FERV39k: A Large-Scale Multi-Scene Dataset for Facial Expression Recognition in Videos

September 26, 2022 · View on GitHub

Project Page | Videos | Paper

FERV39k: A Large-Scale Multi-Scene Dataset for Facial Expression Recognition in Videos
Yan Wang ¹, Yixuan Sun ¹, Yiwen Huang ², Zhongying Liu ², Shuyong Gao ², Wei Zhang ²
, Weifeng Ge ², Wenqiang Zhang ^{1, 2}

¹ Academy for Engineering & Technology, Fudan University ² School of Computer Science, Fudan University

总体介绍

News!

08/2022: Notice!

Letter to researchers:
    
    Thank you for your attention in the FERV39k dataset.
    We have released the FERV39k dataset. 
  
  With kind regards,
  Group FERV39K

03/2022: The Paper FERV39k: A Large-Scale Multi-Scene Dataset for Facial Expression Recognition in Videos is accepted by CVPR 2022.

Model Zoo

Model	Backbone	Pre-trained	Fine-tuned	WAR/UAR	Trained-model	Val-Results	# of Parameters
ResNet18	ResNet18	-	-	39.33/30.30	FERV39k-train-ResNet18	FERV39k-val-ResNet18	17M
ResNet50	ResNet50	-	-	30.57/22.47	FERV39k-train-ResNet50	FERV39k-val-ResNet50	124M
VGG13	VGG13	-	-	41.02/31.19	FERV39k-train-VGG13	FERV39k-val-VGG13	128M
VGG16	VGG16	-	-	41.66/32.01	FERV39k-train-VGG16	FERV39k-val-VGG16	134M
R18-LSTM	ResNet18	-	-	42.59/30.92	FERV39k-train-R18-LSTM	FERV39k-val-R18-LSTM	132M
R50-LSTM	ResNet50	-	-	40.75/32.12	FERV39k-train-R50-LSTM	FERV39k-val-R50-LSTM	57M
VGG13-LSTM	VGG13	-	-	43.37/32.41	FERV39k-train-VGG13-LSTM	FERV39k-val-VGG13-LSTM	133M
VGG16-LSTM	VGG16	-	-	41.70/30.93	FERV39k-train-VGG16-LSTM	FERV39k-val-VGG16-LSTM	138M
C3D	C3D	-	-	31.69/22.68	FERV39k-train-C3D	FERV39k-val-C3D	78M
I3D	I3D	-	-	38.78/30.17	FERV39k-train-I3D	FERV39k-val-I3D	12M
3D-R18	ResNet18	-	-	37.57/26.67	FERV39k-train-3D-R18	FERV39k-val-3D-R18	33M
Two C3D	C3D	-	-	41.77/30.72	FERV39k-train-Two-C3D	FERV39k-val-Two-C3D	97M
Two I3D	I3D	-	-	41.30/31.01	FERV39k-train-Two-I3D	FERV39k-val-Two-I3D	26M
Two 3D-R18	ResNet18	-	-	42.28/30.55	FERV39k-train-Two-3D-R18	FERV39k-val-Two-3D-R18	67M
Two R18-LSTM	ResNet18	-	-	43.20/31.28	FERV39k-train-Two-R18-LSTM	FERV39k-val-Two-R18-LSTM	27M
Two VGG13-LSTM	VGG13	-	-	44.54/32.79	FERV39k-train-Two-VGG13-LSTM	FERV39k-val-Two-VGG13-LSTM	144M
RS18-LSTM	ResNet18	MS-Celeb-1M	FERV39k	41.XX/31.XX	FERV39k-train-R18-LSTM-MS-Celeb-1M	FERV39k-val-LSTM-MS-Celeb-1M	132M
RS18-LSTM	ResNet18	DFEW	FERV39k	41.XX/29.XX	FERV39k-train-R18-LSTM-DFEW	FERV39k-val-LSTM-DFEW	132M
RS50-LSTM	ResNet50	MS-Celeb-1M	FERV39k	46.XX/34.XX	FERV39k-train-R50-LSTM-MS-Celeb-1M	FERV39k-val-LSTM-MS-Celeb-1M	57M
RS50-LSTM	ResNet50	DFEW	FERV39k	41.XX/29.XX	FERV39k-train-R50-LSTM-DFEW	FERV39k-val-LSTM-DFEW	57M

Terms & Conditions

The dataset is available for non-commercial research purposes only.
All images of the dataset are obtained from the Internet and some papers which are not property of fudanroilab, Fudan University. The fudanroilab is not responsible for the content nor the meaning of these images.
You agree not to reproduce, duplicate, copy, sell, trade, resell or exploit for any commercial purposes, any portion of the images and any portion of derived data.
You agree not to further copy, publish or distribute any portion of the fudanroilab. Except, for internal use at a single site within the same organization it is allowed to make copies of the dataset.
The SOL reserves the right to terminate your access to the fudanroilab at any time.

How to get the Password (Dataset applications can be available)

This dataset is publicly available. It is free for professors and researcher scientists affiliated to a University.

Baidu Drive: (Available after we receive the email of application to download the FERV39k Dataset

(FERV39k) Download Link (链接): xxxx (Expiration date is 30 days)
Extract Code(提取码)： xxxx

Google Drive:

Download Link (FERV39k) : xxxx (Can be available soon due to the upload capacity limit)

Permission to use but not reproduce or distribute our database is granted to all researchers given that the following steps are properly followed: Send an e-mail to Yan Wang (yanwang19@fudan.edu.cn) and Wenqiang Zhang (wqzhang@fudan.edu.cn) before downloading the dataset. You will need a password to access the files of this dataset. Your Email MUST be set from a valid University account and MUST include the following text:

1. Subject: (fudanroilab) Application to download the FERV39k Dataset          
2. Name: <your first and last name>
3. Affiliation: <University where you work>
4. Department: <your department>
5. Position: <your job title>
6. Email: <must be the email at the above mentioned institution>

I have read and agree to the terms and conditions specified in the FERV39K database webpage. 
This database will only be used for research purposes. 
I will not make any part of this database available to a third party. 
I'll not sell any part of this database or make any profit from its use.

Citation

@inproceedings{wang2022ferv39k,
  title={FERV39k: A Large-Scale Multi-Scene Dataset for Facial Expression Recognition in Videos},
  author={Wang, Yan and Sun, Yixuan and Huang, Yiwen and Liu, Zhongying and Gao, Shuyong and Zhang, Wei and Ge, Weifeng and Zhang, Wenqiang},
  booktitle={Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition},
  pages={20922--20931},
  year={2022}
}