Point Cloud Processing and Coding using Deep Learning
Point Clouds are becoming one of the most common data structures to represent 3D scenes as it enables six degrees of freedom (6DoF) viewing experience. Point Cloud Data has been used in many applications, from immersive media, and autonomous driving to healthcare.
However, a typical point cloud contains millions of 3D points and requires a huge amount of storage. Hence, efficient Point Cloud Compression (PCC) methods are just inevitable in order to bring point cloud into practical applications. With the help of recent advances in deep learning techniques [Minkowski Engine, PointNet, …], in this thesis, we will tackle challenges in Point Cloud Processing/Coding including sparsity, high dimensional volume, and irregularity. The possible thesis includes:
- Studying Point Cloud representations (Voxel, Octree, Point, tri-soup, …)
- Computational complexity optimization for existing point cloud coding methods [SparseVoxelDNN]
- Neural network-based point cloud processing (classification, segmentation, …)
- Point cloud texture generation
- …
Professor:
Prof. Dr.-Ing. André Kaup
Supervisior:
M.Sc. Dat Nguyen, Room 06.026, dat.thanh.nguyen@fau.de
Prerequisites:
Python, experiences with Deep Learning,
Pytorch and Pytorch Lightning (optional)
Available:
Immediately
References:
[SparseVoxelDNN] Nguyen, Dat Thanh, and Andre Kaup. “Learning-based Lossless Point Cloud Geometry Coding using Sparse Representations.” arXiv preprint arXiv:2204.05043 (2022).
[MinkowskiEngine] Choy, Christopher, JunYoung Gwak, and Silvio Savarese. “4d spatio-temporal convnets: Minkowski convolutional neural networks.” Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition. 2019.