Synthetic X-ray Data Generation for Neural Network Training
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
The limited availability of high-quality X-ray data for neural network learning, particularly in medical and customs detection fields, due to the scarcity of X-ray images and the need for additional processing and annotation, hinders the development of effective artificial intelligence models.
Innovation Solution
An apparatus and method for generating X-ray data involving a processor that extracts 3D data, projects it onto a 2D plane, performs data augmentation, composes it with background data, and applies post-processing to create output data suitable for neural network learning, utilizing a Generative Adversarial Network (GAN) for improved heterogeneity resolution.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Reliability
If X-ray data is acquired in actual environment, then data authenticity is improved, but data quantity and diversity are limited
Solution Approach 1:
The patent creates synthetic X-ray images by copying and transforming real 3D object data through virtual projection and composition processes. The system generates artificial training data that mimics real X-ray images while overcoming the limitations of actual environment acquisition, thereby increasing data quantity and diversity without compromising the essential characteristics needed for neural network training.
2Manufacturing precision
If additional processing and annotation are performed on acquired data, then learning data quality is improved, but processing time and complexity increase
Solution Approach 1:
The patent performs data processing and annotation in advance during the synthetic data generation process. By pre-processing the 3D object data and creating annotated training images before they are needed for neural network training, the system eliminates the need for time-consuming post-acquisition processing, thereby reducing processing time while maintaining high data quality.
3Adaptability or versatility
If more X-ray images are acquired to increase training data diversity, then neural network learning performance is improved, but acquisition difficulty and cost increase
Solution Approach 1:
The patent transitions from acquiring multiple 2D X-ray images from different angles to generating diverse training data by projecting 3D object data onto 2D planes with various parameters. This dimensional approach allows the system to create unlimited diverse images from a single 3D data source, eliminating the need for complex multi-angle acquisition setups while achieving the same diversity goal.
Data Source
AI summary
The apparatus for an X-ray data generation according to an embodiment of the inventive concept includes a processor that receives 3D data to generate output data and a buffer, and the processor includes an extraction unit that extracts raw object data from the 3D data and projects the raw object data onto a 2D plane to generate first object data, an augmentation unit that performs data augmentation on the first object data to generate second object data, a composition unit that synthesizes the second object data and background data to generate composite data, and a post-processing unit that performs post-processing on the composite data to generate the output data, and the buffer stores a plurality of parameters related to generation of the first object data, the second object data, the composite data, and the output data.