Automated Analytics for Non-Tabular Image Data
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Current data analytics tools for image data are expensive, time-consuming, and require specialized expertise, limiting their accessibility and efficiency in various industries.
Innovation Solution
The development of automated machine learning techniques and computer vision tools that use two-stage models to analyze and interpret image data, determining feature importance and generating visual explanations, while also detecting data drift and integrating with non-image data for enhanced predictive capabilities.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Ease of operation
If automated machine learning techniques are used for image data analytics, then the need for specialized expertise is reduced and development costs are lowered, but the complexity of processing non-tabular image data increases
Solution Approach 1:
The patent segments image data into constituent image features (e.g., individual pixels, regions, or objects) and processes them separately through automated machine learning models. This segmentation allows the system to handle complex non-tabular image data by breaking it down into manageable units that can be analyzed independently and then aggregated, reducing the operational complexity while maintaining processing capability.
Solution Approach 2:
The patent introduces an intermediary layer that translates non-tabular image data into a tabular format with constituent features. This intermediary representation layer acts as a bridge between the complex image data and the automated machine learning models, enabling standard ML techniques to process image data without requiring specialized computer vision expertise.
2Measurement precision
If traditional data analytics processes are used for image data, then measurement precision can be maintained, but the duration of action and time required for analysis increases
Solution Approach 1:
The patent performs preliminary extraction of constituent image features from image data before the main analytics processing. By pre-processing the image data to identify and extract relevant features (such as edges, textures, or object characteristics) in advance, the system reduces the computational burden during the actual analytics phase, thereby maintaining measurement precision while significantly reducing the overall analysis time.
Solution Approach 2:
The patent extracts only the necessary constituent features from the full image data rather than processing the entire image. This selective extraction of relevant features (taking out only what is needed) maintains the precision required for accurate analytics while reducing the volume of data that needs to be processed, thus decreasing the time required for analysis.
3Measurement precision
If feature importance scoring is performed for all image features, then the accuracy of predictive analytics is improved, but the computational power and processing time increase
Solution Approach 1:
The patent applies local quality by determining feature importance scores selectively for specific constituent image features rather than uniformly for all features. The system identifies and focuses computational resources on the most relevant features (those with higher importance scores), allowing accurate predictive analytics to be achieved with reduced computational power by concentrating analysis on locally important features rather than processing all features with equal intensity.
Data Source
AI summary
Automated data analytics techniques for non-tabular data sets may include methods and systems for (1) automatically developing models that perform tasks in the domains of computer vision, audio processing, speech processing, text processing, or natural language processing; (2) automatically developing models that analyze heterogeneous data sets containing image data and non-image data, and/or heterogeneous data sets containing tabular data and non-tabular data; (3) determining the importance of an image feature with respect to a modeling task, (4) explaining the value of a modeling target based at least in part on an image feature, and (5) detecting drift in image data. In some cases, multi-stage models may be developed, wherein a pre-trained feature extraction model extracts low-, mid-, high-, and/or highest-level features of non-tabular data, and a data analytics models uses those features (or features derived therefrom) to perform a data analytics task.