I want to ask a question. How have you labeled or annoted and created the dataset?

[rambo1111]

How have you created the JSON file for the images?

I want make the same project but for badminton and table tennis.

[yastrebksv]

For each image I used this script https://github.com/ArtLabss/tennis-tracking/blob/main/court_detector.py with some modification to get court annotation. But beause there were a lot of mistakes, I filtered resutls manually.

[fm64hylian]

For each image I used this script https://github.com/ArtLabss/tennis-tracking/blob/main/court_detector.py with some modification to get court annotation. But beause there were a lot of mistakes, I filtered resutls manually.

I'm sorry, I am trying to train the dataset as well and you mentioned that you modified court_detector.py to get the annotation, however, there is one "metric" key in data_train.json that I don't know where is coming from. Could you please explain what's used for and how is it calculated?

[yastrebksv]

"Metric" is something like intersection over union between white lines of court in original image and reference one