Emotion recognition is an upcoming research area with strong applications in the next-generation communications using digital modes. To develop robust algorithms for this, we are looking for some good datasets.
One of the most complete datasets for this, is the Aff-Wild2 dataset (https://ibug.doc.ic.ac.uk/resources/aff-wild2/), with 558 video containing around 2.8 millions frames, with valence arousal estimation, action unit detection and basic expression classification as annotations.
Another popular (although a bit older) dataset is the extended Cohn-Kanade (CK+) dataset, containing 593 image sequences, annotated with facial expression labels and action unit labels (Conference Paper The Extended Cohn-Kanade Dataset (CK+): A complete dataset f...
, http://www.jeffcohn.net/Resources/).
this should help you: https://paperswithcode.com/task/emotion-recognition/latest
Have a look at https://www.robots.ox.ac.uk/~vgg/data/ and https://www.kaggle.com/c/emotion-detection-from-facial-expressions
https://www.visualdata.io/discovery
- Badges
- Science topic