Bounding-Box Auto Segmentation for Lower-Cost Food Image Labeling
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Current food identification technologies using deep learning for image recognition face challenges due to the need for large amounts of manually labeled data, which is time-consuming and costly, and the low accuracy of food type identification, leading to errors in calorie counting and inefficient food management.
Innovation Solution
A method and apparatus for auto segmentation using a bounding box through weakly-supervised learning, where a segmentation model is trained using bounding box labels and color similarity maps to classify objects and backgrounds, with auto-labeling through user inspection for prediction errors.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If manual labeling is used to obtain data for deep learning model, then data quality can be ensured, but time cost and cost increase significantly
Solution Approach 1:
The system uses automatic segmentation models to perform labeling without human intervention. The model processes images, detects objects, and generates bounding boxes and segmentation masks automatically, making the system self-sufficient in data preparation while maintaining high data quality through continuous model improvement
Solution Approach 2:
The system incorporates feedback mechanisms where segmentation results are evaluated and used to refine the model. Prediction confidence scores and error analysis feed back into the system to automatically adjust and improve labeling accuracy over time, reducing the need for manual verification
2Manufacturing precision
If segmentation labeling is performed by drawing polygons, then precise object segmentation is achieved, but the number of clicks and time required increase overwhelmingly
Solution Approach 1:
The patent replaces the mechanical manual process of drawing polygons with an automated computer vision system. The segmentation model uses deep learning algorithms to automatically detect object boundaries and generate precise masks without requiring human interaction, substituting manual mechanical actions with automated computational processes
Solution Approach 2:
The system divides the image processing task into multiple segmentation stages using hierarchical models. The segmentation model first identifies coarse regions, then refines boundaries at finer levels, achieving precise object segmentation through multi-scale analysis rather than requiring manual polygon drawing
3Reliability
If deep learning model is trained with large amount of data, then model performance improves, but data accumulation beyond certain amount becomes less effective than data quality improvement
Solution Approach 1:
The system changes the focus from data quantity to data quality by implementing automated filtering and validation mechanisms. The segmentation model processes and selects only high-quality labeled data, adjusting the effective data parameters to prioritize accuracy and relevance over sheer volume, thereby optimizing model performance
Data Source
AI summary
Provided are a method and an apparatus for auto segmentation using a bounding box. A method for auto segmentation using a bounding box according to one embodiment of the present disclosure comprises receiving a first object image including an object labeled with a bounding box, which is a pre-learning target, learning a segmentation model by classifying an object and a background from the bounding box of the received first object image, and segmenting an object from a second object image, which is an identification target, using the learned segmentation model.