Prompt-Guided Object Detection With Mask Clipping for Reduced Annotation Cost
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing object detection methods using machine learning models require significant cost and time to add ground truths to training data, making them inefficient.
Innovation Solution
An object detection method that includes an object mask generation, image clipping, and object identification process, utilizing prompts to generate masks and clipped images for accurate object detection without extensive training.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If existing object detection methods using machine learning models are used, then object detection can be performed, but it requires significant cost and time to add ground truths to training data
Solution Approach 1:
The patent uses a segmentation model to generate pseudo ground truths by copying and adapting existing segmentation capabilities to create training data automatically, eliminating the need for manual ground truth annotation while maintaining detection accuracy
Solution Approach 2:
The system performs self-service by automatically generating its own training data through the segmentation model, which creates masks and pseudo ground truths without external human intervention, thus resolving the time-consuming ground truth addition problem
2Measurement precision
If existing object detection methods using machine learning models are used, then object detection can be performed, but it requires significant cost to add ground truths to training data
Solution Approach 1:
The segmentation model copies its masking capability to generate pseudo ground truths automatically, replacing expensive manual annotation processes with an automated system that produces training data at minimal cost
Solution Approach 2:
The system generates its own training data through self-service mechanisms, where the segmentation model automatically creates pseudo ground truths without requiring external human labor, thereby eliminating the cost burden of ground truth addition
Data Source
AI summary
In an object detection device, the object mask generation means outputs a mask indicating an area of an object included in an image and a detection score indicating detection accuracy of the area, based on the image and a prompt. The image clipping means outputs a clipped image obtained by clipping the area of the object from the image, based on the mask and the detection score. The object identification means identifies a category of the object based on the clipped image. The output means outputs the area of the object and the category of the object, as an object detection result.