In the recent version of the paper it says that due to long tail issue of the keystep recognition task, only 278 keysteps were used for baselines by thresholding the classes with number of segments >= 20. Since the dataset does not come with test annotations, it is not clear how we can figure out the keysteps used in the baselines.
Can you please provide the list of keystep (names or ids) used in the paper?
Also, is it possible to share the baseline checkpoints and training/validation code?
Thank you very much for your excellent work.
In the recent version of the paper it says that due to long tail issue of the keystep recognition task, only 278 keysteps were used for baselines by thresholding the classes with number of segments >= 20. Since the dataset does not come with test annotations, it is not clear how we can figure out the keysteps used in the baselines.
Can you please provide the list of keystep (names or ids) used in the paper? Also, is it possible to share the baseline checkpoints and training/validation code?
Thanks, Deep