Evaluation details

[HaozhouPang]

Hello,

Thanks for sharing this amazing work along with the evaluation script. The evaluation section of the paper says

We evaluate our performance in monadic gesture generation on the recently introduced BEAT dataset. The test set includes 2492 5-sec motion sequences and includes a set of 5 unseen speakers.

Could you clarify how the 5-sec clips were splitted? can you provide the script to generate the evaluation data? or even better can you share the dataset that you use in the evaluation?

Any help will be appreciated.