Emissive media are often challenging to render: in thin regions where only few scattering events occur the emission is poorly sampled, while sampling events for emission can be disadvantageous due to ...absorption in dense regions. We extend the standard path space measurement contribution to also collect emission along path segments, not only at vertices. We apply this extension to two estimators: extending paths via scattering and distance sampling, and next event estimation. In order to do so, we unify the two approaches and derive the corresponding Monte Carlo estimators to interpret next event estimation as a solid angle sampling technique. We avoid connecting paths to vertices hidden behind dense absorbing layers of smoke by also including transmittance sampling into next event estimation. We demonstrate the advantages of our line integration approach which generates estimators with lower variance since entire segments are accounted for. Also, our novel forward next event estimation technique yields faster run times compared to previous next event estimation as it penetrates less deeply into dense volumes.
Inspired by vector field topology, an established tool for the extraction and identification of important features of flows and vector fields, we develop means for the analysis of the structure of ...light transport. For that, we derive an analogy to vector field topology that defines coherent structures in light transport. We also introduce Finite‐Time Path Deflection (FTPD), a scalar quantity that represents the deflection characteristic of all light transport paths passing through a given point in space. For virtual scenes, the FTPD can be computed directly using path‐space Monte Carlo integration. We visualize the FTPD field for several example scenes and discuss the revealed structures. Lastly, we show that the coherent regions visualized by the FTPD are closely related to the coherent regions in our new topologically‐motivated analysis of light transport. FTPD visualizations are thus also visualizations of the structure of light transport.
Markov Chain Monte Carlo (MCMC) rendering is extensively studied, yet it remains largely unused in practice. We propose solutions to several practicability issues, opening up path space MCMC to ...become an adaptive sampling framework around established Monte Carlo (MC) techniques. We address non-uniform image quality by deriving an analytic target function for imagespace sample stratification. The function is based on a novel connection between variance and path differentials, allowing analytic variance estimates for MC samples, with potential uses in other adaptive algorithms outside MCMC. We simplify these estimates down to simple expressions using only quantities known in any MC renderer. We also address the issue that most existing MCMC renderers rely on bi-directional path tracing and reciprocal transport, which can be too costly and/or too complex in practice. Instead, we apply our theoretical framework to optimize an adaptive MCMC algorithm that only uses forward path construction. Notably, we construct our algorithm by adapting (with minimal changes) a full-featured path tracer into a single-path state space Markov Chain, bridging another gap between MCMC and existing MC techniques.
Displacement mapping is routinely used to add geometric details in a fast and easy‐to‐control way, both in offline rendering as well as recently in interactive applications such as games. However, it ...went largely unnoticed (with the exception of McGuire and Whitson MW08) that, when applying displacement mapping to a surface with a low‐distortion parametrization, this parametrization is distorted as the geometry was changed by the displacement mapping. Typical resulting artifacts are “rubber band”‐like distortion patterns in areas of strong displacement change where a small isotropic area in texture space is mapped to a large anisotropic area in world space. We describe a fast, fully GPU‐based two‐step procedure to resolve this problem. First, a correction deformation is computed from the displacement map. Second, two variants to apply this correction when computing displacement mapping are proposed. The first variant is backward‐compatible and can resolve the artifact in any rendering pipeline without modifying it and without requiring additional computation at render time, but only works for bijective parametrizations. The second variant works for more general parametrizations, but requires to modify the rendering code and incurs a very small computational overhead.
In this paper we present a novel GPU-friendly real-time voxelization technique for rendering homogeneous media that is defined by particles, e.g., fluids obtained from particle-based simulations such ...as Smoothed Particle Hydrodynamics (SPH). Our method computes view-adaptive binary voxelizations with on-the-fly compression of a tiled perspective voxel grid, achieving higher resolutions than previous approaches. It allows for interactive generation of realistic images, enabling advanced rendering techniques such as ray casting-based refraction and reflection, light scattering and absorption, and ambient occlusion. In contrast to previous methods, it does not rely on preprocessing such as expensive, and often coarse, scalar field conversion or mesh generation steps. Our method directly takes unsorted particle data as input. It can be further accelerated by identifying fully populated simulation cells during simulation. The extracted surface can be filtered to achieve smooth surface appearance. Finally, we provide a new scheme for accelerated ray casting inside the voxelization.
Samples with high contribution but low probability density, often called fireflies, occur in all practical Monte Carlo estimators and are part of computing unbiased estimates. For finite‐sample ...estimates, however, they can lead to excessive variance. Rejecting all samples classified as outliers, as suggested in previous work, leads to estimates that are too low and can cause undesirable artefacts. In this paper, we show how samples can be re‐weighted depending on their contribution and sampling frequency such that the finite‐sample estimate gets closer to the correct expected value and the variance can be controlled. For this, we first derive a theory for how samples should ideally be re‐weighted and that this would require the probability density function of the optimal sampling strategy. As this probability density function is generally unknown, we show how the discrepancy between the optimal and the actual sampling strategy can be estimated and used for re‐weighting in practice. We describe an efficient algorithm that allows for the necessary analysis of per‐pixel sample distributions in the context of Monte Carlo rendering without storing any individual samples, with only minimal changes to the rendering algorithm. It causes negligible runtime overhead, works in constant memory and is well suited for parallel and progressive rendering. The re‐weighting runs as a fast post‐process, can be controlled interactively and our approach is non‐destructive in that the unbiased result can be reconstructed at any time.
Samples with high contribution but low probability density, often called fireflies, occur in all practical Monte Carlo estimators and are part of computing unbiased estimates. For finite‐sample estimates, however, they can lead to excessive variance. Rejecting all samples classified as outliers, as suggested in previous work, leads to estimates that are too low and can cause undesirable artefacts. In this paper, we show how samples can be re‐weighted depending on their contribution and sampling frequency such that the finite‐sample estimate gets closer to the correct expected value and the variance can be controlled. For this, we first derive a theory for how samples should ideally be re‐weighted and that this would require the probability density function of the optimal sampling strategy. As this probability density function is generally unknown, we show how the discrepancy between the optimal and the actual sampling strategy can be estimated and used for re‐weighting in practice. We describe an efficient algorithm that allows for the necessary analysis of per‐pixel sample distributions in the context of Monte Carlo rendering without storing any individual samples, with only minimal changes to the rendering algorithm.
We introduce a practical general-purpose neural appearance filtering pipeline for physically-based rendering. We tackle the previously difficult challenge of aggregating visibility across many levels ...of detail from local information only, without relying on learning visibility for the entire scene. The high adaptivity of neural representations allows us to retain geometric correlations along rays and thus avoid light leaks. Common approaches to prefiltering decompose the appearance of a scene into volumetric representations with physically-motivated parameters, where the inflexibility of the fitted models limits rendering accuracy. We avoid assumptions on particular types of geometry or materials, bypassing any special-case decompositions. Instead, we directly learn a compressed representation of the intra-voxel light transport. For such high-dimensional functions, neural networks have proven to be useful representations. To satisfy the opposing constraints of prefiltered appearance and correlation-preserving point-to-point visibility, we use two small independent networks on a sparse multi-level voxel grid. Each network requires 10--20 minutes of training to learn the appearance of an asset across levels of detail. Our method achieves 70--95% compression ratios and around 25% of quality improvements over previous work. We reach interactive to real-time framerates, depending on the level of detail.
We present a novel method to optimize the attenuation of light for the single scattering model in direct volume rendering. A common problem of single scattering is the high dynamic range between lit ...and shadowed regions due to the exponential attenuation of light along a ray. Moreover, light is often attenuated too strong between a sample point and the camera, hampering the visibility of important features. Our algorithm employs an importance function to selectively illuminate important structures and make them visible from the camera. With the importance function, more light can be transmitted to the features of interest, while contextual structures cast shadows which provide visual cues for perception of depth. At the same time, more scattered light is transmitted from the sample point to the camera to improve the primary visibility of important features. We formulate a minimization problem that automatically determines the extinction along a view or shadow ray to obtain a good balance between sufficient transmittance and attenuation. In contrast to previous approaches, we do not require a computationally expensive solution of a global optimization, but instead provide a closed-form solution for each sampled extinction value along a view or shadow ray and thus achieve interactive performance.
Hardware Acceleration of Neural Graphics Mubarik, Muhammad Husnain; Kanungo, Ramakrishna; Zirr, Tobias ...
Proceedings of the 50th Annual International Symposium on Computer Architecture,
06/2023
Conference Proceeding
Rendering and inverse rendering techniques have recently attained powerful new capabilities and building blocks in the form of neural representations (NR), with derived rendering techniques quickly ...becoming indispensable tools next to classic computer graphics algorithms, covering a wide range of functions throughout the full pipeline from sensing to pixels. NRs have recently been used to directly learn the geometric and appearance properties of scenes that were previously hard to capture, and to re-synthesize photo realistic imagery based on this information, thereby promising simplifications and replacements for several complex traditional computer graphics problems and algorithms with scalable quality and predictable performance. In this work we ask the question: Does neural graphics (graphics based on NRs) need hardware support? We studied four representative neural graphics applications (NeRF, NSDF, NVR, and GIA) showing that, if we want to render 4k resolution frames at 60 frames per second (FPS) there is a gap of ~ 1.51× to 55.50× in the desired performance on current GPUs. For AR and VR applications, there is an even larger gap of ~ 2--4 orders of magnitude (OOM) between the desired performance and the required system power. We identify that the input encoding and the multi-layer perceptron kernels are the performance bottlenecks, consuming 72.37%, 60.0% and 59.96% of application time for multi resolution hashgrid encoding, multi resolution densegrid encoding and low resolution densegrid encoding, respectively. We propose a neural graphics processing cluster (NGPC) - a scalable and flexible hardware architecture that directly accelerates the input encoding and multi-layer perceptron kernels through dedicated engines and supports a wide range of neural graphics applications. To achieve good overall application level performance improvements, we also accelerate the rest of the kernels by fusion into a single kernel, leading to a ~ 9.94× speedup compared to previous optimized implementations 17 which is sufficient to remove this performance bottleneck. Our results show that, NGPC gives up to 58.36× end-to-end application-level performance improvement, for multi resolution hashgrid encoding on average across the four neural graphics applications, the performance benefits are 12.94×, 20.85×, 33.73× and 39.04× for the hardware scaling factor of 8, 16, 32 and 64, respectively. Our results show that with multi resolution hashgrid encoding, NGPC enables the rendering of 4k Ultra HD resolution frames at 30 FPS for NeRF and 8k Ultra HD resolution frames at 120 FPS for all our other neural graphics applications.
In the wake of many new ML-inspired approaches for reconstructing and representing high-quality 3D content, recent hybrid and explicitly learned representations exhibit promising performance and ...quality characteristics. However, their scaling to higher dimensions is challenging, e.g. when accounting for dynamic content with respect to additional parameters such as material properties, illumination, or time. In this paper, we tackle these challenges for an explicit representations based on Gaussian mixture models. With our solutions, we arrive at efficient fitting of compact N-dimensional Gaussian mixtures and enable efficient evaluation at render time: For fast fitting and evaluation, we introduce a high-dimensional culling scheme that efficiently bounds N-D Gaussians, inspired by Locality Sensitive Hashing. For adaptive refinement yet compact representation, we introduce a loss-adaptive density control scheme that incrementally guides the use of additional capacity towards missing details. With these tools we can for the first time represent complex appearance that depends on many input dimensions beyond position or viewing angle within a compact, explicit representation optimized in minutes and rendered in milliseconds.