Markerless Rat Behavior Quantification With Cascade Neural Network

Quantifying rat behavior through video surveillance is crucial for medicine, neuroscience and other fields. In this paper, we focus on the challenging problem of estimating the landmark points such as rat’s eyes and joints only with the image processing, and quantify the motion behavior of the rat. Firstly, we put the rat on a special running machine and use a high frame rate camera to capture its motion. Secondly, we designed the cascade convolution network (CCN) and cascade hourglass network (CHN) two structures to extract the feature of the images respectively, and used fully connected regression (FCR), heatmap maximum position (HMP), heatmap integral regression (HIR) three coordinate calculation methods to locate the coordinates of the landmark points. Thirdly, though a strict normalized evaluation criterion, we analyzed the accuracy of different structures and coordinate calculation methods for rat landmark points estimation in different feature map sizes. Finally, according to the results, we found that CCN structure with HIR method have the highest estimation accuracy 75%, which is enough to track and quantify rat joint motion.