Development and Validation of an Automated Video Tracking Model for Stabled Horses.

摘要: Changes in behaviour are often caused by painful conditions. Therefore, the assessment of is important for recognition pain, but also quality life. Automated detection movement and a horse box stall should represent significant advancement. In this study, videos horses an animal hospital were recorded using action camera time-lapse mode. These processed convolutional neural network Loopy automated prediction body parts. Development model was carried out several steps, including annotation key points, training to generate checking its accuracy. The points nose, withers tail detected with sensitivity more than 80% error rate between 2 7%, depending on point. By means case possibility further analysis acquired data investigated. results will significantly improve pain help develop algorithms machine learning.