Data Augmentation Object Analysis Model Learning Apparatus
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
The challenge lies in efficiently generating a model for classifying objects in images, particularly in online shopping platforms where there are millions of products, requiring automated keyword extraction from product images, which is hindered by the time-consuming process of labeling data and the need for high-quality learning data.
Innovation Solution
A data augmentation-based object analysis model learning apparatus and method that generates and labels augmented learning data by transforming pixel information in images, using bounding boxes and noise addition to enhance data quality and quantity, thereby improving model performance and reducing labeling time.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If the amount of learning data is increased to improve model performance, then the model accuracy is improved, but the time required for labeling data increases significantly
Solution Approach 1:
The patent applies data augmentation techniques to create copies of existing learning data by transforming images through various operations (rotation, flipping, color adjustments, noise addition). These synthetic copies expand the dataset without requiring manual relabeling, as the labels are inherited from the original images. This resolves the contradiction by increasing data quantity while maintaining labeling efficiency.
Solution Approach 2:
The system uses a pre-trained model to automatically label augmented data by inferring labels from the original labeled images. The augmentation process itself serves the dual purpose of expanding data while preserving labels through inheritance. This self-service approach eliminates the need for manual relabeling of augmented data, resolving the time efficiency problem.
2Quantity of substance
If data augmentation is applied to increase learning data quantity, then the model performance is improved, but the complexity of the data processing system increases
Solution Approach 1:
The patent segments the data processing pipeline into distinct modules: image augmentation module, labeling module, and model training module. Each module performs a specific function independently, making the complex system more manageable and easier to implement. The segmentation allows for modular complexity management while achieving the goal of increasing learning data quantity.
3Productivity
If automated labeling is implemented to reduce labeling time, then the productivity is improved, but the quality of labeled data may deteriorate
Solution Approach 1:
The patent implements a feedback mechanism where the automatically labeled augmented data is fed back into the training process. The model continuously refines its labeling based on the augmented data distribution. This feedback loop ensures that automated labeling quality improves over time while maintaining high productivity, resolving the quality-speed trade-off.
Data Source
AI summary
Disclosed is a data augmentation-based object analysis model learning apparatus including one or more processors, wherein the operation performed by the processor includes acquiring a first space image including a first object image and generating a second space image by changing pixel information included in the first space image, specifying a bounding box in a region including the first object image in the first space image and labeling a first class specifying the first object image in the bounding box, and primarily learning a weight of a model designed based on a predetermined object detection algorithm, for deriving a correlation between the first object image in the bounding box and the first class, by inputting the first space image to the model, specifying an object image included in a space image based on the correlation.