|Topic:||Accelerating Unified Neural Implicit Surfaces and Radiance Fields with Hash-Encoded Neural Networks|
|Supervisor:||Mgr. Jonáš Kulhánek|
|Announce as:||Diplomová práce, Bakalářská práce, Semestrální projekt|
|Description:||Novel view synthesis is a long-standing computer vision problem with a significant impact in other fields such as computer graphics and robotics. In this thesis, the student will work with the newest state-of-the-art neural rendering methods such as [1,2]. Current Neural Radiance Field (NeRF) methods use neural networks to represent scenes and a differentiable rendering algorithm to optimize the representation. Although the methods often achieve photorealistic results , the internal representation often cannot be reliably used to reconstruct the underlying geometry, and the inference speed without caching is slow. In , the NeRF was combined with surface rendering to achieve more meaningful representation and faster inference. In , the authors achieved extremely fast training and inference speed by optimizing a hash-encoded embedding and neural network for novel GPU architectures. This thesis aims to combine the two and use neural graphical primitives from  to accelerate the UNISURF method . In theory, the final method could be used to efficiently reconstruct the underlying geometry from images with known camera parameters.
Although this thesis can be challenging, if it goes as expected, the results can be presented or published in a prestigious conference or journal (e.g., CVPR, ICCV), which can benefit the student’s career greatly.
|Bibliography:||: Müller, T., Evans, A., Schied, C., & Keller, A. (2022). Instant Neural Graphics Primitives with a Multiresolution Hash Encoding. arXiv:2201. 05989.
: Oechsle, M., Peng, S., & Geiger, A. (2021). Unisurf: Unifying neural implicit surfaces and radiance fields for multi-view reconstruction. In Proceedings of the IEEE/CVF International Conference on Computer Vision (pp. 5589-5599).
: Barron, Jonathan T., et al. 'Mip-NeRF: A Multiscale Representation for Anti-Aliasing Neural Radiance Fields.' arXiv preprint arXiv:2103.13415 (2021).