Automated ML Training Data Collection via Multi-Sensor Event Detection
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Developing generalized machine learning (ML) models for novel applications is hindered by the difficulty in obtaining large and varied datasets, as existing datasets often fail to capture the diverse situations encountered in real-world scenarios.
Innovation Solution
A training data collection device that uses environmental sensors to automatically detect and label events, generating training data by combining data from different sensor types, allowing for the creation of diverse and extensive datasets without manual curation.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Quantity of substance
If manual curation efforts are used to create large labeled training datasets, then dataset size and quality are improved, but time consumption and labor requirements increase significantly
Solution Approach 1:
The system enables automated self-service data collection by having the computing device perform its own environmental sensing, event detection, and training data generation without external manual intervention. The device uses its environmental sensors to collect data, applies configured event detectors to identify events, and automatically generates labeled training data, eliminating the need for manual curation while maintaining large dataset sizes
Solution Approach 2:
The patent replaces the mechanical manual labeling process with an automated electronic system. Instead of human operators manually annotating data, the system uses environmental sensors to capture data, event detectors to identify events automatically, and software to generate labeled training data, substituting physical human labor with electronic automation
2Loss of time
If existing publicly available datasets are used for training, then time and effort are reduced, but the datasets fail to capture varied real-world situations for novel applications
Solution Approach 1:
The system performs preliminary automated data collection and labeling actions before ML model training is needed. By pre-configuring event detectors with training event detection data and having the device automatically collect environmental data and generate labeled training data in advance, the system prepares customized diverse datasets specific to novel applications without requiring time-consuming manual efforts later
Solution Approach 2:
The system changes the parameters of data collection by using multiple environmental sensors to capture different types of environmental data (audio, visual, contextual information). This multi-parameter approach enables the collection of diverse real-world situations that single-source datasets cannot provide, while automation maintains efficiency
3Adaptability or versatility
If multiple environmental sensors are used to collect diverse data, then dataset variety and ML model generalization are improved, but device complexity increases
Solution Approach 1:
The computing device is designed with multi-functionality, serving both as an environmental sensing platform and an ML model training data generator. The device uses its environmental sensors not only for their primary sensing functions but also for collecting training data, and the same processing unit that runs applications also configures event detectors and generates training data, reducing the need for separate dedicated hardware
4Productivity
If automated event detection is implemented, then manual labeling effort is reduced, but the complexity of configuring and maintaining event detectors increases
Solution Approach 1:
The event detection system operates as a self-service automated process. The computing device automatically receives training event detection data, configures event detectors using this data, applies the configured detectors to environmental data to identify events, and generates labeled training data without requiring external manual intervention at any stage, maintaining high productivity while managing complexity through automation
Data Source
AI summary
Methods and systems for automatically generating training data for use in machine learning are disclosed. The methods can involve the use of environmental data derived from first and second environmental sensors for a single event. The environmental data types derived from each environmental sensor are different. The event is detected based on first environmental data derived from the first environmental sensor, and a portion of second environmental data derived from the second environmental sensor is selected to generate training data for the detected event. The resulting training data can be employed to train machine learning models.