A comprehensive dataset of Infant Facial expressions of Pain Intensity

Facial expression plays an import role in daily life routine whether people dealing each other or diagnosing internal feelings of one another. Facial expression of pain is very much import to recognize and properly handle. There is a large number of database representing facial expression of pain, including adult’s pain databases and infant’s pain databases. Most of the databases are focusing on adults pain expressions while some deals in infant’s pain expressions. This paper represent a new infants’ pains’ facial expression database called “Infant Facial Expression of Pain Intensity” (IFEPI). This database mainly focuses on four-pain levels i.e. No pain, Mild Pain, Moderate Pain and Severe Pain. The database consist of large collection of 25000 spontaneous images collected from 120 infants of two different hospitals of Khyber Pakhtunkhwa, Pakistan, works under Health department of Pakistan Government and UNICEF. The paper represent the process from setting environment to images/frame collections to the preprocessing and refinement of the images. This database is the largest database available on pain, also there is no database, available in our knowledge, of infant which deals with four level of pain intensity. To check the validation of the proposed database we use 4 state of the art machine learning models including logistic regression, SVM, KNN and Random Forest and 2 state of the art deep learning models including LeNet and AlexNet. As the database is the largest database available, the deep learning models i.e LeNet achieves training accuracy and testing accuracy of 98.2% and 86.3% respectively while AlexNet secure training and testing accuracy of 99.1% and 89.1% respectively