Multimodal Work Recognition Model Filtering for Real-Time Accuracy
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing work recognition systems face challenges in performing advanced work recognition at high speed while maintaining recognition accuracy, particularly when multiple models are required, leading to increased processing time.
Innovation Solution
A work recognition system that utilizes a plurality of work component recognition models, incorporating information such as movement, touch, hearing, position, camera, and equipment tools, and includes a product selection filtering unit, a progress filtering unit, and a work recognition unit to select and apply the appropriate models for real-time work recognition.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If multiple work component recognition models are used for advanced work recognition, then recognition accuracy is improved, but processing time increases
Solution Approach 1:
The system segments the work recognition process into distinct phases: product selection filtering, progress filtering, and work recognition. Each phase uses appropriate models selectively, avoiding the need to run all models simultaneously, thus maintaining accuracy while reducing processing time.
Solution Approach 2:
The system performs preliminary filtering actions before the main work recognition process. The product selection filtering unit and progress filtering unit pre-process the data by eliminating irrelevant models based on production instructions and work progress, so that only necessary models are executed, reducing overall processing time while maintaining recognition accuracy.
2Adaptability or versatility
If multiple work component recognition models are applied simultaneously, then comprehensive work recognition is achieved, but system complexity increases
Solution Approach 1:
The system dynamically adjusts which recognition models are active based on real-time conditions such as production instructions and work progress. This dynamic selection mechanism allows comprehensive work recognition when needed while simplifying the system by deactivating unnecessary models, thus managing complexity adaptively.
Solution Approach 2:
The system extracts and removes unnecessary recognition models from the processing pipeline based on filtering criteria. By taking out only the relevant models needed for the current work context, the system maintains comprehensive recognition capability while reducing the complexity of simultaneously managing all possible models.
Data Source
AI summary
A work recognition system that uses a plurality of work component recognition models into which information of movement, touch, hearing, position, camera, and equipment, and tools at the time of a work performed by a worker is input, the work recognition system including: a product selection filtering unit configured to select the work component recognition model based on a production instruction; a progress filtering unit configured to select the work component recognition model in accordance with progress in the work; and a work recognition unit configured to recognize the work from the selected work component recognition model is provided.