Adaptive Data Augmentation for Robust Self-Supervised Learning
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing data augmentation methods in self-supervised learning require manually preset parameters, leading to limited diversity in feature learning and reduced robustness, affecting the performance of downstream tasks.
Innovation Solution
A data augmentation device and method that employs a combination of data augmentation modules with fixed and non-fixed parameter groups, generating diverse augmentation images to enhance feature diversity and stability.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Ease of operation
If manually preset fixed parameters are used in data augmentation modules, then the device complexity is reduced and ease of operation is improved, but the feature diversity learned by the self-supervised learning module is limited and robustness deteriorates
Solution Approach 1:
The patent applies dynamics by making the parameter groups adaptive rather than fixed. The system dynamically adjusts parameters based on the input data characteristics, allowing the data augmentation modules to adapt their transformation parameters (such as rotation angles, scaling factors, or color jitter values) according to the specific properties of the data being processed, thereby improving both robustness and feature diversity while maintaining ease of operation through automated adaptation.
2Ease of manufacture
If manually designed parameters are used in the augmentation pipeline, then the ease of manufacture is improved, but the diversity of augmentation images is limited and the performance on downstream tasks deteriorates
Solution Approach 1:
The patent implements self-service by enabling the data augmentation system to automatically configure and adjust its own parameters without manual intervention. The system uses the input data itself to determine appropriate augmentation parameters, allowing the augmentation pipeline to serve itself by adapting to different data types and characteristics automatically, thus improving downstream task performance while maintaining ease of manufacture through automated parameter selection.
3Device complexity
If fixed parameter groups are used in all data augmentation modules, then the device complexity is reduced, but the adaptability to different data characteristics deteriorates
Solution Approach 1:
The patent applies local quality by allowing different parameter groups to have different levels of flexibility based on their specific requirements. Some parameter groups are configured as fixed values for simplicity, while others are set as non-fixed to adapt to local data characteristics. This selective approach enables the system to maintain low overall complexity while achieving adaptability where needed, by applying different parameter strategies to different parts of the augmentation pipeline based on their specific needs.
Data Source
AI summary
A data augmentation device including a processor is disclosed. The processor is configured to capture at least one sample image from an original image, and the processor is configured to input the at least one sample image to at least two data augmentation module, so as to generate at least two augmentation image group through the at least two data augmentation module. The at least two data augmentation module include a first data augmentation module and a second data augmentation module, in which a first parameter group of the first data augmentation module is a fixed value, and a second parameter group of the second data augmentation module is not the fixed value.