The article outlines an integrated framework for studying brain activity linked to natural behavior in freely moving mice. This includes Miniature Two-Photon Microscopy (mTPM) for neural imaging, deep learning-based 3D behavioral tracking, unsupervised behavior decomposition using the Social Behavior Atlas (SBeA) and Behavior Atlas (BeA), and CEBRA, a method for joint mapping of neural and behavioral data. The goal is to decode real-world behavior from neural population dynamics. The research used mTPM system for simultaneous high-resolution neural imaging and behavior capture in freely behaving mice, achieving synchronization with a TTL signal. This allows stable imaging of individual neurons. Behavioral tracking was done with ADPT, a transformer-based pose tracker that reduces drift and reconstructs 3D poses for interacting mice. By employing unsupervised clustering methods, human bias is minimized. CEBRA produces joint latent spaces to illustrate the relationship between neural activity and behavior.