Facial Expression Databases From Other Research Groups

 

We list some face databases widely used for facial expression related studies, and summarize the specifications of these databases as below.

 

1. Binghamton University facial expression databases.  

o       Source: The Binghamton University facial expression databases are built by Dr. Lijun Yin at Binghamton University and other collaborators,

o       Purpose: The Binghamton University facial expression databases record images or videos of subjects with various facial expressions. There are multiple types of subsets. Some subsets contain 4D facial data. Some subsets contain multi-modality facial data.  

o       Properties:

Properties

Descriptions

# of subjects

Number of subjects varies with different data subsets.

# of images/videos

-

Static/Videos

Static and videos.

Single/Multiple faces

Single

Gray/Color

color

Resolution

-

Face pose

-

Facial expression

Various expressions.

Illumination

-

3D data

3D face scann

Ground truth

Facial expression and facial action unit annotations. Some data subsets contain tracked facial landmark locations.

o       Reference: refer to the website: http://www.cs.binghamton.edu/~lijun/Research/3DFE/3DFE_Analysis.html

 

 

2. Acted Facial Expressions in the Wild (AFEW) and Static Facial Expressions in the Wild (SFEW) databases.  

o       Source: The AFEW and SFEW databases are built by Australian National University, University of Canberra, and Commonwealth Scientific and Industrial Research Organisation, Australia ,

o       Purpose: Acted Facial Expressions In The Wild (AFEW) is a dynamic temporal facial expressions data corpus consisting of close to real world environment extracted from movies. Static Facial Expressions in the Wild (SFEW) has been developed by selecting frames from AFEW.  

o       Properties:

Properties

Descriptions

# of subjects

330

# of images/videos

1426 video sequences in AFEW database. 700 images in SFEW database (SPI category).

Static/Videos

Videos in AFEW, Static images in SFEW.

Single/Multiple faces

Multiple

Gray/Color

color

Resolution

-

Face pose

Various poses

Facial expression

Angry, Disgust, Fear, Happy, Neutral, Sad, Surprise.

Illumination

Various illuminations

3D data

coarse head pose label

Ground truth

5 facial landmark annotations for some images

o       Reference: refer to the paper: Abhinav Dhall, Roland Goecke, Simon Lucey, Tom Gedeon, Collecting Large, "Richly Annotated Facial-Expression Databases from Movies", IEEE Multimedia 2012. Abhinav Dhall, Roland Goecke, Simon Lucey, and Tom Gedeon, "Static Facial Expressions in Tough Conditions: Data, Evaluation Protocol And Benchmark", First IEEE International Workshop on Benchmarking Facial Image Analysis Technologies BeFIT, IEEE International Conference on Computer Vision ICCV2011, Barcelona, Spain, 6-13 November 2011.

 

  1. The Second Emotion Recognition In The Wild Challenge and Workshop (EmotiW 2014) dataset

o         Source: this database is provided by the Second Emotion Recognition In The Wild Challenge and Workshop.

o         Purpose: this database is primarily used to evaluate the emotion recognition methods in real-world conditions.

o         Properties:

Properties

Descriptions

# of subjects

N/A

# of images/videos

Training (578 videos), validataion (383 videos), and test sets (N/A)

Static/Videos

Video and audio

Single/Multiple faces

Single

Gray/Color

Color

Resolution

N/A

Face pose

various face poses

Facial expression

7 basic facial expressions: Anger, Disgust, Fear, Happiness, Neutral, Sadness and Surprise.

Description of facial expression

video clips from movies.

Illumination

N/A

Accessories

N/A

3D data

N/A

Frame rate

N/A

Ground truth

Facial expression label for each video

 

  1. Cohn-Kanade AU-Coded Facial Expression Database

o         Source: this database is provided by Jeff Cohn from Carnegie Mellon University.

o         Purpose: this database is widely used as the standard database to evaluate the facial action unit recognition systems. It may also be used for facial expression recognition and face recognition.

o         Properties:

Properties

Descriptions

# of subjects

100 university students (released version)

65% female, 15% African-American, and 3% percent Asian or Latino.

# of images/videos

486

Static/Videos

Videos

Single/Multiple faces

Single

Gray/Color

Eight-bit gray

Resolution

640* 490

Face pose

Frontal-view only

Facial expression

23 facial displays including single AUs or combinations of AUs

Description of facial expression

Neutral to apex;

Posed facial expressions

Illumination

N/A

Accessories

N/A

3D data

N/A

Frame rate

12 frame/sec

Ground truth

AU label for final frame in each image sequence

Identifications of subjects

 

  1. MMI Database

o         Source: this database is provided by M. Pantic and M. F. Valstar.

o         Purpose: this database is primarily used to evaluate the facial action unit recognition systems. It may also be used for face recognition.

o         Properties:

Properties

Descriptions

# of subjects

43

# of images/videos

1280 videos and over 250 images

Static/Videos

Videos and static images

Single/Multiple faces

Single

Gray/Color

Color

Resolution

720* 576

Face pose

Frontal-view or dual-view (frontal and profile) captured by two cameras simultaneously

Facial expression

79 facial displays including single AUs or combinations of AUs

Description of facial expression

Neutral-apex-neutral;

Posed facial expressions

Illumination

N/A

Accessories

N/A

3D data

N/A

Frame rate

24 frame/sec

Ground truth

AU label for the image frame with apex facial expression in each image sequence;

Some image sequences have been FACS coded for each frame;

Lot of metadata of subjects

 

  1. Japanese Female Facial Expression (JAFFE) Database

o         Source: this database was planned and assembled by Miyuki Kamachi, Michael Lyons, and Jiro Gyoba.

o         Purpose: this database is primarily used to evaluate the facial expression recognition systems. It may also be used for face recognition.

o         Properties:

Properties

Descriptions

# of subjects

10

# of images/videos

213

Static/Videos

Static

Single/Multiple faces

Single

Gray/Color

Eight-bit gray

Resolution

256* 256

Face pose

Frontal-view

Facial expression

7 facial expressions:

neutral, sadness, surprise, happiness, fear, anger, and disgust

Description of facial expression

Posed facial expressions

Illumination

N/A

Accessories

N/A

3D data

N/A

Frame rate

N/A

Ground truth

Facial expression label

Identifications of subjects

o         Reference: Please refer to the paper: Michael J. Lyons, Shigeru Akamatsu, Miyuki Kamachi & Jiro Gyoba, Coding Facial Expressions with Gabor Wavelets, Proc. of FGR98, pp. 200-205, 1998.

  1. Belfast Naturalistic Database

o         Source: this database is created by Queen's University of Belfast.

o         Purpose: This database is primarily used for emotion recognition. It may also be used to evaluate the algorithms for facial expression and facial action unit recognition under spontaneous conditions.

o         Properties:

Properties

Descriptions

# of subjects

125 (31 males and 94 females)

# of images/videos

>250

Static/Videos

Videos (audio-visual)

Single/Multiple faces

Single

Gray/Color

Color

Resolution

N/A

Face pose

Various

Facial expression

Wide range of facial expressions

Description of facial expression

Neutral-apex-neutral;

Spontaneous facial expressions

Illumination

Indoor

Accessories

N/A

3D data

N/A

Frame rate

N/A

Ground truth

Identifications of subjects

Emotional descriptors of each sequence

o         Reference: Please refer to the paper: E. Douglas-Cowie and R. Cowie and M. Schroeder, The description of naturally occurring emotional speech, 15th ICPhS, Barcelona.