Sports Neural Network Codec for Real-Time Player Segmentation
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Conventional sports vision analysis systems are computationally redundant and time-consuming due to vertically structured modules that independently encode and decode visual information, and they lack the ability to support real-time object detection and foreground identification.
Innovation Solution
A unified machine learning architecture is employed, combining an object detection module with a mask subnet to perform both object detection and foreground identification in a single framework, allowing for real-time processing of sports video streams.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Reliability
If conventional vertically structured modules are used for sports vision analysis, then each module can independently perform its function, but the system becomes computationally redundant and time-consuming
Solution Approach 1:
The patent combines object detection and foreground identification into a single unified neural network model. The object detection module and mask subnet are integrated to process video frames simultaneously, eliminating the need for separate independent modules and reducing computational redundancy while maintaining processing reliability
2Reliability
If conventional vertically structured modules are used for sports vision analysis, then each module can independently perform its function, but the system becomes computationally redundant
Solution Approach 1:
The patent merges the object detection module and mask subnet into a single integrated neural network architecture. This consolidation allows both object detection and foreground identification to be performed simultaneously on the same video frames, eliminating duplicate computational operations and reducing energy consumption while maintaining independent functional capabilities
3Device complexity
If conventional systems are used, then processing can be done in stages, but real-time object detection and foreground identification cannot be supported
Solution Approach 1:
The patent integrates object detection and foreground identification into a single unified processing pipeline within the neural network. The object detection module and mask subnet operate simultaneously on input frames, producing both detection results and foreground masks in real-time, eliminating the delays associated with sequential stage-based processing
Data Source
AI summary
A computing system receives a broadcast video stream of a game. A codec module of the computing system extracts image level features from the broadcast video stream. The codec module includes an object detection portion configured to detect players in the broadcast video stream and a subnet portion attached to the object detection portion. The subnet portion is configured to identify foreground information of the detected players. The codec module provides the image level features to a plurality of task specific modules for analysis. The plurality of task specific modules generates a plurality of outputs based on the image level features.