This project is under the supervision of Tim Murphy lab at UBC: https://murphylab.med.ubc.ca/
To develop a method that extract fully dense, high-dimensional representation of mice behaviour and unified them in the same scale, we develop this project that perform dense tracking on mice's body based on the GAN-supervised spatial transformation network from https://github.com/wpeebles/gangealing
Our model can spatially transform the uncongealed mice to a general mice's representation. Using this information, we can densly map each individual pixel back to the uncongealed mice's body in each video frames.
This method allows analysis of dense-pose and highly fine-detailed motion on mice's body
Sample are demonstrated below. Training details, model checkpoints and analysis code are provided by request
