Need help setting up google collab model for deep lap cut to analyze regions of interest for mice behavior on an elevated zero maze

I am currently trying to apply the deep lab cut pose estimation software to the simBA program for my research project that has recorded mice movement on an elevated zero maze. The videos are from an overhead angle where the 2 open regions of interest are in clear view, however the mouse is often occluded by the walls of the maze and out of sight but the open regions are clearly visible and these are the regions of interest. Additionally, the angle of the videos changes slightly several times over the course of the project. I would need someone to set up a shared google colab model with the deep lab cut package in it that would generate accurate labeled CSV files to be used in the simBA program for advanced analytics for mice movement. There would likely have to be multiple iterations of this model given the angle changes would require new training for the different sets of videos. This would require you to extract the csv files from the deep lab cut to run them through simBA and find the necessary parameters for outlier correction and visualization generation for accurate pose estimation.