Workpiece Pick Training Data Generation Using Real Arrangement Images
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Generating large amounts of training data for machine learning to accurately estimate the pick position and orientation of workpieces is inefficient and often inaccurate due to the difficulty in replicating real-world conditions, especially with large or heavy workpieces, as existing simulation methods produce idealized images that do not account for variations and deformations.
Innovation Solution
A training data generator system that includes a measuring instrument and movement device to capture real-world images and change the arrangement pattern of workpieces, generating training data that reflects actual conditions by acquiring images and pick position information through repeated measurement and movement processes.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If a large number of training data are generated to improve machine learning accuracy, then the accuracy of estimating pick position and orientation is improved, but the time and labor required for data generation increases significantly
Solution Approach 1:
The patent uses a simulation device to create virtual copies of workpieces and their arrangement patterns. The simulation device generates training data by virtually arranging workpiece models in various patterns and generating corresponding images, eliminating the need to physically handle real workpieces for each training data sample. This copying approach allows rapid generation of diverse training data without time-consuming manual intervention.
Solution Approach 2:
The system pre-generates multiple arrangement patterns of workpieces using the simulation device before actual machine learning training. By preliminarily creating a diverse set of virtual arrangement patterns and their corresponding images, the system prepares comprehensive training data in advance, reducing the time needed during actual model training and deployment phases.
2Adaptability or versatility
If manual methods are used to change arrangement patterns of large or heavy workpieces, then diverse training data can be obtained, but the labor burden and efficiency decrease significantly
Solution Approach 1:
The simulation device creates virtual copies of workpieces that can be freely arranged without physical constraints. These digital models can be positioned, rotated, and stacked in any configuration instantly, providing diverse arrangement patterns without the labor burden of physically moving heavy real workpieces. The virtual copies maintain all necessary geometric and visual properties for effective machine learning training.
Solution Approach 2:
The patent replaces the mechanical system of physically handling and arranging real workpieces with a computational simulation system. Instead of manually moving heavy objects to create different arrangements, the system uses software to virtually position workpiece models, eliminating the need for physical labor while maintaining the ability to generate diverse arrangement patterns efficiently.
3Productivity
If simulation devices are used to generate training data, then the time and labor burden are reduced, but the accuracy decreases due to idealized images not reflecting real-world conditions
Solution Approach 1:
The simulation device is configured to adjust various parameters including lighting conditions, camera angles, and workpiece positions to better match real-world imaging scenarios. By changing these parameters to reflect actual operating conditions, the generated training data becomes more realistic while maintaining the efficiency advantages of simulation-based generation.
Data Source
AI summary
This device for generating learning data is provided with a measuring device, such as a visual sensor, that measures an arrangement region of a plurality of workpieces and acquires an image. The device for generating learning data is provided with a learning data generation unit that generates learning data including an image acquired by the visual sensor and a workpiece removal position. The device for generating learning data generates a plurality of sets of learning data by repeating: control of the movement of the workpieces using a movement device so as to change the arrangement pattern of the workpieces; measurement of the arrangement region of the plurality of workpieces by the visual sensor; and generation of learning data by the learning data generation unit.