Fig 1.
Methodology steps for building the machine vision algorithm for pain assessment in horses.
Fig 2.
Horse face regions and pain levels for evaluation by computer modeling.
Ears: Considers ear position, distance between them and direction. Eyes: Considers size of the orbital area, muscular tension above the eye, visibility of underlying bone surfaces. Mouth and nostrils: Considers level of mouth and nostrils straining, tension, size of mouth-collum (line between the upper and lower lips) and nostril dilation.
Fig 3.
Window application software for the labeling process.
Table 1.
Composition of final image database.
Table 2.
Model architecture based on the convolutional neural network that was used to classify the pain level for each evaluated parameter (ears, eye and mouth and nostrils).
Table 3.
Hyperparameters used in the final classifier model for each evaluated parameter (ears, eye, and mouth and nostrils).
Table 4.
Confusion matrix between the data classified by the CNN-based model and HGS for the ears images.
Table 5.
Confusion matrix between the data classified by the CNN-based model and HGS for the eyes images.
Table 6.
Confusion matrix between the data classified by the CNN-based model and HGS for the mouth and nostrils images.
Table 7.
Confusion matrix between the data classified by the ANN-based classifier and HGS, described on three levels of pain.
Table 8.
Confusion matrix between the data classified by the ANN-based classifier and HGS, described on two levels of pain.