Classification Parallelization for Resource-Limited Object Detection
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Unmanned systems with limited computation resources face challenges in using third-party machine learning models for object detection, as they require efficient resource management to operate effectively in operational scenarios.
Innovation Solution
An AI gateway system processes images through multiple machine learning models, allowing operators to select specific object types for detection, and dynamically manages resource utilization by activating only necessary models, adjusting processing based on model efficiency, and throttling or stopping less efficient models to conserve resources.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Adaptability or versatility
If multiple machine learning models are deployed for comprehensive object detection, then detection capability and versatility are improved, but resource consumption and device complexity increase
Solution Approach 1:
The system dynamically activates or deactivates machine learning models based on operational requirements and resource availability. The AI gateway receives indications of object types to recognize and selectively enables only the necessary models, allowing the system to adapt its detection capability to match current operational needs while conserving computational resources.
Solution Approach 2:
The system changes operational parameters by adjusting which machine learning models are active based on the detected object types and resource conditions. This parameter change allows the system to optimize between detection versatility and resource consumption by enabling only the subset of models needed for current operational scenarios.
2Adaptability or versatility
If multiple machine learning models are deployed for comprehensive object detection, then detection capability is improved, but device complexity increases
Solution Approach 1:
The AI gateway serves as an intermediary layer between the machine learning models and the unmanned system. It manages the complexity by receiving high-level indications of object types from operators and automatically determining which specific models to activate, thereby shielding the rest of the system from model management complexity while maintaining versatile detection capability.
Solution Approach 2:
The system segments the machine learning model management into distinct functional components: the AI gateway handles model selection and activation, while individual models handle specific object type detection. This segmentation reduces overall system complexity by dividing the complex task of multi-model management into manageable, specialized components.
3Reliability
If all machine learning models are actively processing frames, then detection coverage is improved, but processing efficiency decreases due to resource constraints
Solution Approach 1:
The system applies partial action by activating only the necessary subset of machine learning models required for current operational needs rather than running all models continuously. This approach maintains sufficient detection coverage for the current scenario while significantly improving processing efficiency by reducing the computational load on resource-constrained devices.
Data Source
AI summary
Methods and systems are described herein for hosting and arbitrating algorithms for the generation of structured frames of data from one or more sources of unstructured input frames. A plurality of frames may be received from a recording device and a plurality of object types to be recognized in the plurality of frames may be determined. A determination may be made of multiple machine learning models for recognizing the object types. The frames may be sequentially input into the machine learning models to obtain a plurality of sets of objects from the plurality of machine learning models and object indicators may be received from those machine learning models. A set of composite frames with the plurality of indicators corresponding to the plurality of objects may be generated, and an output stream may be generated including the set of composite frames to be played back in chronological order.