Is it allowed to manually label the train data?
I specifically want to create bounding boxes around the training images, to train a object detection model, that will detect the texts.
Is that allowed, given that the training images are available to everyone.
If permitted, do I also need to share that dataset with all the competitors?
Good Day,
Some great questions here; thank you for bringing them up. You are free to supplement the data as you wish, which will be yours; you don’t need to share them with the rest of the contestants. Just note that you won’t be able to manually label the holdout data for the final evaluation. So make sure your solution isn’t dependent on these additional labels.
Team Xeek
1 Like