Sample Label Generation Using Icon Variation and Template Placement
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
The existing methods for creating datasets for computer vision systems, particularly in medical and pharmaceutical industries, are labor-intensive and time-consuming due to the need for manual acquisition and annotation of a large number of images, and they often result in unbalanced datasets, hindering the deployment of machine learning models.
Innovation Solution
A method and system for generating high-fidelity sample labels by receiving user inputs for icons and backgrounds, applying pre-augmentation operations, and using an occupancy region map to optimize the placement of icons, thereby synthesizing and annotating images automatically, reducing manual effort and time.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If manual acquisition and annotation of images is used to create training datasets, then the quality and accuracy of annotated data is improved, but the time consumption and manual effort increase significantly
Solution Approach 1:
The patent uses template-based copying where standardized label templates are reused and populated with different icon sets. Instead of manually creating each label from scratch, the system copies proven template structures and automatically fills them with selected icons, maintaining consistency and quality while dramatically reducing manual effort
Solution Approach 2:
The patent performs preliminary organization of icons into categories and creates standardized label templates in advance. This pre-processing allows the dataset generation to proceed by simply selecting and combining pre-organized elements, eliminating the need for real-time manual annotation decisions during dataset creation
2Adaptability or versatility
If a large number of diverse images are acquired manually to create balanced datasets, then the diversity and balance of the dataset is improved, but the complexity and time required for data collection increases
Solution Approach 1:
The patent segments the dataset creation process into independent modules: icon selection, template selection, combination logic, and generation. Each module handles a specific aspect of diversity (icon variety, template variety, arrangement variety) independently, making the overall system manageable despite the complexity of generating diverse datasets
Solution Approach 2:
The patent creates a universal template system that can accommodate multiple types of icons and labels through a single framework. The same template structure serves multiple purposes by accepting different icon combinations, eliminating the need for separate collection processes for different label types
3Ease of operation
If existing ML models or conventional image processing techniques are used for automatic annotation, then the manual effort is reduced, but the computational resources and time required increase
Solution Approach 1:
The patent uses template copying instead of computational annotation. Pre-defined label templates are directly instantiated with selected icons, eliminating the need for computationally intensive image processing or ML model inference that would be required to automatically detect and annotate icons in existing images
Solution Approach 2:
The system performs self-service dataset generation by automatically combining selected icons with templates according to predefined rules. No external ML models or complex processing pipelines are needed - the system generates annotated labels directly through rule-based combination of pre-prepared elements
Data Source
AI summary
The method and system for generating sample labels is disclosed. The method includes receiving, from a user, a selection of: one or more icons from a plurality of icons; and one or more backgrounds from a plurality of backgrounds. The method further includes creating a first plurality of variation-icons corresponding to each of the one or more icons, by applying one or more pre-augmentation operations to each of the one or more icons and selecting a set of variation-icons from a second plurality of variation-icons corresponding to the one or more icons, based on dimensions of each variation-icon of the set of variation-icons and predefined dimensions of a sample label template. The method further includes applying a background to the sample label template and positioning the set of variation-icons in the sample label template over the background, to generate a sample label.