Abstract Spatiotemporal resampling methods such as ReSTIR decrease noise in Monte Carlo rendering of dynamic content by reusing paths across frames and pixels. Standard ReSTIR reuses spatially from a small number of randomly selected neighbors. This reuse suffers when few neighbors contain contributing samples, reducing quality toward that of the underlying path sampler. This commonly occurs during camera or object motion, as regions not present in prior frames are revealed. Increasing the number of spatial neighbors helps but also increases cost. We propose a novel spatial neighbor selection technique, stochastic pairwise MIS, which enables unbiased reuse from many neighbors in real time and focuses reuse on pixels with contributing samples. This provides a significant increase in image quality overall, especially in regions with poor input samples.

Monte Carlo integration is widely used in computer graphics, especially in rendering, but we identify two key limitations. First, for each sample, we can often obtain rich auxiliary information, such as sample position, or geometric information. However, a classical Monte Carlo estimator cannot effectively use this information and only averages the function values. Second, the Monte Carlo formulation makes it difficult to adapt the sampling distribution toward truly informative regions, which can be critical for reconstructing the signal. To address these limitations, we argue that a sampler and integrator beyond the standard Monte Carlo methods is needed. We therefore propose an end-to-end sampling-integration approach that jointly learns both a sampler and an integrator using neural networks, enabling samples to be drawn and used in a more coupled and principled manner. By training on a dataset of integrands, the estimator can further use learned priors over integrand structure and specialize to a family of problems. We evaluate our method on diverse applications in lighting, transmittance, generalized winding number, and walk-on-spheres, spanning both linear and nonlinear cases, and a broad range of low- and high-dimensional settings. Even though the networks add computational overheads, in the equal-sample setting, our method achieves substantial improvements, providing a powerful alternative to traditional quadrature rules and sampling methods.

Computer graphics applications are sophisticated heterogeneous programs that orchestrate numerous components, including shaders, pipelines, resource descriptors, and command streams. In modern graphics workflows, communication between these components is mediated by resources such as shader I/O, storage buffers, and resource bindings. To make this communication correct, developers must uphold resource contracts. These contracts cover agreements on type, layout, binding locations, synchronization requirements, and related properties. In standard practice, these contracts are satisfied implicitly through conventions. As a result, contract breaches are often detected too late, at runtime by validation layers, or worse, as visual artifacts in rendered output. In this work, we present RCGP 1 , a system that mechanistically enforces resource contracts between components of a graphics program. RCGP formalizes contracts , which are centralized through single-source-of-truth resource declarations, as statically introspectable units that can be verified. We map objects from distinct program components to modules that import and export contracts. Modules are then composed into larger components via combinators , which check compatibility between the corresponding contracts and compute the residual contracts for the result. Finally, RCGP upholds these contracts through a set of witnesses , which encode the necessary promises. Using a prototype implementation in C++ and Vulkan, we demonstrate that this formulation converts common classes of late failures, such as descriptor-type mismatches, layout drift, and missing synchronization barriers, into early, actionable compile-time diagnostics.

We propose a method, HOTSPOT, for optimizing neural signed distance functions. Existing losses, such as the eikonal loss, act as necessary but insufficient constraints and cannot guarantee that the recovered implicit function represents a true distance function, even if the output minimizes these losses almost everywhere. Furthermore, the eikonal loss suffers from stability issues in optimization. Finally, in conventional methods, regularization losses that penalize surface area distort the reconstructed signed distance function. We address these challenges by designing a loss function using the solution of a screened Poisson equation. Our loss, when minimized, provides an asymptotically sufficient condition to ensure the output converges to a true distance function. Our loss also leads to stable optimization and naturally penalizes large surface areas. We present theoretical analysis and experiments on both challenging 2D and 3D datasets and show that our method provides better surface reconstruction and a more accurate distance approximation.

End-to-end optimization, which simultaneously optimizes optics and algorithms, has emerged as a powerful data-driven method for computational imaging system design. This method achieves joint optimization through backpropagation by incorporating differentiable optics simulators to generate measurements and algorithms to extract information from measurements. However, due to high computational costs, it is challenging to model both aberration and diffraction in light transport for end-to-end optimization of compound optics. Therefore, most existing methods compromise physical accuracy by neglecting wave optics effects or off-axis aberrations, which raises concerns about the robustness of the resulting designs. In this paper, we propose a differentiable optics simulator that efficiently models both aberration and diffraction for compound optics. Using the simulator, we conduct end-to-end optimization on scene reconstruction and classification. Experimental results demonstrate that both lenses and algorithms adopt different configurations depending on whether wave optics is modeled. We also show that systems optimized without wave optics suffer from performance degradation when wave optics effects are introduced during testing. These findings underscore the importance of accurate wave optics modeling in optimizing imaging systems for robust, high-performance applications.

Neural fields excel at representing continuous visual signals but typically operate at a single, fixed resolution. We present a simple yet powerful method to optimize neural fields that can be prefiltered in a single forward pass. Key innovations and features include: (1) We perform convolutional filtering in the input domain by analytically scaling Fourier feature embeddings with the filter’s frequency response. (2) This closed-form modulation generalizes beyond Gaussian filtering and supports other parametric filters (Box and Lanczos) that are unseen at training time. (3) We train the neural field using single-sample Monte Carlo estimates of the filtered signal. Our method is fast during both training and inference, and imposes no additional constraints on the network architecture. We show quantitative and qualitative improvements over existing methods for neural-field filtering.

We present a novel method to differentiate integrals of discontinuous functions, which are common in inverse graphics, computer vision, and machine learning applications. Previous methods either require specialized routines to sample the discontinuous boundaries of predetermined primitives, or use reparameterization techniques that suffer from high variance. In contrast, our method handles general discontinuous functions, expressed as shader programs, without requiring manually specified boundary sampling routines. We achieve this through a program transformation that converts discontinuous functions into piecewise constant ones, enabling efficient boundary sampling through a novel segment snapping technique, and accurate derivatives at the boundary by simply comparing values on both sides of the discontinuity. Our method handles both explicit boundaries (polygons, ellipses, Bézier curves) and implicit ones (neural networks, noise-based functions, swept surfaces). We demonstrate that our system supports a wide range of applications, including painterly rendering, raster image fitting, constructive solid geometry, swept surfaces, mosaicing, and ray marching.

In recent years, reconstruction methods have been developed that can recover strand-level hair geometry from images. However, these methods recover a vast number of individual hair strands that are difficult to edit and simulate. Many methods also rely on neural priors to infer non-visible inner hair, which can result in poor inner hair structure for complex hairstyles, such as curly hair. We propose an inverse hair grooming pipeline that transforms the imperfect 3D strands from these reconstruction methods into procedural hair grooms that consist of a small set of guide strands and hair grooming operators, inspired by pipelines used by artists in popular 3D modeling tools such as Blender and Houdini. We take a probabilistic view of these hair grooms and design various optimization strategies and loss functions to optimize for the guide strands and operator parameters. Due to the proceduralism, our resulting grooms can naturally represent challenging hairstyles, have structurally sound inner hair, and are easily editable.

Recent spatiotemporal resampling algorithms (ReSTIR) accelerate real-time path tracing by reusing samples between pixels and frames. However, existing methods are limited by the sampling quality of path tracing, making them inefficient for scenes with caustics and hard-to-reach lights. We develop a ReSTIR variant incorporating bidirectional path tracing that significantly improves the sampling quality in these scenes. Combining bidirectional path tracing and ReSTIR introduces multiple challenges: the generalized resampled importance sampling (GRIS) behind ReSTIR is, by default, not aware of how a path was sampled, which complicates reuse of bidirectional paths. Light tracing is also challenging since light subpaths can contribute to all pixels. To address these challenges, we apply GRIS in a sampling technique-aware extended path space, design a bidirectional hybrid shift mapping, and introduce caustics reservoirs that can accumulate caustics across frames. Our method takes around 50ms per frame across our test scenes, and achieves significantly lower error compared to prior unidirectional ReSTIR variants running in equal time.

Radiance fields have gained tremendous success with applications ranging from novel view synthesis to geometry reconstruction, especially with the advent of Gaussian splatting. However, they sacrifice modeling of material reflective properties and lighting conditions, leading to significant geometric ambiguities and the inability to easily perform relighting. One way to address these limitations is to incorporate physically-based rendering, but it has been prohibitively expensive to include full global illumination within the inner loop of the optimization. Therefore, previous works adopt simplifications that make the whole optimization with global illumination effects efficient but less accurate. In this work, we adopt Gaussian surfels as the primitives and build an efficient framework for differentiable light transport, inspired from the classic radiosity theory. The whole framework operates in the coefficient space of spherical harmonics, enabling both diffuse and specular materials. We extend the classic radiosity into non-binary visibility and semi-opaque primitives, propose novel solvers to efficiently solve the light transport, and derive the backward pass for gradient optimizations, which is more efficient than auto-differentiation. During inference, we achieve view-independent rendering where light transport need not be recomputed under viewpoint changes, enabling hundreds of FPS for global illumination effects, including view-dependent reflections using a spherical harmonics representation. Through extensive qualitative and quantitative experiments, we demonstrate superior geometry reconstruction, view synthesis and relighting than previous inverse rendering baselines, or data-driven baselines given relatively sparse datasets with known or unknown lighting conditions.

Integral linear operators play a key role in many graphics problems, but solutions obtained via Monte Carlo methods often suffer from high variance. A common strategy to improve the efficiency of integration across various inputs is to precompute the kernel function. Traditional methods typically rely on basis expansions for both the input and output functions. However, using fixed output bases can restrict the precision of output reconstruction and limit the compactness of the kernel representation. In this work, we introduce a new method that approximates both the kernel and the input function using Gaussian mixtures. This formulation allows the integral operator to be evaluated analytically, leading to improved flexibility in kernel storage and output representation. Moreover, our method naturally supports the sequential application of multiple operators and enables closed-form operator composition, which is particularly beneficial in tasks involving chains of operators. We demonstrate the versatility and effectiveness of our approach across a variety of graphics problems, including environment map relighting, boundary value problems, and fluorescence rendering.

Variance reduction techniques are widely used for reducing the noise of Monte Carlo integration. However, these techniques are typically designed with the assumption that the integrand is scalar-valued. Recognizing that rendering and inverse rendering broadly involve vector-valued integrands, we identify the limitations of classical variance reduction methods in this context. To address this, we introduce ratio control variates, an estimator that leverages a ratio-based approach instead of the conventional difference-based control variates. Our analysis and experiments demonstrate that ratio control variables can significantly reduce the mean squared error of vector-valued integration compared to existing methods and are broadly applicable to various rendering and inverse rendering tasks.

In offline reinforcement learning (RL), updating the value function with the discrete-time Bellman Equation often encounters challenges due to the limited scope of available data. This limitation stems from the Bellman Equation, which cannot accurately predict the value of unvisited states. To address this issue, we have introduced an innovative solution that bridges the continuous- and discrete-time RL methods, capitalizing on their advantages. Our method uses a discrete-time RL algorithm to derive the value function from a dataset while ensuring that the function’s first derivative aligns with the local characteristics of states and actions, as defined by the Hamilton-Jacobi-Bellman equation in continuous RL. We provide practical algorithms for both deterministic policy gradient methods and stochastic policy gradient methods. Experiments on the D4RL dataset show that incorporating the first-order information significantly improves policy performance for offline RL problems.

SA Conference Papers ’24, December 03–06, 2024, Tokyo, Japan

Abstract Conventional rendering techniques are primarily designed and optimized for single‐frame rendering. In practical applications, such as scene editing and animation rendering, users frequently encounter scenes where only a small portion is modified between consecutive frames. In this paper, we develop a novel approach to incremental re‐rendering of scenes with dynamic objects, where only a small part of a scene moves from one frame to the next. We formulate the difference (or residual) in the image between two frames as a (correlated) light‐transport integral which we call the residual path integral. Efficient numerical solution of this integral then involves (1) devising importance sampling strategies to focus on paths with non‐zero residual‐transport contributions and (2) choosing appropriate mappings between the native path spaces of the two frames. We introduce a set of path importance sampling strategies that trace from the moving object(s) which are the sources of residual energy. We explore path mapping strategies that generalize those from gradient‐domain path tracing to our importance sampling techniques specially for dynamic scenes. Additionally, our formulation can be applied to material editing as a simpler special case. We demonstrate speed‐ups over previous correlated sampling of path differences and over rendering the new frame independently. Our formulation brings new insights into the re‐rendering problem and paves the way for devising new types of sampling techniques and path mappings with different trade‐offs.

Computations in physical simulation, computer graphics, and probabilistic inference often require the differentiation of discontinuous processes due to contact, occlusion, and changes at a point in time. Popular differentiable programming languages, such as PyTorch and JAX, ignore discontinuities during differentiation. This is incorrect for parametric discontinuities —conditionals containing at least one real-valued parameter and at least one variable of integration. We introduce Potto, the first differentiable first-order programming language to soundly differentiate parametric discontinuities. We present a denotational semantics for programs and program derivatives and show the two accord. We describe the implementation of Potto, which enables separate compilation of programs. Our prototype implementation overcomes previous compile-time bottlenecks achieving an 88.1x and 441.2x speed up in compile time and a 2.5x and 7.9x speed up in runtime, respectively, on two increasingly large image stylization benchmarks. We showcase Potto by implementing a prototype differentiable renderer with separately compiled shaders.

We propose a set of techniques to efficiently importance sample the derivatives of a wide range of Bidirectional Reflectance Distribution Function (BRDF) models. In differentiable rendering, BRDFs are replaced by their differential BRDF counterparts, which are real-valued and can have negative values. This leads to a new source of variance arising from their change in sign. Real-valued functions cannot be perfectly importance sampled by a positive-valued PDF, and the direct application of BRDF sampling leads to high variance. Previous attempts at antithetic sampling only addressed the derivative with the roughness parameter of isotropic microfacet BRDFs. Our work generalizes BRDF derivative sampling to anisotropic microfacet models, mixture BRDFs, Oren-Nayar, Hanrahan-Krueger, among other analytic BRDFs. Our method first decomposes the real-valued differential BRDF into a sum of single-signed functions, eliminating variance from a change in sign. Next, we importance sample each of the resulting single-signed functions separately. The first decomposition, positivization, partitions the real-valued function based on its sign, and is effective at variance reduction when applicable. However, it requires analytic knowledge of the roots of the differential BRDF, and for it to be analytically integrable too. Our key insight is that the single-signed functions can have overlapping support, which significantly broadens the ways we can decompose a real-valued function. Our product and mixture decompositions exploit this property, and they allow us to support several BRDF derivatives that positivization could not handle. For a wide variety of BRDF derivatives, our method significantly reduces the variance (up to 58× in some cases) at equal computation cost and enables better recovery of spatially varying textures through gradient-descent-based inverse rendering.

Metropolis Light Transport (MLT) is a global illumination algorithm that is well-known for rendering challenging scenes with intricate light paths. However, MLT methods tend to produce unpredictable correlation artifacts in images, which can introduce visual inconsistencies for animation rendering. This drawback also makes it challenging to denoise MLT renderings while maintaining temporal stability. We tackle this issue with modern learning-based methods and build a sequence denoiser combining the recurrent connections with the cutting-edge vision transformer architecture. We demonstrate that our sophisticated denoiser can consistently improve the quality and temporal stability of MLT renderings with difficult light paths. Our method is efficient and scalable for complex scene renderings that require high sample counts.

We propose a real-time path guiding method, Voxel Path Guiding (VXPG), that significantly improves fitting efficiency under limited sampling budget. Our key idea is to use a spatial irradiance voxel data structure across all shading points to guide the location of path vertices. For each frame, we first populate the voxel data structure with irradiance and geometry information. To sample from the data structure for a shading point, we need to select a voxel with high contribution to that point. To importance sample the voxels while taking visibility into consideration, we adapt techniques from offline many-lights rendering by clustering pairs of shading points and voxels. Finally, we unbiasedly sample within the selected voxel while taking the geometry inside into consideration. Our experiments show that VXPG achieves significantly lower perceptual error compared to other real-time path guiding and virtual point light methods under equal-time comparison. Furthermore, our method does not rely on temporal information, but can be used together with other temporal reuse sampling techniques such as ReSTIR to further improve sampling efficiency.

Inverse path tracing has recently been applied to joint material and lighting estimation, given geometry and multi-view HDR observations of an indoor scene. However, it has two major limitations: path tracing is expensive to compute, and ambiguities exist between reflection and emission. Our Factorized Inverse Path Tracing (FIPT) addresses these challenges by using a factored light transport formulation and finds emitters driven by rendering errors. Our algorithm enables accurate material and lighting optimization faster than previous work, and is more effective at resolving ambiguities. The exhaustive experiments on synthetic scenes show that our method (1) outperforms state-of-the-art indoor inverse rendering and relighting methods particularly in the presence of complex illumination effects; (2) speeds up inverse path tracing optimization to less than an hour. We further demonstrate robustness to noisy inputs through material and lighting estimates that allow plausible relighting in a real scene. The source code is available at: https://github.com/lwwu2/fipt

A single panel of a comic book can say a lot: it can depict not only where the characters currently are, but also their motions, their motivations, their emotions, and what they might do next. More generally, humans routinely infer complex sequences of past and future events from a *static snapshot* of a *dynamic scene*, even in situations they have never seen before. In this paper, we model how humans make such rapid and flexible inferences. Building on a long line of work in cognitive science, we offer a Monte Carlo algorithm whose inferences correlate well with human intuitions in a wide variety of domains, while only using a small, cognitively-plausible number of samples. Our key technical insight is a surprising connection between our inference problem and Monte Carlo path tracing, which allows us to apply decades of ideas from the computer graphics community to this seemingly-unrelated theory of mind task.

Abstract Precomputed Radiance Transfer (PRT) remains an attractive solution for real‐time rendering of complex light transport effects such as glossy global illumination. After precomputation, we can relight the scene with new environment maps while changing viewpoint in real‐time. However, practical PRT methods are usually limited to low‐frequency spherical harmonic lighting. All‐frequency techniques using wavelets are promising but have so far had little practical impact. The curse of dimensionality and much higher data requirements have typically limited them to relighting with fixed view or only direct lighting with triple product integrals. In this paper, we demonstrate a hybrid neural‐wavelet PRT solution to high‐frequency indirect illumination, including glossy reflection, for relighting with changing view. Specifically, we seek to represent the light transport function in the Haar wavelet basis. For global illumination, we learn the wavelet transport using a small multi‐layer perceptron (MLP) applied to a feature field as a function of spatial location and wavelet index, with reflected direction and material parameters being other MLP inputs. We optimize/learn the feature field (compactly represented by a tensor decomposition) and MLP parameters from multiple images of the scene under different lighting and viewing conditions. We demonstrate real‐time (512 x 512 at 24 FPS, 800 x 600 at 13 FPS) precomputed rendering of challenging scenes involving view‐dependent reflections and even caustics.

We introduce SLANG.D, an extension to the Slang shading language that incorporates first-class automatic differentiation support. The new shading language allows us to transform a Direct3D-based path tracer to be fully differentiable with minor modifications to existing code. SLANG.D enables a shared ecosystem between machine learning frameworks and pre-existing graphics hardware API-based rendering systems, promoting the interchange of components and ideas across these two domains. Our contributions include a differentiable type system designed to ensure type safety and semantic clarity in codebases that blend differentiable and non-differentiable code, language primitives that automatically generate both forward and reverse gradient propagation methods, and a compiler architecture that generates efficient derivative propagation shader code for graphics pipelines. Our compiler supports differentiating code that involves arbitrary control-flow, dynamic dispatch, generics and higher-order differentiation, while providing developers flexible control of checkpointing and gradient aggregation strategies for best performance. Our system allows us to differentiate an existing real-time path tracer, Falcor, with minimal change to its shader code. We show that the compiler-generated derivative kernels perform as efficiently as handwritten ones. In several benchmarks, the SLANG.D code achieves significant speedup when compared to prior automatic differentiation systems.

Neural image representations offer the possibility of high fidelity, compact storage, and resolution-independent accuracy, providing an attractive alternative to traditional pixel- and grid-based representations. However, coordinate neural networks fail to capture discontinuities present in the image and tend to blur across them; we aim to address this challenge. In many cases, such as rendered images, vector graphics, diffusion curves, or solutions to partial differential equations, the locations of the discontinuities are known. We take those locations as input, represented as linear, quadratic, or cubic Bézier curves, and construct a feature field that is discontinuous across these locations and smooth everywhere else. Finally, we use a shallow multi-layer perceptron to decode the features into the signal value. To construct the feature field, we develop a new data structure based on a curved triangular mesh, with features stored on the vertices and on a subset of the edges that are marked as discontinuous. We show that our method can be used to compress a 100, 000 2 -pixel rendered image into a 25MB file; can be used as a new diffusion-curve solver by combining with Monte-Carlo-based methods or directly supervised by the diffusion-curve energy; or can be used for compressing 2D physics simulation data.

Physics-based differentiable rendering is becoming increasingly crucial for tasks in inverse rendering and machine learning pipelines. To address discontinuities caused by geometric boundaries and occlusion, two classes of methods have been proposed: 1) the edge-sampling methods that directly sample light paths at the scene discontinuity boundaries, which require nontrivial data structures and precomputation to select the edges, and 2) the reparameterization methods that avoid discontinuity sampling but are currently limited to hemispherical integrals and unidirectional path tracing. We introduce a new mathematical formulation that enjoys the benefits of both classes of methods. Unlike previous reparameterization work that focused on hemispherical integral, we derive the reparameterization in the path space. As a result, to estimate derivatives using our formulation, we can apply advanced Monte Carlo rendering methods, such as bidirectional path tracing, while avoiding explicit sampling of discontinuity boundaries. We show differentiable rendering and inverse rendering results to demonstrate the effectiveness of our method.

We present a method to automatically compute correct gradients with respect to geometric scene parameters in neural SDF renderers. Recent physically-based differentiable rendering techniques for meshes have used edge-sampling to handle discontinuities, particularly at object silhouettes, but SDFs do not have a simple parametric form amenable to sampling. Instead, our approach builds on area-sampling techniques and develops a continuous warping function for SDFs to account for these discontinuities. Our method leverages the distance to surface encoded in an SDF and uses quadrature on sphere tracer points to compute this warping function. We further show that this can be done by subsampling the points to make the method tractable for neural SDFs. Our differentiable renderer can be used to optimize neural shapes from multi-view images and produces comparable 3D reconstructions to recent SDF-based inverse rendering methods, without the need for 2D segmentation masks to guide the geometry optimization and no volumetric approximations to the geometry.

We present a method to automatically synthesize efficient, high-quality demosaicking algorithms, across a range of computational budgets, given a loss function and training data. It performs a multi-objective, discrete-continuous optimization which simultaneously solves for the program structure and parameters that best tradeoff computational cost and image quality. We design the method to exploit domain-specific structure for search efficiency. We apply it to several tasks, including demosaicking both Bayer and Fuji X-Trans color filter patterns, as well as joint demosaicking and super-resolution. In a few days on 8 GPUs, it produces a family of algorithms that significantly improves image quality relative to the prior state-of-the-art across a range of computational budgets from 10 s to 1000 s of operations per pixel (1 dB–3 dB higher quality at the same cost, or 8.5–200× higher throughput at same or better quality). The resulting programs combine features of both classical and deep learning-based demosaicking algorithms into more efficient hybrid combinations, which are bandwidth-efficient and vectorizable by construction. Finally, our method automatically schedules and compiles all generated programs into optimized SIMD code for modern processors.

We present a new algorithm to quickly generate high-performance GPU implementations of complex imaging and vision pipelines, directly from high-level Halide algorithm code. It is fully automatic, requiring no schedule templates or hand-optimized kernels. We address the scalability challenge of extending search-based automatic scheduling to map large real-world programs to the deep hierarchies of memory and parallelism on GPU architectures in reasonable compile time. We achieve this using (1) a two-phase search algorithm that first ‘freezes’ decisions for the lowest cost sections of a program, allowing relatively more time to be spent on the important stages, (2) a hierarchical sampling strategy that groups schedules based on their structural similarity, then samples representatives to be evaluated, allowing us to explore a large space with few samples, and (3) memoization of repeated partial schedules, amortizing their cost over all their occurrences. We guide the process with an efficient cost model combining machine learning, program analysis, and GPU architecture knowledge. We evaluate our method’s performance on a diverse suite of real-world imaging and vision pipelines. Our scalability optimizations lead to average compile time speedups of 49x (up to 530x). We find schedules that are on average 1.7x faster than existing automatic solutions (up to 5x), and competitive with what the best human experts were able to achieve in an active effort to beat our automatic results.

Emerging research in computer graphics, inverse problems, and machine learning requires us to differentiate and optimize parametric discontinuities. These discontinuities appear in object boundaries, occlusion, contact, and sudden change over time. In many domains, such as rendering and physics simulation, we differentiate the parameters of models that are expressed as integrals over discontinuous functions. Ignoring the discontinuities during differentiation often has a significant impact on the optimization process. Previous approaches either apply specialized hand-derived solutions, smooth out the discontinuities, or rely on incorrect automatic differentiation. We propose a systematic approach to differentiating integrals with discontinuous integrands, by developing a new differentiable programming language. We introduce integration as a language primitive and account for the Dirac delta contribution from differentiating parametric discontinuities in the integrand. We formally define the language semantics and prove the correctness and closure under the differentiation, allowing the generation of gradients and higher-order derivatives. We also build a system, Teg, implementing these semantics. Our approach is widely applicable to a variety of tasks, including image stylization, fitting shader parameters, trajectory optimization, and optimizing physical designs.

We present an unbiased online Monte Carlo method for rendering with many lights. Our method adapts both the hierarchical light clustering and the sampling distribution to our collected samples. Designing such a method requires us to make clustering decisions under noisy observation, and making sure that the sampling distribution adapts to our target. Our method is based on two key ideas: a coarse-to-fine clustering scheme that can find good clustering configurations even with noisy samples, and a discrete stochastic successive approximation method that starts from a prior distribution and provably converges to a target distribution. We compare to other state-of-the-art light sampling methods, and show better results both numerically and visually.

Differentiable rendering computes derivatives of the light transport equation with respect to arbitrary 3D scene parameters, and enables various applications in inverse rendering and machine learning. We present an unbiased and efficient differentiable rendering algorithm that does not require explicit boundary sampling. We apply the divergence theorem to the derivative of the rendering integral to convert the boundary integral into an area integral. We rewrite the converted area integral to a form that is suitable for Monte Carlo rendering. We then develop an efficient Monte Carlo sampling algorithm for solving the area integral. Our method can be easily plugged into a traditional path tracer and does not require dedicated data structures for sampling boundaries. We analyze the convergence properties through bias-variance metrics, and demonstrate our estimator's advantages over existing methods for some synthetic inverse rendering examples.

We introduce a differentiable rasterizer that bridges the vector graphics and raster image domains, enabling powerful raster-based loss functions, optimization procedures, and machine learning techniques to edit and generate vector content. We observe that vector graphics rasterization is differentiable after pixel prefiltering. Our differentiable rasterizer offers two prefiltering options: an analytical prefiltering technique and a multisampling anti-aliasing technique. The analytical variant is faster but can suffer from artifacts such as conflation. The multisampling variant is still efficient, and can render high-quality images while computing unbiased gradients for each pixel with respect to curve parameters. We demonstrate that our rasterizer enables new applications, including a vector graphics editor guided by image metrics, a painterly rendering algorithm that fits vector primitives to an image by minimizing a deep perceptual loss function, new vector graphics editing algorithms that exploit well-known image processing methods such as seam carving, and deep generative models that generate vector content from raster-only supervision under a VAE or GAN training objective.

Modern computer vision algorithms have brought significant advancement to 3D geometry reconstruction. However, illumination and material reconstruction remain less studied, with current approaches assuming very simplified models for materials and illumination. We introduce Inverse Path Tracing, a novel approach to jointly estimate the material properties of objects and light sources in indoor scenes by using an invertible light transport simulation. We assume a coarse geometry scan, along with corresponding images and camera poses. The key contribution of this work is an accurate and simultaneous retrieval of light sources and physically based material properties (e.g., diffuse reflectance, specular reflectance, roughness, etc.) for the purpose of editing and re-rendering the scene under new conditions. To this end, we introduce a novel optimization method using a differentiable Monte Carlo renderer that computes derivatives with respect to the estimated unknown illumination and material properties. This enables joint optimization for physically correct light transport and material models using a tailored stochastic gradient descent.

We present a new algorithm to automatically schedule Halide programs for high-performance image processing and deep learning. We significantly improve upon the performance of previous methods, which considered a limited subset of schedules. We define a parameterization of possible schedules much larger than prior methods and use a variant of beam search to search over it. The search optimizes runtime predicted by a cost model based on a combination of new derived features and machine learning. We train the cost model by generating and featurizing hundreds of thousands of random programs and schedules. We show that this approach operates effectively with or without autotuning. It produces schedules which are on average almost twice as fast as the existing Halide autoscheduler without autotuning, or more than twice as fast with, and is the first automatic scheduling algorithm to significantly outperform human experts on average.

Denoising has proven to be useful to efficiently generate high-quality Monte Carlo renderings. Traditional pixel-based denoisers exploit summary statistics of a pixel's sample distributions, which discards much of the samples' information and limits their denoising power. On the other hand, sample-based techniques tend to be slow and have difficulties handling general transport scenarios. We present the first convolutional network that can learn to denoise Monte Carlo renderings directly from the samples. Learning the mapping between samples and images creates new challenges for the network architecture design: the order of the samples is arbitrary, and they should be treated in a permutation invariant manner. To address these challenges, we develop a novel kernel-predicting architecture that splats individual samples onto nearby pixels. Splatting is a natural solution to situations such as motion blur, depth-of-field and many light transport paths, where it is easier to predict which pixels a sample contributes to, rather than a gather approach that needs to figure out, for each pixel, which samples (or nearby pixels) are relevant. Compared to previous state-of-the-art methods, ours is robust to the severe noise of low-sample count images (e.g. 8 samples per pixel) and yields higher-quality results both visually and numerically. Our approach retains the generality and efficiency of pixel-space methods while enjoying the expressiveness and accuracy of the more complex sample-based approaches.

3D visual computing data are often spatially sparse. To exploit such sparsity, people have developed hierarchical sparse data structures, such as multi-level sparse voxel grids, particles, and 3D hash tables. However, developing and using these high-performance sparse data structures is challenging, due to their intrinsic complexity and overhead. We propose Taichi , a new data-oriented programming language for efficiently authoring, accessing, and maintaining such data structures. The language offers a high-level, data structure-agnostic interface for writing computation code. The user independently specifies the data structure. We provide several elementary components with different sparsity properties that can be arbitrarily composed to create a wide range of multi-level sparse data structures. This decoupling of data structures from computation makes it easy to experiment with different data structures without changing computation code, and allows users to write computation as if they are working with a dense array. Our compiler then uses the semantics of the data structure and index analysis to automatically optimize for locality, remove redundant operations for coherent accesses, maintain sparsity and memory allocations, and generate efficient parallel and vectorized instructions for CPUs and GPUs. Our approach yields competitive performance on common computational kernels such as stencil applications, neighbor lookups, and particle scattering. We demonstrate our language by implementing simulation, rendering, and vision tasks including a material point method simulation, finite element analysis, a multigrid Poisson solver for pressure projection, volumetric path tracing, and 3D convolution on sparse grids. Our computation-data structure decoupling allows us to quickly experiment with different data arrangements, and to develop high-performance data structures tailored for specific computational tasks. With 1 1 0 th as many lines of code, we achieve 4.55× higher performance on average, compared to hand-optimized reference implementations.

Gradient-based methods are becoming increasingly important for computer graphics, machine learning, and computer vision. The ability to compute gradients is crucial to optimization, inverse problems, and deep learning. In rendering, the gradient is required with respect to variables such as camera parameters, light sources, scene geometry, or material appearance. However, computing the gradient of rendering is challenging because the rendering integral includes visibility terms that are not differentiable. Previous work on differentiable rendering has focused on approximate solutions. They often do not handle secondary effects such as shadows or global illumination, or they do not provide the gradient with respect to variables other than pixel coordinates. We introduce a general-purpose differentiable ray tracer, which, to our knowledge, is the first comprehensive solution that is able to compute derivatives of scalar functions over a rendered image with respect to arbitrary scene parameters such as camera pose, scene geometry, materials, and lighting parameters. The key to our method is a novel edge sampling algorithm that directly samples the Dirac delta functions introduced by the derivatives of the discontinuous integrand. We also develop efficient importance sampling methods based on spatial hierarchies. Our method can generate gradients in times running from seconds to minutes depending on scene complexity and desired precision. We interface our differentiable ray tracer with the deep learning library PyTorch and show prototype applications in inverse rendering and the generation of adversarial examples for neural networks.

Gradient-based optimization has enabled dramatic advances in computational imaging through techniques like deep learning and nonlinear optimization. These methods require gradients not just of simple mathematical functions, but of general programs which encode complex transformations of images and graphical data. Unfortunately, practitioners have traditionally been limited to either hand-deriving gradients of complex computations, or composing programs from a limited set of coarse-grained operators in deep learning frameworks. At the same time, writing programs with the level of performance needed for imaging and deep learning is prohibitively difficult for most programmers. We extend the image processing language Halide with general reverse-mode automatic differentiation (AD), and the ability to automatically optimize the implementation of gradient computations. This enables automatic computation of the gradients of arbitrary Halide programs, at high performance, with little programmer effort. A key challenge is to structure the gradient code to retain parallelism. We define a simple algorithm to automatically schedule these pipelines, and show how Halide's existing scheduling primitives can express and extend the key AD optimization of "checkpointing." Using this new tool, we show how to easily define new neural network layers which automatically compile to high-performance GPU implementations, and how to solve nonlinear inverse problems from computational imaging. Finally, we show how differentiable programming enables dramatically improving the quality of even traditional, feed-forward image processing algorithms, blurring the distinction between classical and deep methods.

Implementing Monte Carlo integration requires significant domain expertise. While simple samplers, such as unidirectional path tracing, are relatively forgiving, more complex algorithms, such as bidirectional path tracing or Metropolis methods, are notoriously difficult to implement correctly. We propose Aether, an embedded domain specific language for Monte Carlo integration, which offers primitives for writing concise and correct-by-construction sampling and probability code. The user is tasked with writing sampling code, while our compiler automatically generates the code necessary for evaluating PDFs as well as the book keeping and combination of multiple sampling strategies. Our language focuses on ease of implementation for rapid exploration, at the cost of run time performance. We demonstrate the effectiveness of the language by implementing several challenging rendering algorithms as well as a new algorithm, which would otherwise be prohibitively difficult.

The simulation of light transport in the presence of multi-bounce glossy effects and motion is challenging because the integrand is high dimensional and areas of high-contribution tend to be narrow and hard to sample. We present a Markov Chain Monte Carlo (MCMC) rendering algorithm that extends Metropolis Light Transport by automatically and explicitly adapting to the local shape of the integrand, thereby increasing the acceptance rate. Our algorithm characterizes the local behavior of throughput in path space using its gradient as well as its Hessian. In particular, the Hessian is able to capture the strong anisotropy of the integrand. We obtain the derivatives using automatic differentiation, which makes our solution general and easy to extend to additional sampling dimensions such as time. However, the resulting second order Taylor expansion is not a proper distribution and cannot be used directly for importance sampling. Instead, we use ideas from Hamiltonian Monte-Carlo and simulate the Hamiltonian dynamics in a flipped version of the Taylor expansion where gravity pulls particles towards the high-contribution region. Whereas such methods usually require numerical integration, we show that our quadratic landscape leads to a closed-form anisotropic Gaussian distribution for the final particle positions, and it results in a standard Metropolis-Hastings algorithm. Our method excels at rendering glossy-to-glossy reflections on small and highly curved surfaces. Furthermore, unlike previous work that derives sampling anisotropy with pen and paper and only considers specific effects such as specular BSDFs, we characterize the local shape of throughput through automatic differentiation. This makes our approach very general. In particular, our method is the first MCMC rendering algorithm that is able to resolve the anisotropy in the time dimension and render difficult moving caustics.

This paper introduces a scalable algorithm for rendering translucent materials with complex lighting. We represent the light transport with a diffusion approximation by a dual-matrix representation with the Light-to-Surface and Surface-to-Camera matrices. By exploiting the structures within the matrices, the proposed method can locate surface samples with little contribution by using only subsampled matrices and avoid wasting computation on these samples. The decoupled estimation of irradiance and diffuse BSSRDFs also allows us to have a tight error bound, making the adaptive diffusion approximation more efficient and accurate. Experiments show that our method outperforms previous methods for translucent material rendering, especially in large scenes with massive translucent surfaces shaded by complex illumination.

We apply Stein's Unbiased Risk Estimator (SURE) to adaptive sampling and reconstruction to reduce noise in Monte Carlo rendering. SURE is a general unbiased estimator for mean squared error (MSE) in statistics. With SURE, we are able to estimate error for an arbitrary reconstruction kernel, enabling us to use more effective kernels rather than being restricted to the symmetric ones used in previous work. It also allows us to allocate more samples to areas with higher estimated MSE. Adaptive sampling and reconstruction can therefore be processed within an optimization framework. We also propose an efficient and memory-friendly approach to reduce the impact of noisy geometry features where there is depth of field or motion blur. Experiments show that our method produces images with less noise and crisper details than previous methods.