Image Localizability Classification for Context-Aware Asset Screening
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Conventional systems for localizing assets in organizational documentation are inefficient and inaccurate due to the failure to detect text elements like font color, text size, and background color, and lack of context consideration, leading to over-inclusive and slow workflows.
Innovation Solution
A Convolutional Neural Network (CNN)-based classifier is trained to analyze assets for localization needs, identifying features and contexts to determine whether assets should be localized, thereby excluding unnecessary images from the localization process.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Extent of automation
If conventional OCR systems are used to identify assets for localization, then the process can be automated, but the accuracy is poor and it causes workflow bottlenecks
Solution Approach 1:
The patent replaces conventional mechanical OCR systems with a deep learning-based CNN classifier that uses convolutional neural networks to analyze image features. This substitution enables the system to accurately identify assets requiring localization by analyzing visual context, font characteristics, and document structure, thereby achieving both automation and high reliability simultaneously.
Solution Approach 2:
The patent changes the approach from text-based OCR to image-based deep learning classification. By transforming the input from extracted text to full image analysis with multiple feature parameters (visual context, font properties, layout), the system achieves superior accuracy in determining which assets require localization while maintaining automation.
2Quantity of substance
If all assets are included in the localization process, then completeness is maintained, but productivity decreases due to unnecessary processing
Solution Approach 1:
The patent extracts and removes assets from the localization process that do not require it, based on the CNN classifier's prediction. By identifying and excluding unnecessary assets (such as those with no text or where localization is not needed), the system maintains completeness for relevant assets while significantly improving productivity by avoiding wasted processing on irrelevant ones.
Solution Approach 2:
The system applies partial action by selectively processing only the subset of assets that the classifier determines require localization. This partial processing approach, guided by the deep learning model's accuracy, prevents excessive processing of all assets while ensuring that necessary localization tasks are completed, thereby optimizing productivity.
3Reliability
If manual review of each asset is performed, then accuracy is high, but time consumption increases
Solution Approach 1:
The patent implements self-service by enabling the system to automatically classify assets for localization using the trained CNN model. The system performs self-evaluation of asset localizability through deep learning, eliminating the need for manual review while maintaining high accuracy. This self-service mechanism significantly reduces time consumption compared to manual inspection of each asset.
Data Source
AI summary
In a computer-implemented workflow, a submission of an asset localized for a first location is received. The asset may be intended for dissemination to a second location. A trained neural network is applied to the asset to determine a probability of recommending localization of the asset for the second location. This determination can be based on a plurality of features indicating contextual aspects of a document, which are identified in accordance with a plurality of transformations performed on the asset utilizing the trained neural network. Responsive to determining that the probability satisfies a condition, such as being a percentage above a threshold value, a recommendation is provided to exclude the asset from being localized to the second location.