Driving Image Identification With Text-Picture Interaction

Resolve Bottlenecks,
Find Innovative SolutionsGenerate Solutions

Solution Overview

Problem

Conventional image capture devices lack effective methods to identify driving images for implementing driving assistance functions.

Innovation Solution

An identification system utilizing a storage device and processor with an identification module that includes a text encoder, computing module, and an attentive pairwise interaction network model to convert input data between text and picture formats, trained using contrastive language-image pre-training and attentive pairwise interaction network models.

Engineering Contradictions & Design Principles

VSEngineering Contradiction Analysis

1Adaptability or versatility

If conventional image capture devices are used, then image recording function is provided, but effective identification of driving images cannot be implemented

Engineering Contradiction:

Improveidentification capabilityVSAvoidsystem structure

Core Design Contradiction:

Adaptability or versatilityVSDevice complexity

Solution Approach 1:

The patent combines the text encoder, computing module, and attentive pairwise interaction network model into an integrated identification module that works together to process image and text data, enabling comprehensive identification functionality within a unified system architecture

Inventive Principle:

Principle #5Merging (Combining)

Solution Approach 2:

The identification module is designed to handle multiple types of data (image data and text data) and perform various identification tasks including driving scene recognition, object detection, and text-image matching, making the system versatile across different driving assistance applications

Inventive Principle:

Principle #6Universality (Multi-functionality)

2Adaptability or versatility

If simple image recording is implemented, then device complexity is low, but driving assistance functions cannot be effectively provided

Engineering Contradiction:

Improvedriving assistance functionalityVSAvoidprocessing system

Core Design Contradiction:

Adaptability or versatilityVSDevice complexity

Solution Approach 1:

The identification module is divided into distinct functional components: a text encoder for processing text data, a computing module for executing calculations, and an attentive pairwise interaction network model for analyzing relationships between images and text, allowing each component to be optimized independently while contributing to overall driving assistance functionality

Inventive Principle:

Principle #1Segmentation

Solution Approach 2:

The attentive pairwise interaction network model acts as an intermediary that processes and integrates information from both image data and text data, enabling the system to perform complex driving assistance tasks by mediating between different data types and producing coordinated identification results

Inventive Principle:

Principle #24Intermediary (Mediator)

Data Source

PatentUS20250308248A1Identification system and identification method

Publication Date: 2025.10.02 VIA TECH INC

AI summary

An identification system and an identification method are provided. The identification system includes a storage device and a processor. The storage device stores an identification module. The identification module includes a text encoder, a computing module, and an attentive pairwise interaction network model. The processor is coupled to the storage device and executes the identification module. The processor inputs the input data to the identification module, so that the identification module generates output data according to the input data. The input data is one of text data and picture data, and the output data is the other one of text data and picture data. Encoding data output by the text encoder or the attentive pairwise interaction network model is used as the input data of the computing module. The computing module generates output data according to the input data.