Object Detection Training via Density-Based Sample Selection
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Conventional object detection systems face challenges with manual labeling of training samples being time-consuming and error-prone, leading to false positive and false negative detections.
Innovation Solution
Utilizing density-based clustering functions to autonomously select positive and negative training samples by analyzing object locations over time, generating accurate probability density functions, and refining object detection systems through autonomous image capture and selection.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If manual labeling of training samples is used, then training data can be obtained, but the process is time-consuming and error-prone
Solution Approach 1:
The system performs self-labeling by automatically generating training samples through density-based clustering of detection results. The object detection system clusters detected objects over time to identify false positives and generate both positive and negative training samples autonomously, eliminating the need for manual human labeling while maintaining high accuracy.
2Measurement precision
If manual labeling is used to generate training samples, then detection accuracy can be improved, but the process becomes error-prone
Solution Approach 1:
The system uses feedback from detection results to continuously improve training data quality. By clustering detection results over time and using the clustered data to generate new training samples, the system creates a closed-loop feedback mechanism that automatically corrects errors and improves reliability without human intervention.
3Productivity
If conventional training methods are used, then object detection can be performed, but false positives and false negatives occur
Solution Approach 1:
The system performs preliminary clustering of detection results to identify false positives before generating training samples. By pre-processing detection data through density-based clustering and using the results to create balanced positive and negative training samples, the system prevents false detections before they occur during actual operation.
Data Source
AI summary
A method for training an object detection system includes estimating a location of a first object in an environment based on a density cluster map generated from a plurality of images of the environment. The method also includes generating one or more negative training samples of the first object in the environment based on the plurality of images, each of the one or more negative training samples corresponding to a second object at a location in the environment that is different than the estimated location of the first object. The method further includes generating positive training samples from a set of images of the first object. The method also includes training the object detection system to detect the first object based on the positive training samples and the negative training sample.