Deep Learning Data Expansion via Image Aliasing
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
The limited availability and high cost of manually annotated data sets hinder the widespread application of deep learning networks, as existing public data sets are not suitable for practical scenarios involving specific and fine-grained objects.
Innovation Solution
An automatic data enhancement method and system that uses a mechanical arm and camera controlled by an embedded device to collect and process video data, extracting seed images, performing image enhancement operations, and generating composite images to create a diverse and authentic training data set for deep learning networks, such as Yolov3, thereby reducing the need for extensive original materials and lowering costs.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If manually annotated data sets are used for deep learning training, then recognition accuracy is improved, but time and capital costs increase significantly
Solution Approach 1:
The patent uses image aliasing enhancement to generate composite images that copy and combine features from multiple seed images. This creates synthetic training data that mimics real object variations without requiring manual annotation of each sample, thereby maintaining recognition accuracy while reducing data preparation time and costs.
Solution Approach 2:
The system performs preliminary data enhancement by generating composite images from seed images before actual training. This preliminary action creates a expanded data set with diverse object features, reducing the need for extensive manual annotation later and accelerating the overall training process.
2Measurement precision
If manually annotated data sets are used for deep learning training, then recognition accuracy is improved, but capital costs increase significantly
Solution Approach 1:
The patent uses image aliasing enhancement to generate composite images that copy and combine features from multiple seed images. This creates synthetic training data that mimics real object variations without requiring manual annotation of each sample, thereby maintaining recognition accuracy while reducing data preparation time and costs.
Solution Approach 2:
The system performs automatic data enhancement and composite image generation without human intervention. The automated pipeline extracts seed images, applies enhancement operations, and generates composite images autonomously, eliminating the need for expensive manual annotation services while maintaining data quality.
3Quantity of substance
If public data sets are used, then data availability is improved, but adaptability to specific practical scenarios deteriorates
Solution Approach 1:
The patent extracts seed images containing specific local features from original images and uses these as basis for generating composite images. This local quality approach ensures that the generated data set contains features specific to the practical scenario while maintaining the benefits of automated generation, thereby improving scenario adaptability.
Solution Approach 2:
The system applies various image enhancement operations that change parameters such as brightness, contrast, and composition to generate diverse composite images from seed images. This parameter transformation allows the data set to adapt to different practical scenarios while being generated automatically from limited original materials.
4Productivity
If automated data enhancement is implemented, then data expansion efficiency is improved, but system complexity increases
Solution Approach 1:
The patent divides the data enhancement process into distinct modules: seed image extraction, image enhancement operation, and composite image generation. This segmentation allows each module to be optimized independently while working together to achieve high data expansion efficiency, managing system complexity through modular design.
Data Source
AI summary
A data enhancement expansion method, recognition method and system for deep learning, the data enhancement expansion method includes the following steps: collecting original video data of a target to be recognized, and extracting original images of several recognized targets from the original video data; extracting seed images of RoI outlines of the recognized targets from the original images; performing an image enhancement operation on the seed images of the RoI outlines of the recognized targets, and randomly extracting the seed images subjected the image enhancement operation for image aliasing enhancement to obtain several composite images; and generating a data set based on the original images of the several recognized targets and the several composite images. Original data materials are easy to obtain with extremely low cost and high authenticity, and can be really put to a deep learning network to achieve good recognition results.