rvi101 / deepball-keras Goto Github PK

Hi,

thanks for your repository and the great work. I was looking for an example of the "hard negative mining strategy" they employ in the paper but couldnt find one. It is just not clear to me how to mitigate the class imbalance when all you can do is feed the complete image into the network ?

Do you have any suggestions on this ?

Thanks,
Alexander

Hey, I'm studying existing ball detection solutions, came across your repository and gave it a test with a sample I picked from youtube, it is a wide shot, filming the entire pitch with a pretty shallow angle, and the ball is very rarely detected accurately. I'm merely posting as feedback, but I'll be curious if you have an idea what makes this sort of footage a bad fit. The sample is up at mathieudu.com/misc/clay.avi :)

Hi, I'm a student who is interested in ball tracking solutions. I was impressed with your methods and results, so I decided to try out the code on my own. But I got poor results on the ISSIA dataset video (5th one), it rarely detected the ball. Aren't the models pre-trained with the ISSIA dataset? Or is there anything else I'm missing?

Also, if I want to apply the model to a video with a different ball size, would changing the bg_kernel value help? It would be appreciated if you can give me some suggestions.

Thanks in advance!

I am trying to train another model but it's giving me very high loss and very poor precision. Also, I would like to understand what you meant by ''The dictionaries are to account for the fact that some of the sequences are horizontally flipped but the coordinate systems are not, so ball data will be wrong without storing this."