“APhotoEveryday” (APE) dataset contains the faces (images or videos) of 98 individuals with large appearance variations across time. The APE dataset was acquired by the University of Cagliari and consists of facial images extracted from YouTube Daily Photo Projects. Faces were captured in the frontal pose and most of these have a controlled expression. The images of each user are labelled with a number that indicates the temporal progression of the sequence. The APE dataset images are characterized by many temporal variations of the facial appearance thanks to the Daily Photo Project (See for example: https://www.youtube.com/watch?v=RBPYDIzEbYk). The number of images per user varies between 92 and 3592 and the acquisition time varies between less than one year and seventeen years.
The dataset has two versions:
- v1: Contains 45 users
- v2: The dataset has been extended up to 98 users.
The url of the youtube videos used for the acquisition of the dataset are indicated in this link.
To obtain the dataset with the images of the faces extracted and labeled, follow the instructions below.
- Download license agreement
- Compile and send us (livdet.diee@unica.it) the License Agreements signed by the Requestor of the data and by a legal representative of you institution. The document must be certified by the stamp of your company or institution or by an electronic signature.
G. Orrù, G. L. Marcialis and F. Roli, "An experimental investigation on self adaptive facial recognition algorithms using a long time span data set," 2018 Eighth International Conference on Image Processing Theory, Tools and Applications (IPTA), Xi'an, 2018, pp. 1-6, doi: 10.1109/IPTA.2018.8608134.
G. Orrù, M. Micheletto, J. Fierrez and G. L. Marcialis “Are Adaptive Face Recognition Systems still Necessary? Experiments on the APE Dataset” 2020 Fourth IEEE International Conference on Image Processing, Applications and Systems (IPAS), Genova, 2020, in press. Arxiv: https://arxiv.org/pdf/2010.04072.pdf