Physical Marker Labeling for Automated Image Training Data
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing methods for generating labeled training data for machine learning, particularly for classifying local features of objects in images, are time-consuming, costly, and inefficient, especially for objects with reflective or transparent surfaces and varying lighting conditions, requiring large datasets that are difficult to obtain.
Innovation Solution
A method using physical marker devices adjacent to the local features, combined with camera imaging, to automatically generate training data by detecting and computing regions of interest, thereby reducing human intervention and enhancing data generation speed and quality.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If manual labeling by human experts is used to generate training data, then labeling accuracy can be maintained, but the process becomes time-consuming and costly
Solution Approach 1:
Physical marker devices are introduced as intermediaries between the object and the camera system. These markers serve as detectable reference points that enable automatic identification and labeling of regions of interest, replacing the need for manual expert labeling while maintaining accuracy through precise marker detection and coordinate transformation
Solution Approach 2:
The manual mechanical process of drawing boxes around regions of interest is replaced by an automated computer vision system. The system uses camera imaging to capture marker positions, computes corresponding regions of interest through coordinate transformations, and automatically generates labels without human intervention
2Reliability
If large datasets are collected to improve model reliability for challenging visual conditions, then classification reliability improves, but the cost and time of data collection increases significantly
Solution Approach 1:
The system enables self-service data generation where the physical markers on objects serve as their own reference points. The camera system automatically detects these markers and generates corresponding labels without requiring external manual intervention, allowing rapid collection of large datasets for training robust models
Solution Approach 2:
The system changes the parameter of object representation by introducing detectable physical markers with specific visual characteristics. This transformation enables reliable detection and labeling even under varying lighting conditions and for challenging surfaces, as the markers provide consistent reference points that can be reliably detected and used to generate training data
3Productivity
If simulation and rendering tools are used to generate training data, then large amounts of labeled data can be produced efficiently, but the complexity of the system and cost of software tools increases
Solution Approach 1:
Instead of creating complex virtual simulations, the system uses physical markers as simple, tangible copies or proxies that can be directly detected by camera systems. This approach captures real-world lighting and appearance conditions without requiring elaborate 3D modeling and rendering pipelines, reducing system complexity while maintaining data generation efficiency
Data Source
Figure 1
Figure 2
Figure 3~5
AI summary
Methods and the systems for generating image data for generating training information, and for generating training information for an automated image analysis related to a local feature in an image. A method comprises applying at least one physical marker device adjacent to the local feature of an object. The method then acquires with at least one camera sensor, a plurality of images of the object and stores the plurality of images. A processor detects the at least one physical marker device in at least one image of the acquired plurality of images. The processor computes, for each detected at least one physical marker device, a region of interest in the at least one image based on predetermined relative location information associated with the at least one physical marker device. The processor then generates mask information based on the computed region of interest and stores the generated mask information associated with the at least one image as training information; and generates the classification information for detecting, segmenting, classifying, identifying, or determining a regression for the local feature by training a model using the stored training information.