Facial Landmark-Based Emotion Recognition via Directed Graph Neural Network

Round 1

Reviewer 1 Report

The introduction of the paper need to be further improved.

Literature review is poor. Authors need to explain them with the recent literature.

Did the authors use the "state of the art" images and compare this result with the other methods?

Sad and contempt are classified poorly compared to the other methods, please explain the reason.

Neutral and happy are also gave poor results, please explain.

Please improve the conclusion and discussion section.

Author Response

Please see the attachment.

Author Response File: Author Response.pdf

Reviewer 2 Report

The authors use DGNN to recognize facial emotion with landmarks. The work is interesting and the results are convincing.

My comments on this paper are as the following:

1) Why does the proposed model work? The authors should highlight the motivation of the paper.

2) The recognition network is built on landmarks. If the methods to find landmarks change, what is its influence on emotion recognition results?

3) In Figure 2, what is the position of subfigure (c) in subfigure (a)?

4) Please check Equation (4) and what does a_{t,i,j} denote?

5) If the accuracy of CK+ is given in Abstract, the authors should also give the accuracy of MMI as well as AFEW in Abstract.

6) The authors should report the running time of the proposed network.

7) The authors should analyze the limitations of the proposed network.

Author Response

Please see the attachment.

Author Response File: Author Response.pdf

Round 2

Reviewer 1 Report

The authors have addressed all my comments satisfactorily and the paper can be considered for publication.

Author Response

Thank you very much for good advice. 

Author Response File: Author Response.pdf