Annotation of a subset of the AR Face database : 40 images of frontal faces have been manually labeled 10 times each.
We labeled the following faces (name of the AR images, because we cannot distribute the database) :
m-046-14 m-047-14 m-048-14 m-049-14 m-051-14 m-052-14 m-053-14 m-054-14 m-055-14 m-056-14 m-058-14 m-059-14 m-060-14 m-061-14 m-065-14 m-066-14 m-067-14 m-069-14 m-070-14 m-071-14 m-072-14 m-073-14 m-074-14 m-075-14 m-076-14 m-001-1 m-002-1 m-004-1 m-005-1 m-009-1 m-010-1 m-012-1 m-013-1 m-014-1 m-015-1 m-018-1 m-020-1 m-022-1 m-023-1 m-024-1
We used the following model triangulation :
Extraction and annotations of some images that belong to the MMI Face Database
We selected from the MMI database, 6 different identities each performing 18 isolated facial actions. Each image has been manually labelled with a 35-points shape model. The retained facial actions are :
| Action Unit | Description |
|---|---|
| AU 34 | Cheek puff |
| AU 33 | Cheek blow |
| AU 35 | Cheek suck |
| AU 36L | Tongue bulge (Left) |
| AU 36R | Tongue bulge (Right) |
| AU 36T | Tongue bulge (Top) |
| AU 36B | Tongue bulge (Bottom) |
| AU 12+6 | Smile |
| AU 12+6 | Smile, visible teeth |
| AU 7 | Eye lid tight |
| AU 43 | Eyes closed |
| AU 5 | Eyes wide-opened |
| AU 2 | Eyebrow raised |
| AU 15 | ?? |
| AU 26 | Opened mouth |
| AU 19 | Visible tongue |
| AU 18 | Lips |
| AU 28 | Lips pinched |
Archive containing the 19 x 6 manual labellings