Hey i have a question so i wanted to use python + machine leaning to cut a football highlight video into different sub-categories like shots,goals,fouls
It should be a pretty interesting task, but you didn’t write the question.
My question was could you give me pointers to make the software I was thinking that I can use computer vision to give each player a skeleton and track their movements
I think the best starting point would be a course on computer vision, and the first library that comes to mind when we talk about computer vision is
I did a quick search and found this one, I haven’t tried it but it looks promising:
To provide more detail on my project i made a visual analysis to explain how each player in each video will given artificial skeleton to tell the machine learning model if the player is making a shot
Hi! There are two common approaches for performing such a task.
The first one is to use pipeline training (you are doing it now), i.e. split your task to few subtasks and train models for each subtask: detect the ball, players, estimate their poses and using all these data classify (by another one model) if the player is making a shot or just running. This is the more traditional approach.
The second one is end-to-end learning: you just feed your model with data and resulting data (i.e. is this video frame contains a shot or not), and if your dataset and model’s hypothesis space are large enough, you’ll be able to solve your task using only one model.
As far as I know, tasks related to classifying sports activities are already able to be solved by the end-to-end approach. As a starting point, I recommend you to investigate these articles: (see the post below)
Thanks if you have more information please contact me
Hello Guys i want to announce that this project is still up and going and i really need more of your ideas to assist me, so far i have gotten two amazing input and i will be glad to receive more input to use the best of my knowledge alongside yours