Why Use T2M for Evaluating FID, Diversity, and R-Precision Metrics?

[kangshwan]

Dear Authors,

First of all, thank you for your great work on diffusion-based motion generation models.

I have some questions regarding the evaluation metrics used in your paper, specifically about using the T2M dataset to evaluate FID, Diversity, and R-Precision:

Use of T2M for Evaluation Metrics: I noticed that you used the T2M (Text-to-Motion) dataset to calculate FID, Diversity, and R-Precision. Is this because T2M already provides matched text and motion pairs, making it suitable for calculating R-Precision?

Application of FID and Diversity: If that's the case, I'm still curious about the use of T2M for FID and Diversity metrics. Could you please explain why you chose to use T2M instead of applying FID and Diversity directly to your dataset or generated motions?

Thank you for your time, and I appreciate any insights you can provide on these questions.