-
GSN: Generalisable Segmentation in Neural Radiance Field
Authors:
Vinayak Gupta,
Rahul Goel,
Sirikonda Dhawal,
P. J. Narayanan
Abstract:
Traditional Radiance Field (RF) representations capture details of a specific scene and must be trained afresh on each scene. Semantic feature fields have been added to RFs to facilitate several segmentation tasks. Generalised RF representations learn the principles of view interpolation. A generalised RF can render new views of an unknown and untrained scene, given a few views. We present a way t…
▽ More
Traditional Radiance Field (RF) representations capture details of a specific scene and must be trained afresh on each scene. Semantic feature fields have been added to RFs to facilitate several segmentation tasks. Generalised RF representations learn the principles of view interpolation. A generalised RF can render new views of an unknown and untrained scene, given a few views. We present a way to distil feature fields into the generalised GNT representation. Our GSN representation generates new views of unseen scenes on the fly along with consistent, per-pixel semantic features. This enables multi-view segmentation of arbitrary new scenes. We show different semantic features being distilled into generalised RFs. Our multi-view segmentation results are on par with methods that use traditional RFs. GSN closes the gap between standard and generalisable RF methods significantly. Project Page: https://vinayak-vg.github.io/GSN/
△ Less
Submitted 7 February, 2024;
originally announced February 2024.
-
StyleTRF: Stylizing Tensorial Radiance Fields
Authors:
Rahul Goel,
Sirikonda Dhawal,
Saurabh Saini,
P. J. Narayanan
Abstract:
Stylized view generation of scenes captured casually using a camera has received much attention recently. The geometry and appearance of the scene are typically captured as neural point sets or neural radiance fields in the previous work. An image stylization method is used to stylize the captured appearance by training its network jointly or iteratively with the structure capture network. The sta…
▽ More
Stylized view generation of scenes captured casually using a camera has received much attention recently. The geometry and appearance of the scene are typically captured as neural point sets or neural radiance fields in the previous work. An image stylization method is used to stylize the captured appearance by training its network jointly or iteratively with the structure capture network. The state-of-the-art SNeRF method trains the NeRF and stylization network in an alternating manner. These methods have high training time and require joint optimization. In this work, we present StyleTRF, a compact, quick-to-optimize strategy for stylized view generation using TensoRF. The appearance part is fine-tuned using sparse stylized priors of a few views rendered using the TensoRF representation for a few iterations. Our method thus effectively decouples style-adaption from view capture and is much faster than the previous methods. We show state-of-the-art results on several scenes used for this purpose.
△ Less
Submitted 19 December, 2022;
originally announced December 2022.
-
Real-Time Rendering of Arbitrary Surface Geometries using Learnt Transfer
Authors:
Sirikonda Dhawal,
Aakash KT,
P. J. Narayanan
Abstract:
Precomputed Radiance Transfer (PRT) is widely used for real-time photorealistic effects. PRT disentangles the rendering equation into transfer and lighting, enabling their precomputation. Transfer accounts for the cosine-weighted visibility of points in the scene while lighting for emitted radiance from the environment. Prior art stored precomputed transfer in a tabulated manner, either in vertex…
▽ More
Precomputed Radiance Transfer (PRT) is widely used for real-time photorealistic effects. PRT disentangles the rendering equation into transfer and lighting, enabling their precomputation. Transfer accounts for the cosine-weighted visibility of points in the scene while lighting for emitted radiance from the environment. Prior art stored precomputed transfer in a tabulated manner, either in vertex or texture space. These values are fetched with interpolation at each point for shading. Vertex space methods require densely tessellated mesh vertices for high quality images. Texture space methods require non-overlap** and area-preserving UV map** to be available. They also require a high-resolution texture to avoid rendering artifacts. In this paper, we propose a compact transfer representation that is learnt directly on scene geometry points. Specifically, we train a small multi-layer perceptron (MLP) to predict the transfer at sampled surface points. Our approach is most beneficial where inherent mesh storage structure and natural UV map** are not available, such as Implicit Surfaces as it learns the transfer values directly on the surface. We demonstrate real-time, photorealistic renderings of diffuse and glossy materials on SDF geometries with PRT using our approach.
△ Less
Submitted 19 December, 2022;
originally announced December 2022.
-
PRTT: Precomputed Radiance Transfer Textures
Authors:
Sirikonda Dhawal,
Aakash KT,
P. J. Narayanan
Abstract:
Precomputed Radiance Transfer (PRT) can achieve high quality renders of glossy materials at real-time framerates. PRT involves precomputing a k-dimensional transfer vector of Spherical Harmonic (SH) coefficients at specific points for a scene. Most prior art precomputes transfer at vertices of the mesh and interpolates color for interior points. They require finer mesh tessellations for high quali…
▽ More
Precomputed Radiance Transfer (PRT) can achieve high quality renders of glossy materials at real-time framerates. PRT involves precomputing a k-dimensional transfer vector of Spherical Harmonic (SH) coefficients at specific points for a scene. Most prior art precomputes transfer at vertices of the mesh and interpolates color for interior points. They require finer mesh tessellations for high quality renderings. In this paper, we explore and present the use of textures for storing transfer. Using transfer textures decouples mesh resolution from transfer storage and sampling which is useful especially for glossy renders. We further demonstrate glossy inter-reflections by precomputing additional textures. We thoroughly discuss practical aspects of transfer textures and analyze their performance in real-time rendering applications. We show equivalent or higher render quality and FPS and demonstrate results on several challenging scenes.
△ Less
Submitted 23 March, 2022;
originally announced March 2022.
