This study focuses on the issue of automatic Facial
Expression Recognition (FER) on little databases of 2D faces. Convolutional
Neural Networks (CNN) is a relatively new classification technique, which
reaches the state of the art on big databases; however, the use of CNN with a
scarce number of samples is still an open and interesting challenge. Following
the classical machine learning approach, we considered different combination of
appearance based projection methods, feature extraction techniques and classifiers,
and we compared their performances with special designed CNN. Experimental
results underline the drawback of CNN with scares labeled data.
Little databases facial expression sparse representation based classifier convolutional neural net
Bölüm | Computer Engineering |
---|---|
Yazarlar | |
Yayımlanma Tarihi | 20 Eylül 2017 |
Yayımlandığı Sayı | Yıl 2017 Cilt: 30 Sayı: 3 |