ICACT20210175 Question.1
Questioner: baniyab@gram.edu 2021-02-09 오후 6:54:29 |
ICACT20210175 Answer.1
Answer by Auhor quandh@vnu.edu.vn 2021-02-09 오후 6:54:29
Chrome  Click!! |
| Have you tried to calculate the training time (of each classifier for real time facial recognition)? Among them which classifier takes less training time as compared with other classifiers (or computational time? |
Thank you for your question. Comparison of training time is not a goal in our survey. Deep learning methods normally take more time and computing resources compared to traditional machine learning-based methods. However, as I know, a CNN based approach can recently be process well in real time for both Face detection and facial expression recognition tasks. Its main advantage is the ability to detect faces/recognite expressions in different positions (including side views) and different lighting scenarios. |
ICACT20210175 Question.2
Questioner: iglee@sungshin.ac.kr 2021-02-09 오후 4:50:46 |
ICACT20210175 Answer.2
Answer by Auhor quandh@vnu.edu.vn 2021-02-09 오후 4:50:46
Chrome Click!! |
| Thanks for sharing an interesting study. When using deep learning algorithms in the image processing field, a lot of time and computing resources are consumed. In what environment did you experiment in the paper? And how long did it take? |
It is my pleasure to answer your question as follows. Our work is a survey, and we only considered a performance measurement, accuracy, for each referred experiment presented.
It is true that a deep learning algorithm normally takes more time and computing resources to process, however, the advances in hardware mostly GPUs is accelerating deep learning models, especially those based on CNN architecture and its variations. It is worth noting two interesting papers: (1) Shaohuai Shi el al., 2016: Benchmarking State-of-the-Art Deep Learning Software Tools; (2) Eugenio Gianniti et al., 2019: Performance Prediction of GPU-based Deep Learning Applications; in which they demonstrated deep learning models in different types of GPUs. |
ICACT20210175 Question.3
Questioner: baniyab@gram.edu 2021-02-09 오후 5:14:16 |
ICACT20210175 Answer.3
Answer by Auhor quandh@vnu.edu.vn 2021-02-09 오후 5:14:16
Chrome Click!! |
| Have you tried the implement the feature selection technique for facial expression classification? |
Thanks for your question. A feature selection technique can be applied after the feature extraction phase, probably in order to reduce irrelevant or redundant features while it also decreases the training time of classifier. However, in our survey, we did not study algorithms for feature selection. I highly recommend this paper: Ghosh, M. et al., 2019: Feature selection for facial emotion recognition using late hill-climbing based memetic algorithm. The author introduced a good relevant review of facial feature selection techniques. |
ICACT20210175 Question.4
Questioner: alihussain.dream@gmail.com 2021-02-09 오후 7:01:48 |
ICACT20210175 Answer.4
Answer by Auhor quandh@vnu.edu.vn 2021-02-09 오후 7:01:48
Chrome Click!! |
| I want to know, Your proposed method how long did it take to train the data. |
It is my pleasure to answer your question. Comparison of training time is not a goal in our survey, and we only considered a performance measurement, accuracy, for each referred experiment presented. |
ICACT20210175 Question.5
Questioner: alihussain.dream@gmail.com 2021-02-09 오후 7:00:36 |
ICACT20210175 Answer.5
Answer by Auhor quandh@vnu.edu.vn 2021-02-09 오후 7:00:36
Chrome Click!! |
| Do you think your proposed method is efficient as compared to previous ones? |
Thank you for the question. Our work is a survey. We didn't propose a new method for FER problem. Our main contributions are as follows. Different from other surveys, in this work, we conduct more specific and detailed research from 2014 to 2020 with an aim to not only comprehensively highlight their differences and similarities between the two types, but also the new trend of hybrid and ensemble learning in FER systems. We then extend our survey to current application scenarios in Vietnam. |
ICACT20210175 Question.6
Questioner: kobiljonikromjanov@gmail.com 2021-02-09 오후 5:19:31 |
ICACT20210175 Answer.6
Answer by Auhor quandh@vnu.edu.vn 2021-02-09 오후 5:19:31
Chrome Click!! |
| Thank you very much for submitting your research article and welcome to IEEE/ ICACT2021. Can you please explain your contribution in this article? |
It is my pleasure to explain our contribution in the work:
(1) We comprehensively highlight the differences and similarities of the three FER approaches: Traditional machine learning – based approaches; Deep learning – based approaches; Hybrid and Ensemble learning. It is worth noting that several surveys on automatic expression analysis have been published in recent years, but those study focus only either on traditional machine learning-based methods, or recently deep learning-based methods.
(2) Detail competitive performances and experimental comparisons of researches from 2014 to 2020 on widely used datasets.
(3) We discuss our current FER application scenarios in Vietnam. |
ICACT20210175 Question.7
Questioner: shubhamkj2405@gmail.com 2021-02-09 오후 10:33:36 |
ICACT20210175 Answer.7
Answer by Auhor quandh@vnu.edu.vn 2021-02-09 오후 10:33:36
Chrome Click!! |
| Dear Author, thanks for a brilliant presentation. I really enjoyed reading through the slides. I was interested in the findings of the dataset that your FER in Vietnam collected. Do you plan to use that for the next phase of this research and would that be open source at some point in the future? Also, what method do you think would best suit your FER dataset? |
|
ICACT20210175 Question.8
Questioner: 17862328070@163.com 2021-02-10 오후 5:21:47 |
ICACT20210175 Answer.8
Answer by Auhor quandh@vnu.edu.vn 2021-02-10 오후 5:21:47
Chrome Click!! |
| Thank you for sharing an interesting study. Can you introduce the research on how to protect privacy in face recognition applications |
|
ICACT20210175 Question.9
Questioner: fatou.mbodji@esp.sn 2021-02-11 오전 1:24:51 |
ICACT20210175 Answer.9
Answer by Auhor quandh@vnu.edu.vn 2021-02-11 오전 1:24:51
Chrome Click!! |
| Dear authors, sorry for asking again. I have asked this question but it seems that the system deleted it.
Thank you for your slides. My question is about the analysis of différents subljects:(male/female , your/old, ethnicity) mentioned in your slide 37. Since these subjects have visible characteristics. Couldn't we include the characteristics of these different subjects in the FER in order to detect the class of the subject and to make a FER conforming to the detected class? |
|
