Credits

Smile detection runs entirely in your browser. Your camera image never leaves your device for it. Google's MediaPipe Face Landmarker reads facial expression scores and face shape, and a small model of ours decides whether you are smiling. That model was built and tested with the datasets below.

http://mplab.ucsd.edu, The MPLab GENKI Database, GENKI-4K Subset. Machine Perception Laboratory, University of California, San Diego (MPLab GENKI Database). Used to train and test the smile model.

Open Images Dataset by Google. Annotations licensed under CC BY 4.0; images are listed as having a CC BY 2.0 license by their Flickr authors. We used human-verified "Smile" and "Laugh" labels on face crops to train and test the smile model; no images are shown or redistributed here. Kuznetsova et al., "The Open Images Dataset V4", IJCV 2020.

Kärkkäinen & Joo, "FairFace: Face Attribute Dataset for Balanced Race, Gender, and Age", WACV 2021 (FairFace, CC BY 4.0). Used to check that faces are detected about equally well across groups.

MediaPipe (Apache License 2.0) runs face tracking and laugh-sound detection in your browser.