Distinguishing Teams in Tracking Players on Field

I’m using this blog Track Football Players with Computer Vision to help me with my project to track American football players on the field. However, I would also like to separate the two teams, and is it possible for me to find the offensive and defensive players on the field? I can make it so that all the offensive players start out on the right hand side on the screen, but I would just like to identify and distinguish between teams.

Hey @Mustela_Erminea

Yes, its possible, but it’ll require a bit of work on your part. This isn’t a exhaustive list of solutions, but here are some thoughts that could help guide you:

  • You could use the dataset linked in the post to create your own dataset with two (or as many as you want) classes for each of the teams. (I found this Universe dataset which distinguishes the referee, in a similar concept. I recommend searching through Universe)
  • You could create a classification model then set up a two-stage model that classifies each player into a team
  • You could use some code (such as from this Stack Overflow post) to find the dominant color from within the image bounding box, then use that to identify the teams.

Hope this helps

This topic was automatically closed 21 days after the last reply. New replies are no longer allowed.