spoonsso
commented
2 years ago We are actively iterating on the postprocessing analysis pipeline (mostly to make it more user-friendly and easily scalable to very large datasets), but you can refer to my CAPTURE fork for a script that will run the pipeline on DANNCE mouse data (https://github.com/spoonsso/CAPTURE_demo/blob/master/Species_specific_files/pd_analysis_demo.m). This is an example for a 20-landmark mouse dataset we are using in the lab. To use it with a 22-marker mouse dataset (i.e., like the DANNCE demo), you will need to replace all instances of taddy_mouse in pd_analysis_demo.m with kyle_mouse. Any deviations from the supported skeletons will require that you create a custom Utility/load_*_anglestruct.m that defines joint angles in terms of the landmarks your skeleton uses. You can see any exsample of this in https://github.com/spoonsso/CAPTURE_demo/blob/master/Utility/load_mouse_kyle_anglestruct.m
Hi! May I ask if your team has integrated the whole postprocessing analysis, including DANNCE and CAPTURE? I found there are some different descriptions between DANNCE and CAPTURE which make me a bit confused, (e.g. impute method, movement threshold...). So I would like to know if you have any scripts for generating the behavior map based on DANNCE predicted models. Any suggestions will be appreciated. Many thanks!