Automated Image Label Generation Using Template and Reference Data
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
The labor-intensive and time-consuming process of creating large amounts of labeled data for machine learning research, particularly in image processing, hinders efficient training of image processing models.
Innovation Solution
An image processing method that generates labeled images by using a template label image and a plurality of reference images, where the processor generates a target image with a contour and color/texture based on the template label image and reference images, enabling automatic generation of pixel-wise labeled images for object segmentation.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If manual labeling is used to create labeled data for machine learning, then the quality and accuracy of training data can be ensured, but the process becomes labor-intensive and time-consuming
Solution Approach 1:
The patent applies preliminary action by pre-defining template label images that contain standardized label configurations for different object types. These templates are created in advance and can be automatically applied to new images, eliminating the need for manual labeling while maintaining consistency and accuracy. The system retrieves appropriate templates based on image content and applies them automatically, significantly reducing labeling time while preserving quality.
Solution Approach 2:
The patent uses copying by replicating proven label templates across multiple images. Instead of manually creating labels for each image, the system copies pre-validated label configurations from template images to target images based on object detection results. This ensures consistent labeling accuracy while dramatically reducing the time and labor required, as the same high-quality label patterns are reused across the dataset.
2Quantity of substance
If large amounts of labeled data are manually created, then the training data volume increases, but the labor and time costs increase proportionally
Solution Approach 1:
The system performs preliminary action by pre-preparing a library of template label images with various object configurations and label types. When processing new images, the system automatically selects and applies appropriate templates based on detected objects, enabling rapid generation of large volumes of labeled data without proportional increases in manual labor. This approach maintains high productivity while scaling data volume.
Solution Approach 2:
The patent applies universality by designing template label images that can serve multiple purposes and be applied to different image types. A single template library can generate labels for various objects and scenarios, allowing the system to produce diverse labeled datasets from a unified template system. This multi-functionality enables efficient scaling of data volume without requiring separate manual labeling processes for each case.
3Productivity
If automatic label generation is implemented, then the processing speed increases, but the accuracy and quality of labels may deteriorate
Solution Approach 1:
The patent uses copying to replicate proven, high-quality label templates automatically across numerous images. By copying pre-validated label configurations rather than generating labels from scratch, the system maintains consistent label quality and accuracy while achieving high processing speeds. The template copying mechanism ensures that only proven, accurate label patterns are applied, preventing quality deterioration despite automated high-volume processing.
Solution Approach 2:
The system implements feedback by using detection results to automatically select appropriate templates and apply them to images. The feedback loop ensures that the correct template is chosen based on the detected object characteristics, maintaining label accuracy. This automated feedback mechanism allows rapid processing while preserving quality through intelligent template selection and application validation.
Data Source
AI summary
An image processing training method includes the following steps. A template label image is obtained, in which the template label image comprises a label corresponding to a target. A plurality of first reference images are obtained, in which each of the first reference images comprises object image data corresponding to the target. A target image according to the template label image and the first reference images is generated, in which the target image comprises a generated object, a contour of the generated object is generated according to the template label image, and a color or a texture of the target image is generated according to the first reference images.