1. Facial Motion Prior Networks for Facial Expression Recognition

## Algorithm:

## Pros:

- SOTA

- Instead of making the network to learn active areas blindly, we choose to guide it via some pseudo ground-truth masks that are generated by modeling basic facial expressions.
-
- Generate a facial mask to focus on facial muscle moving regions by using the average differences between neutral faces and the corresponding expressive faces as the training guidance.

## Cons:

- The training phase mainly contains two stages, training the generator for generating facial motion mask, and jointly training the whole recognition network.

- Not suitable for real time situations

2.