Universal Visual Explanation Map Generation for AI Models
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing visual intelligence models make it difficult for users to acquire visual explanation information, as the diverse purposes, types, and structures of these models complicate the extraction of such information.
Innovation Solution
A visual explanation information acquisition system that includes a diversification module, a visual intelligence module, an attribute analysis module, an explanation basis derivation module, and an explanation visualization module, which work together to generate visual explanation maps from various visual intelligence models independently of their purposes, types, and structures.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Loss of information
If visual explanation information is extracted by understanding the interior structure of a visual intelligence model, then the explanation information can be obtained, but the complexity of the process increases due to diverse purposes, types, and structures of models
Solution Approach 1:
The patent creates a universal visual explanation information acquisition system that can extract explanations from various types of visual intelligence models (object detection, segmentation, classification models) through a single standardized interface. The system uses common processing steps including input image preparation, model inference, output parsing, and explanation generation that work across different model architectures and purposes, eliminating the need for model-specific extraction procedures.
Solution Approach 2:
The patent introduces an intermediary explanation generation module that sits between the visual intelligence model and the user. This intermediary takes the model's output (detection results, segmentation masks, classification predictions) and automatically generates human-understandable visual explanations without requiring users to directly analyze or understand the complex interior structures of different models. The intermediary standardizes the explanation format regardless of the source model type.
2Device complexity
If a single system is used to acquire visual explanation information from various visual intelligence models, then the system complexity is reduced, but the adaptability to different model types and purposes must be maintained
Solution Approach 1:
The patent designs the acquisition system with universal components that handle multiple model types through standardized processes. The system includes a unified input interface that accepts images regardless of the target model type, a common inference execution module, and a standardized output parsing mechanism that works with detection results, segmentation masks, and classification predictions from various model architectures.
Solution Approach 2:
The patent employs parameter-based adaptation where the system adjusts processing parameters based on the detected model type and purpose. For example, it dynamically selects appropriate explanation generation methods, adjusts output format parameters, and modifies analysis thresholds based on whether the model is for object detection, semantic segmentation, or image classification, all while maintaining a single unified system architecture.
Data Source
AI summary
There are provided a method and a system for acquiring visual explanation information independent of the purpose, type, and structure of a visual intelligence model. The visual explanation information acquisition system of the visual intelligence model according to an embodiment may input N transformed images which are generated by diversifying an input image to a deep learning-based visual intelligence model and may acquire outputted results, may generate attributes of the visual intelligence model from the acquired results, may derive, from losses of the visual intelligence model which are calculated from the generated attributes, basic data for generating a visual explanation map for visually explaining a result derivation rationale of the visual intelligence model, and may generate a visual explanation map from the derived basic data. Accordingly, visual explanation information may be acquired from various visual intelligence models through one system independently of the purpose, type, and structure of the visual intelligence model.