Hierarchical Object Detection for Video Processing
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing AI algorithms for object detection in images require excessive computational time and power when dealing with a large variety of objects or numerous objects, making it inefficient to detect and highlight all objects, especially when users are interested in only a few, and it becomes time-consuming to find the objects of interest amidst numerous highlighted objects.
Innovation Solution
A hierarchical approach to object detection and selection, where a user can command the system to progressively zoom in on objects of interest by detecting and highlighting objects at multiple levels, allowing the user to focus on specific details without unnecessary processing, with each level revealing more granular details of the previously selected object.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Reliability
If AI algorithms detect and highlight all objects in an image, then object detection completeness is improved, but computational time and power consumption increase excessively
Solution Approach 1:
The patent divides object detection into multiple hierarchical levels. At each level, the system detects objects within a specific region of interest rather than the entire image. This segmentation approach allows comprehensive detection to be achieved in a structured manner, maintaining detection completeness while reducing the computational scope at each step.
Solution Approach 2:
The system performs detection on only the necessary portions of the image at each hierarchical level. By focusing computational resources on specific regions containing objects of interest rather than processing the entire image uniformly, the system achieves effective detection with reduced computational time and power consumption.
2Reliability
If AI algorithms detect and highlight all objects in an image, then object detection completeness is improved, but power consumption increases excessively
Solution Approach 1:
The hierarchical detection framework segments the image processing task into multiple levels, where each level processes only relevant regions. This reduces the total computational workload and corresponding power consumption while maintaining comprehensive detection coverage through progressive refinement of detection regions.
Solution Approach 2:
The system performs partial detection actions at each hierarchical level, processing only the necessary image regions rather than the complete image. This selective processing approach significantly reduces power consumption while achieving complete object detection through the cumulative effect of multiple hierarchical levels.
3Reliability
If a large number of objects are highlighted with frames or boxes, then object detection coverage is improved, but user ability to find objects of interest deteriorates
Solution Approach 1:
The patent segments the detection results into hierarchical levels, with each level displaying objects within a specific region of interest. This segmentation reduces the number of highlighted objects visible to the user at any given time, improving ease of operation while maintaining comprehensive detection coverage across all levels.
Solution Approach 2:
The system adds a hierarchical dimension to object presentation, organizing detected objects across multiple levels rather than displaying all objects simultaneously in a single view. This dimensional organization allows users to navigate through hierarchical levels to find objects of interest, improving user experience while maintaining comprehensive detection coverage.
Data Source
AI summary
A processor of an apparatus plays a video on a display device and receives a command one or more times. In response to receiving the command, the processor performs object detection in a hierarchical manner with respect to objects in the video. The processor displays a video image from the video. Upon receiving the command for a first time, the processor detects in the video image a first set of one or more objects at a first hierarchical level and highlights the first set of one or more objects. Upon receiving the command for a second time, the processor detects in the video image a second set of one or more objects at a second hierarchical level below the first hierarchical level and highlights the second set of one or more objects.