Skip to content

Software to analyze images from a camera to watch over sports matches and detect when common injuries occur (such as concusions and ankle sprains).

Notifications You must be signed in to change notification settings

collinmcfadden/OmniView

Folders and files

NameName
Last commit message
Last commit date

Latest commit

 

History

19 Commits
 
 
 
 
 
 
 
 
 
 

Repository files navigation

Our Goal

High school athletics is powered by a passion to compete through all physical pain and mental barriers, but this dedication can lead to chronic health problems and worsened injuries. With the excitement of the game unfolding, many serious injuries go unnoticed or are intentionally ignored to keep players in the game, which can increase health costs for families. Our software analyzes images from sports events in real-time and notifies trainers and coaches when players appear to have experienced injuries to help increase accountability and awareness of these conditions.

Collecting images

Images of injuries of interest were searched in Bing.com and the links to the queried images were scraped from the page using the Google Chrome plugin called "Link Klipper." The relevant URL's were placed in a text file, and were processed by parseBing.cpp to translate the Bing link to the original image link.

Training the model

We populated a machine learning model using the Clarifai API with images from sports and paired these with the condition of the players in the photos. Conditions, such as "concussion", "broken ankle", and our healthy control, were trained with 100-300 images each to allow the model to predict the probability of the injury occurrence.

Injury Notifications

Upon a detected collision that may require medical attention, notifications will be sent to the required personel. Our AI workstation will send a request with the collision detail to a Lambda function through API Gateway. The Lambda function will trigger notifications for our Android application and website using OneSignal.

Future Directions

Our current model is trained by single images, but it would provide more accurate analysis if it compared actions in sequence and was trained on videos of injury occurrences. Use of Amazon Web Services Rekognition API would allow for training on video clips to compare how a succession of clips can be used to understand the impact of a collision on the players.

About

Software to analyze images from a camera to watch over sports matches and detect when common injuries occur (such as concusions and ankle sprains).

Resources

Stars

Watchers

Forks

Releases

No releases published

Packages

No packages published

Contributors 3

  •  
  •  
  •