Sensor Data Segmentation Using Image-Based Training Labels
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Manual segmentation of sensor data from non-image capture devices is time-consuming and costly, making it unsuitable for many applications, especially in autonomous vehicle operations where efficient data processing is crucial.
Innovation Solution
A system that captures image and sensor data concurrently, uses image data to aid in segmenting sensor data by projecting it onto segmented images to create a training dataset, which is then used to train a sensor data segmentation model for automatic segmentation, enabling the use of sensor data in applications like autonomous vehicle navigation.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If manual segmentation is used for sensor data, then segmentation accuracy is improved, but time consumption and cost increase significantly
Solution Approach 1:
The system performs preliminary segmentation on image data captured by image capture devices, then uses these pre-segmented images as training data to train a sensor data segmentation model. This preliminary action on easier-to-segment image data enables automatic segmentation of more difficult sensor data without manual intervention, resolving the contradiction between accuracy and time consumption.
Solution Approach 2:
The patent introduces segmented image data as an intermediary between manual segmentation and automatic sensor data segmentation. The segmented images serve as training data that bridges the gap, enabling the development of automatic segmentation models for sensor data while maintaining high accuracy without requiring manual segmentation of the sensor data itself.
2Manufacturing precision
If manual segmentation is used for sensor data, then segmentation quality is improved, but processing cost increases prohibitively
Solution Approach 1:
The system creates copies of sensor data and projects them onto pre-segmented image data. These projected copies serve as training examples that enable the sensor data segmentation model to learn segmentation patterns without requiring manual segmentation of actual sensor data, thereby maintaining high segmentation quality while reducing processing costs.
Solution Approach 2:
Segmented image data acts as an intermediary that enables automatic sensor data segmentation. By using these segmented images as training data, the system achieves high segmentation quality for sensor data without the prohibitive costs of manual segmentation, as the intermediary image data can be segmented more efficiently.
3Productivity
If automatic image segmentation is used, then processing efficiency is improved, but applicability to other sensor types deteriorates
Solution Approach 1:
The patent develops a universal sensor data segmentation model that can process multiple types of sensor data (LIDAR, radar, ultrasonic, etc.) using the same approach. The model is trained on segmented image data that is then projected onto various sensor data types, enabling one segmentation system to serve multiple sensor types efficiently, thus achieving both processing efficiency and sensor type applicability.
Solution Approach 2:
Segmented image data serves as a universal intermediary that can be projected onto different sensor data types. This intermediary approach allows the same segmentation model to be applied across multiple sensor types, enhancing versatility while maintaining the processing efficiency of automatic segmentation.
4Measurement precision
If manual segmentation is used for sensor data, then segmentation accuracy is improved, but automation level decreases
Solution Approach 1:
The system enables sensor data segmentation to serve itself by using automatically segmented image data as training data. The sensor data segmentation model automatically processes sensor data without manual intervention, achieving both high automation level and segmentation accuracy through self-supervised learning from projected segmented images.
Solution Approach 2:
The system performs preliminary automatic segmentation on image data, then uses these results to train the sensor data segmentation model. This preliminary automatic action eliminates the need for manual segmentation while maintaining high accuracy, thereby increasing the extent of automation without sacrificing segmentation precision.
Data Source
AI summary
A system may include one or more processors configured to receive a plurality of images representing an environment. The images may include image data generated by an image capture device. The processors may also be configured to transmit the image data to an image segmentation network configured to segment the images. The processors may also be configured to receive sensor data associated with the environment including sensor data generated by a sensor of a type different than an image capture device. The processors may be configured to associate the sensor data with segmented images to create a training dataset. The processors may be configured to transmit the training dataset to a machine learning network configured to run a sensor data segmentation model, and train the sensor data segmentation model using the training dataset, such that the sensor data segmentation model is configured to segment sensor data.