/FYP-SAM3D-VPCC

A project which adapts the SAM model to perform point cloud segmentation for VPCC. This project is a Final Year Project undertaken at the School of Computing, National University of Singapore.

Primary LanguageJupyter Notebook

SAM3D-VPCC

The SAM3D-VPCC project utilizes Meta's Segment Anything Model (SAM) to generate 2D projections of 3D point clouds, which can be used by VPCC during point cloud compression.

In doing so, we can hide visual artifacts created during compression by having them appear in more natural boundaries between segments in the point cloud.

Getting Started

To run the example notebook, sam3d, you will need to install segment_anything alongside its dependencies in your local environment, following the official installation instructions in SAM.

Models for SAM

The models used by SAM can be downloaded from the official SAM repository and placed in the models directory.

Point Cloud Data

The point cloud data used, longdress, is provided by 8i Labs at the JPEG Pleno Database.