Editing Eye tracking (section)

=== Safety applications ===
Scientists in 2017 constructed a Deep Integrated Neural Network (DINN) out of a Deep Neural Network and a convolutional neural network.<ref name=":0">{{Cite journal|last1=Zhao|first1=Lei|last2=Wang|first2=Zengcai|last3=Zhang|first3=Guoxin|last4=Qi|first4=Yazhou|last5=Wang|first5=Xiaojin|date=15 November 2017|title=Eye state recognition based on deep integrated neural network and transfer learning|journal=Multimedia Tools and Applications|volume=77|issue=15|pages=19415–19438|doi=10.1007/s11042-017-5380-8|s2cid=20691291|issn=1380-7501}}</ref> The goal was to use [[deep learning]] to examine images of drivers and determine their level of drowsiness by "classify[ing] eye states." With enough images, the proposed DINN could ideally determine when drivers blink, how often they blink, and for how long. From there, it could judge how tired a given driver appears to be, effectively conducting an eye-tracking exercise. The DINN was trained on data from over 2,400 subjects and correctly diagnosed their states 96%-99.5% of the time. Most other artificial intelligence models performed at rates above 90%.<ref name=":0" /> This technology could ideally provide another avenue for [[driver drowsiness detection]].