Image and Boundary Label Augmentation for ML Training Accuracy
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing image augmentation methods for machine learning models often fail to adequately augment boundary labels, requiring manual processes to ensure accuracy, which is inefficient and time-consuming.
Innovation Solution
A method and system that automatically augments both images and boundary labels using predefined augmentation parameters, generating augmented images with corresponding boundary labels that align with a two-dimensional coordinate plane, allowing for efficient training of machine learning models.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Productivity
If image augmentation is performed without label augmentation, then image processing speed is improved, but label accuracy deteriorates requiring manual processes
Solution Approach 1:
The patent creates a blank image as a copy template and defines a model boundary label on it. This model label is then systematically applied to all augmented images through coordinate transformation, eliminating the need for manual label creation while maintaining accuracy. The copying principle allows automated generation of accurate labels across all augmented image variants.
Solution Approach 2:
The patent performs preliminary action by defining the model boundary label on a blank image before augmentation occurs. The model label serves as a template that is subsequently transformed and applied to all augmented images. This preliminary definition of label coordinates and dimensions enables automated label generation for all augmented variants without manual intervention.
2Measurement precision
If manual label augmentation is performed, then label accuracy is improved, but time consumption increases
Solution Approach 1:
The system performs self-service by automatically generating augmented boundary labels through coordinate transformation of the model label. The automated process transforms the model label coordinates according to the same augmentation parameters applied to images, producing accurate labels without human intervention. This self-service mechanism eliminates manual label augmentation while maintaining precision.
Solution Approach 2:
The patent applies parameter changes by transforming the model label coordinates using the same augmentation parameters (rotation, translation, scaling) applied to the images. By changing the coordinate parameters systematically based on augmentation transformations, the system generates accurate labels for all augmented images automatically, eliminating manual processes and reducing time consumption.
3Device complexity
If boundary labels are not augmented with images, then processing complexity is reduced, but model training effectiveness deteriorates
Solution Approach 1:
The patent segments the label processing into two independent components: image augmentation and label augmentation. The model boundary label is defined separately on a blank image and then transformed independently using coordinate calculations based on augmentation parameters. This segmentation allows both images and labels to be augmented systematically, ensuring model training effectiveness while maintaining manageable processing complexity through modular operations.
Solution Approach 2:
The patent adds another dimension to the processing by introducing coordinate transformation space for labels. Instead of simply copying labels, the system transforms label coordinates in a mathematical space based on augmentation parameters (rotation angles, translation vectors, scaling factors). This dimensional transformation ensures labels remain accurately aligned with augmented images, improving training effectiveness without excessive complexity.
Data Source
AI summary
A method includes obtaining an authentic image of an assembly and a boundary label provided with the authentic image. The boundary label is associated with a selected region of the authentic image depicting a selected object. The method includes generating an augmented image based on the authentic image and an augmentation model employing one or more augmentation parameters, defining a model boundary label on a blank image at a region that correlates with the selected region of the authentic image, generating an augmented blank image based on the one or more augmentation parameters employed for the augmented image, identifying, as an augmented boundary label associated with augmented image, the model boundary label in the augmented blank image, and outputting an augmented image data, wherein the augmented image data incudes data indicative of the augmented image and of the augmented boundary label.