This repository is a collection of different work-in-progress approaches in unsupervised machine learning for the quantification of animal behavior.
The main analysis pipeline does or will soon contain:
- markerless pose estimation from video data with DeepLabCut
- multi-view triangulation and 3D reconstruction with Anipose
- UMAP dimensionality reduction
- stochastic modeling of time series data with Hidden Markov Models (HMM)
- behavioral clustering with Variational Embeddings of Animal Motion (VAME), Toeplitz Inverse Covariance-Based Clustering (TICC) and others