A DESIGN OF SMALL SCALE DEEP CNN MODEL FOR FACIAL EXPRESSION RECOGNITION USING THE LOW-RESOLUTION IMAGE DATASETS

Abstract

Since artificial intelligence is becoming an important part of our lives providing incredible facilities, researchers worldwide are suggesting better applications. Researchers working on various subfields of Artificial Intelligence, including Natural language processing, Expert systems, Speech recognition, and Computer vision are trying to make life easier for mankind. Within those, computer vision is one of the most important fields, since cameras are installed in many places nowadays and being installed in more and more places. One of the trending topics would be facial analysis. From this point of view, facial expression recognition has been one of the hot topics among computer vision researchers in recent decades. Surely, many approaches are proposed to tackle the issue. It can be argued that learning the small dataset with Deep CNN has an over fitting problem compared to that with the big dataset. The reason is that smaller images have a small number of features or not clear features, so deeper models would learn the limited number of futures, possibly causing over fitting later. Thus, a more reliable model that works well for face images with low resolution is necessary. To this end, we propose our method which we believe is more reliable for low-resolution images. We show experimental results on FER2013 and FERPlus datasets to prove how well our architecture classifies low-resolution images.