Player-Team Association Using Gallery Vectors in Sports Video
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing systems for player-team association in sports video feeds face challenges due to poor video encoding, varying lighting conditions, player occlusion, and similar appearances across teams, leading to inaccurate team assignments.
Innovation Solution
A system utilizing convolutional neural networks and classifiers to identify player patches and generate vectors, comparing them with updated reference data (gallery vectors) to determine team affiliations, adapting to lighting changes and incorporating player movement tracking.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If traditional team assignment methods are used, then the system is simpler, but accuracy deteriorates due to poor video encoding, lighting conditions, player occlusion, and similar appearances
Solution Approach 1:
The system segments the player detection and team classification into separate stages: first detecting player patches using a first machine learning model, then generating vectors from these patches using a second model, and finally comparing with gallery vectors. This segmentation allows each component to be optimized independently, improving overall accuracy while maintaining manageable complexity.
Solution Approach 2:
The system performs preliminary actions by pre-processing video frames to identify player patches before the main classification task. The first machine learning model identifies player regions in advance, and gallery vectors are pre-computed from reference images, preparing the data structure needed for accurate real-time team classification despite challenging conditions.
2Productivity
If the system processes all video frames, then completeness is improved, but processing time increases due to continuous analysis requirements
Solution Approach 1:
The system extracts only the necessary information from each video frame - specifically identifying player patches and generating their vector representations - rather than processing every pixel and frame in detail. This extraction approach maintains real-time processing capability while capturing the essential data needed for accurate team classification.
Solution Approach 2:
The system applies partial action by analyzing only the player regions of interest within each frame rather than the entire frame content. By focusing computation on identified player patches and their visual features, the system achieves efficient real-time processing without sacrificing classification accuracy.
Data Source
AI summary
A method for associating a player with a team in a sports event, the method including: receiving a video feed of a sporting event; identifying, based on an output of a first machine learning model, a patch of pixels corresponding to a player in a video frame of the video feed of the sporting event; determining, based on an output of a second machine learning model, a vector of the patch; retrieving gallery vectors for each team in the sporting event; determining a set of distances between the vector and each of the gallery vectors; and determining, based on a closest distance of the set of distances, a team identification for the player.