901838A1 patent drawingFigure 2
  • EP3901838A1 patent drawingFigure 3

AI summary

The invention relates to a system (10) for providing trained AI models (M) for various applications (A), comprising the following components: a connection module (1) for electrically and/or data-technically and/or programmatically connecting different data sources (D) to the system (10), a service module (2) for providing suitable drivers for the different data sources (D), a labeling module (3) for providing inputs in the data (B) of the different data sources (D) to form the neurons of an AI model (M), and corresponding outputs to train the AI ​​model (M), and a training module (4) for providing training services to train the AI ​​model (M) according to the inputs and outputs on the training module (4) or on an external IT infrastructure.