The project involved building a machine learning model capable of classifying emotions in mice from video. Such a model is sorely needed in neuroscience, where animal models are prominent and emotions are an ongoing topic of major study [Dolensek et al, 2020]. Current methods rely on human labelling which raises problems such as observer drift, varying correlation between observers, and long analysis times [Nilsson et al, 2020]. This project involved comparing videos of mice injected with Lypopolysaccharide (LPS), which induces symptoms similar to a strong negative emotional state, with a control group in a normal emotional state. A classifier pipeline was designed that used a pose estimation model to get the position of body parts, and then aggregated the temporal pose data in a classifier built with stacked Long Short Term Memory layers. A tool to improve the pose estimation model was also created and an object detection approach was explored. The method developed correctly labelled 7 of the 8 test videos. This confirms that our approach can be used to classify emotions of mice and shows that the model is robust enough to also work on mice and physical environments outside of what it was trained on.