Discussion

{% include list.liquid all=true %}

Paper Goal: Propose MC-ViViT to detect MCI from NC via the facial videos of the interviews (I-CONECT dataset).

Components' Performance: The experiments show that MC-ViViT can predict well with the help of MC and HP Loss.

Good news: Transformer-based model can tell whether or not the people have MCI merely by analyzing the casual chatting videos.

Data's Plus Point: The interview videos from I-CONECT Study is easy to access, low cost, time-saving, flexible, and less restricted than medical images (MRI and CT Scan), collected demographic information, the answers for the specific questions.

It is very valuable to spread and propagate.

Thank you for reading.