The project implements se(3)-TrackNet, a neural network architecture for long-term 6D pose tracking that operates on RGB-D observations combined with synthetic image conditioning. It addresses domain shift between synthetic training data and real-world images through disentangled feature encoding and a Lie Algebra representation for 3D orientation. The system achieves real-time tracking at 90.9Hz for robotic manipulation tasks including model-based RL, AR/VR, and human-robot interaction.