mondrasovic
commented
2 years ago Well, for all I know, your task is doable with this architecture. However, you would be trying to eat a soup with a fork. First off, there is no pre-trained SiamMOT model available that would detect scoreboard and let alone digits in it by default. You would have to train or at least fine-tune the architecture to facilitate the detection of such objects. But my question is the following. Why on Earth would you need a multi-object tracker for your task? I assume you have one scoreboard. So, try to detect and track it using single-object tracking approaches. From what you have described I am not even sure whether the camera is static or moving. Well, long story short. I have more questions than answers. Nevertheless, what I can tell you is this: you are at the wrong place... Unless you provide some information I could not deduce from your initial description which would make me rethink my stance.
Hello: I want to know, in one live volleyball game video. There is one scoreboard in the TV screen, its position is fixed, but the score in the scoreboard is changing quite often. I want to know if I can use the repo to detect the scoreboard, I need its scoreboard position in the screen, and further, can I detect the score change in the scoreboard. For example, at the beginning, the score is: 0-0, if one team scores, then the score becomes 1-0. Can I detect the position of scoreboard and detect the score changes in it? If yes, please give a general instruction on how to do this. Thanks,