posted on 2021-01-05, 05:16authored byGowoon Son, Seung-Jun Yoo, Shinwoo Kang, Ameer Rasheed, Da Hae Jung, Hyunjun Park, Bongki Cho, Harry W. M. Steinbusch, Keun-A Chang, Yoo-Hun Suh, Cheil Moon
Additional file 2: Movie S1 Representative clip of trained moving mouse for analysis of odor detection test using DeepLabCut. Four points of interest (POIs) that we tracked in each frame were the nose (blue), ears (light blue and yellow), and tail (red). Randomly selected 180 frames and manually labeled POIs in those frames, and used them to train and test a neural network model implemented in DeepLabCut. Evaluation of labeling accuracy was achieved by comparing the labels acquired from the convolutional neural network on the test set with manual labels. The model was then used to evaluate all frames in each group of the 20 videos used for training. The resulting x and y coordinates corresponding to the middle position of four POIs within each frame were used to determine location.
Funding
Korea Health Industry Development Institute National Research Foundation of Korea