First, thank you for this excellent work and for making the code available!
I am building upon the ideas from your paper for my thesis research and have found it incredibly inspiring. I'm attempting to run the preprocessing for the Kinetics-Sounds and UCF101 datasets, but I'm facing computational constraints with the full datasets. Would you be able to share your preprocessed versions?
This data is crucial for my academic, non-commercial thesis work, and I will of course cite your paper prominently. Once again, thank you for considering this request and for your contribution to the community.