I am just getting into visual machine learning (currently a mobile developer) and have a challenging project of interest.

It involves using video as an input to then determine if a baseball player should have made a catch. I am imagining this should be viewed as a classification type problem using supervised learning. My available training set is +100,000 examples (video input, YES/NO output)

The new data (video) input I am hoping to use is either a panning or fixed-view (non-panning) video that captures a baseball being hit and an outfielder attempting to catch the ball

The desired classification: Should that have been a catch (not concerned with whether it was a catch), YES or NO

My question is:

I have seen algorithms successfully classify objects in video among a few other video machine learning perception tasks but I have been unsuccessful in identifying something similar to the project I am working on. I am hoping to be pointed in a good direction or it be explained the algorithms, etc relevant in attacking this problem.

Any help or thoughts are greatly appreciated

  • 1
    $\begingroup$ Much of the computer vision literature focuses on the analysis of single images, e.g., is there a car in this image or is the face in this image smiling? However, it seems to me that your task will require analysis of a series of images or at the very least the selection of the most appropriate single image from the series. So look for temporal techniques that account for time. $\endgroup$ – Jeffrey Girard Nov 23 '16 at 1:28

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy

Browse other questions tagged or ask your own question.