Point Cloud Semantic Segmentation is the task of recognize objects inside a three-dimensional scene,
assigning to each point a label that determine to which object the point belongs.
Incremental Learning is a well-known problem in Deep Learning aiming at developing artificially intelligent systems that can continuously learn to address new tasks from new data while preserving knowledge learned from previously learned tasks.

The objective of this thesis is to perform incremental learning in point cloud semantic segmentation. State-of-the-art models (RandLA-Net, RangeNet++, Cylinder3D etc.) and popular datasets (e.g., SemanticKITTI, S3DIS etc.) will be exploit for the task.

For more information,  please contact prof. Simone Milani at simone.milani@dei.unipd.it.