Deep learning pipeline for 6D pose estimation (translation and rotation) of rigid objects from RGB-D images, based on a modified PointNet++ architecture. It leverages both geometric and image features for high-accuracy results tailored for robotic manipulation tasks.

Interactive PointNet++ workshop for point cloud deep learning. Train a deep learning model to classify 3D shapes, visualize features. GPU-optimized Jupyter notebook.