DFME dataset

Background

Micro-expressions(MEs) are involuntary movements of the face that occur spontaneously in a high-stakes environment. Since MEs contain large amounts of significant and effective information about the genuine emotions, automatic MEs analysis has many potential applications such as treatment of depression, business negotiation, interrogations, and security .

Recently, automatic MEs analysis has attracted increasing attention of computer vision researchers. Especially in MEs recognition task and MEs spotting task, various models based on deep neural networks have been proposed.

Generally, MEs analysis is data-driven, which means we need sufficient MEs samples to implement efficient deep learning models with good generalization ability. However, the existing public MEs datasets usually contain few-shot samples, which restricts the development of both MEs spotting and recognition tasks. Therefore, the construction of large-scale MEs dataset is urgent and significant.

Typical ME samples with .gif in DFME

	Happiness
Anger	contempt	Disgust
Fear	Sadness	Surprise

DFME is a novel spontaneous facial micro-expressions dataset that we spent more than 3 years collecting and meticulously annotating, including three sub-data sets Part-A@500fps, Part-B@300fps, Part-C@200fps, and the ME samples are:

• collected from 671 Chinese people (male: 381, female: 290, age from 17 to 40);
• 7,526 well-labeled ME videos;
• classified into 7 emotion classes: Happy, Sad, Anger, Contempt, Fear, Disgust, Surprise;
• Please refer to our paper for more details:
  IEEE arXiv

ABOUT USTC-MEA GROUP

USTC-MEA Group, affiliated with the BDAA Lab, is a multi-disciplinary emotion analysis group, mainly focuses on the micro-reaction analysis with single or multiple modalities.
Our group has carried out exploratory research in many fields, such as micro-reaction perception instruments, micro-reaction physiological stimulation, emotional datasets, facial expression recognition and other related topics, aiming to develop micro-reaction analysis technology for interrogation, clinical diagnosis, financial risk assessment and other applications.

Download

Elements

i = 0;

while (!deck.isInOrder()) {
    print 'Iteration ' + i;
    deck.shuffle();
    i++;
}

print 'It took ' + i + ' iterations to sort the deck.';