SIGGRAPH 2023 论文列表

We present a one-shot method to infer and render a photorealistic 3D representation from a single unposed image (e.g., face portrait) in real-time. Given a single RGB input, our image encoder directly predicts a canonical triplane representation of a neural radiance field for 3D-aware novel view synthesis via volume rendering. Our method is fast (24 fps) on consumer hardware, and produces higher quality results than strong GAN-inversion baselines that require test-time optimization. To train our triplane encoder pipeline, we use only synthetic data, showing how to distill the knowledge from a pretrained 3D GAN into a feedforward encoder. Technical contributions include a Vision Transformer-based triplane encoder, a camera data augmentation strategy, and a well-designed loss function for synthetic data training. We benchmark against the state-of-the-art methods, demonstrating significant improvements in robustness and image quality in challenging real-world settings. We showcase our results on portraits of faces (FFHQ) and cats (AFHQ), but our algorithm can also be applied in the future to other categories with a 3D-aware image generator.

The continuous advancement of photorealism in rendering is accompanied by a growth in texture data and, consequently, increasing storage and memory demands. To address this issue, we propose a novel neural compression technique specifically designed for material textures. We unlock two more levels of detail, i.e., 16× more texels, using low bitrate compression, with image quality that is better than advanced image compression techniques, such as AVIF and JPEG XL. At the same time, our method allows on-demand, real-time decompression with random access similar to block texture compression on GPUs, enabling compression on disk and memory. The key idea behind our approach is compressing multiple material textures and their mipmap chains together, and using a small neural network, that is optimized for each material, to decompress them. Finally, we use a custom training implementation to achieve practical compression speeds, whose performance surpasses that of general frameworks, like PyTorch, by an order of magnitude.

Styled online in-between motion generation has important application scenarios in computer animation and games. Its core challenge lies in the need to satisfy four critical requirements simultaneously: generation speed, motion quality, style diversity, and synthesis controllability. While the first two challenges demand a delicate balance between simple fast models and learning capacity for generation quality, the latter two are rarely investigated together in existing methods, which largely focus on either control without style or uncontrolled stylized motions. To this end, we propose a Real-time Stylized Motion Transition method (RSMT) to achieve all aforementioned goals. Our method consists of two critical, independent components: a general motion manifold model and a style motion sampler. The former acts as a high-quality motion source and the latter synthesizes styled motions on the fly under control signals. Since both components can be trained separately on different datasets, our method provides great flexibility, requires less data, and generalizes well when no/few samples are available for unseen styles. Through exhaustive evaluation, our method proves to be fast, high-quality, versatile, and controllable. The code and data are available at https://github.com/yuyujunjun/RSMT-Realtime-Stylized-Motion-Transition.

We present a simple but effective technique to smooth out textures while preserving the prominent structures. Our method is built upon a key observation---the coarsest level in a Gaussian pyramid often naturally eliminates textures and summarizes the main image structures. This inspires our central idea for texture filtering, which is to progressively upsample the very low-resolution coarsest Gaussian pyramid level to a full-resolution texture smoothing result with well-preserved structures, under the guidance of each fine-scale Gaussian pyramid level and its associated Laplacian pyramid level. We show that our approach is effective to separate structure from texture of different scales, local contrasts, and forms, without degrading structures or introducing visual artifacts. We also demonstrate the applicability of our method on various applications including detail enhancement, image abstraction, HDR tone mapping, inverse halftoning, and LDR image enhancement. Code is available at https://rewindl.github.io/pyramid_texture_filtering/.

Null-collision approaches for estimating transmittance and sampling free-flight distances are the current state-of-the-art for unbiased rendering of general heterogeneous participating media. However, null-collision approaches have a strict requirement for specifying a tightly bounding total extinction in order to remain both robust and performant; in practice this requirement restricts the use of null-collision techniques to only participating media where the density of the medium at every possible point in space is known a-priori. In production rendering, a common case is a medium in which density is defined by a black-box procedural function for which a bounding extinction cannot be determined beforehand. Typically in this case, a bounding extinction must be approximated by using an overly loose and therefore computationally inefficient conservative estimate. We present an analysis of how null-collision techniques degrade when a more aggressive initial guess for a bounding extinction underestimates the true maximum density and turns out to be non-bounding. We then build upon this analysis to arrive at two new techniques: first, a practical, efficient, consistent progressive algorithm that allows us to robustly adapt null-collision techniques for use with procedural media with unknown bounding extinctions, and second, a new importance sampling technique that improves ratio-tracking based on zero-variance sampling.

We introduce a compact, intuitive procedural graph representation for cellular metamaterials, which are small-scale, tileable structures that can be architected to exhibit many useful material properties. Because the structures’ “architectures” vary widely—with elements such as beams, thin shells, and solid bulks—it is difficult to explore them using existing representations. Generic approaches like voxel grids are versatile, but it is cumbersome to represent and edit individual structures; architecture-specific approaches address these issues, but are incompatible with one another. By contrast, our procedural graph succinctly represents the construction process for any structure using a simple skeleton annotated with spatially varying thickness. To express the highly constrained triply periodic minimal surfaces (TPMS) in this manner, we present the first fully automated version of the conjugate surface construction method, which allows novices to create complex TPMS from intuitive input. We demonstrate our representation’s expressiveness, accuracy, and compactness by constructing a wide range of established structures and hundreds of novel structures with diverse architectures and material properties. We also conduct a user study to verify our representation’s ease-of-use and ability to expand engineers’ capacity for exploration.

This paper presents the model and rendering algorithm of Potentially Visible Hidden Volumes (PVHVs) for multi-view image warping. PVHVs are 3D volumes that are occluded at a known source view, but potentially visible at novel views. Given a bound of novel views, we define PVHVs using the edges of foreground fragments from the known view and the bound of novel views. PVHVs can be used to batch-test the visibilities of source fragments without iterating individual novel views in multi-fragment rendering, and thereby, cull redundant fragments prior to warping. We realize the model of PVHVs in Depth Peeling (DP). Our Effective Depth Peeling (EDP) can reduce the number of completely hidden fragments, capture important fragments early, and reduce warping cost. We demonstrate the benefit of our PVHVs and EDP in terms of memory, quality, and performance in multi-view warping.

Creating pose-driven human avatars is about modeling the mapping from the low-frequency driving pose to high-frequency dynamic human appearances, so an effective pose encoding method that can encode high-fidelity human details is essential to human avatar modeling. To this end, we present PoseVocab, a novel pose encoding method that encourages the network to discover the optimal pose embeddings for learning the dynamic human appearance. Given multi-view RGB videos of a character, PoseVocab constructs key poses and latent embeddings based on the training poses. To achieve pose generalization and temporal consistency, we sample key rotations in so(3) of each joint rather than the global pose vectors, and assign a pose embedding to each sampled key rotation. These joint-structured pose embeddings not only encode the dynamic appearances under different key poses, but also factorize the global pose embedding into joint-structured ones to better learn the appearance variation related to the motion of each joint. To improve the representation ability of the pose embedding while maintaining memory efficiency, we introduce feature lines, a compact yet effective 3D representation, to model more fine-grained details of human appearances. Furthermore, given a query pose and a spatial position, a hierarchical query strategy is introduced to interpolate pose embeddings and acquire the conditional pose feature for dynamic human synthesis. Overall, PoseVocab effectively encodes the dynamic details of human appearance and enables realistic and generalized animation under novel poses. Experiments show that our method outperforms other state-of-the-art baselines both qualitatively and quantitatively in terms of synthesis quality. Code is available at https://github.com/lizhe00/PoseVocab.

Cage coordinates are a powerful means to define 2D deformation fields from sparse control points. We introduce Conformal polynomial Coordinates for closed polyhedral cages, enabling segments to be transformed into polynomial curves of any order. Extending classical 2D Green coordinates, our coordinates result in conformal harmonic deformations that are cage-aware. We demonstrate the usefulness of our technique on a variety of 2D deformation scenarios where curves allow artists to perform intuitive deformations with few input parameters. Our method combines the texture preservation property of conformal deformations together with the expressiveness offered by Bezier controls.

Standard liquid simulators apply operator splitting to independently solve for pressure and viscous stresses, a decoupling that induces incorrect free surface boundary conditions. Such methods are unable to simulate fluid phenomena reliant on the balance of pressure and viscous stresses, such as the liquid rope coil instability exhibited by honey. By contrast, unsteady Stokes solvers retain coupling between pressure and viscosity, thus resolving these phenomena, albeit using a much larger and thus more computationally expensive linear system compared to the decoupled approach. To accelerate solving the unsteady Stokes problem, we propose a reduced fluid model wherein interior regions are represented with incompressible polynomial vector fields. Sets of standard grid cells are consolidated into super-cells, each of which are modelled using a quadratic field of 26 degrees of freedom. We demonstrate that the reduced field must necessarily be at least quadratic, with the affine model being unable to correctly capture viscous forces. We reproduce the liquid rope coiling instability, as well as other simulated examples, to show that our reduced model is able to reproduce the same fluid phenomena at a smaller computational cost. Futhermore, we performed a crowdsourced user survey to verify that our method produces imperceptible differences compared to the full unsteady Stokes method.

Authoring high-quality digital materials is key to realism in 3D rendering. Previous generative models for materials have been trained exclusively on synthetic data; such data is limited in availability and has a visual gap to real materials. We circumvent this limitation by proposing PhotoMat: the first material generator trained exclusively on real photos of material samples captured using a cell phone camera with flash. Supervision on individual material maps is not available in this setting. Instead, we train a generator for a neural material representation that is rendered with a learned relighting module to create arbitrarily lit RGB images; these are compared against real photos using a discriminator. We train PhotoMat with a new dataset of 12,000 material photos captured with handheld phone cameras under flash lighting. We demonstrate that our generated materials have better visual quality than previous material generators trained on synthetic data. Moreover, we can fit analytical material models to closely match these generated neural materials, thus allowing for further editing and use in 3D rendering.

Virtual reality (VR) passthrough uses external cameras on the front of a headset to allow the user to see their environment. However, passthrough cameras cannot physically be co-located with the user’s eyes, so the passthrough images have a different perspective than what the user would see without the headset. Although the images can be computationally reprojected into the desired view, errors in depth estimation, view-dependent effects, and missing information at occlusion boundaries can lead to undesirable artifacts. We propose a novel computational camera that directly samples the rays that would have gone into the user’s eye, several centimeters behind the sensor. Our design contains an array of lenses with an aperture behind each lens, and the apertures are strategically placed to allow through only the desired rays. The resulting thin, flat architecture has suitable form factor for VR, and the image reconstruction is computationally lightweight, enabling low-latency passthrough. We demonstrate our approach experimentally in a fully functional binocular passthrough prototype with practical calibration and real-time image reconstruction. Finally, we experimentally validate that our camera captures the correct perspective for VR passthrough, even in the presence of transparent objects, specular highlights, and complex occluding structures.

Modeling perception is critical for many applications and developments in computer graphics to optimize and evaluate content generation techniques. Most of the work to date has focused on central (foveal) vision. However, this is insufficient for novel wide-field-of-view display devices, such as virtual and augmented reality headsets. Furthermore, the perceptual models proposed for the fovea do not readily extend to the off-center, peripheral visual field, where human perception is drastically different. In this article, we focus on modeling the temporal aspect of visual perception in the periphery. We present new psychophysical experiments that measure the sensitivity of human observers to different spatio-temporal stimuli across a wide field of view. We use the collected data to build a perceptual model for the visibility of temporal changes at different eccentricities in complex video content. Finally, we discuss, demonstrate, and evaluate several problems that can be addressed using our technique. First, we show how our model enables injecting new content into the periphery without distracting the viewer, and we discuss the link between the model and human attention. Second, we demonstrate how foveated rendering methods can be evaluated and optimized to limit the visibility of temporal aliasing.

Point patterns are characterized by their density and correlation. While spatial variation of density is well-understood, analysis and synthesis of spatially-varying correlation is an open challenge. No tools are available to intuitively edit such point patterns, primarily due to the lack of a compact representation for spatially varying correlation. We propose a low-dimensional perceptual embedding for point correlations. This embedding can map point patterns to common three-channel raster images, enabling manipulation with off-the-shelf image editing software. To synthesize back point patterns, we propose a novel edge-aware objective that carefully handles sharp variations in density and correlation. The resulting framework allows intuitive and backward-compatible manipulation of point patterns, such as recoloring, relighting to even texture synthesis that have not been available to 2D point pattern design before. Effectiveness of our approach is tested in several user experiments. Code is available at https://github.com/xchhuang/patternshop.

Anime is an abstract art form that is substantially different from the human portrait, leading to a challenging misaligned image translation problem that is beyond the capability of existing methods. This can be boiled down to a highly ambiguous unconstrained translation between two domains. To this end, we design a new anime translation framework by deriving the prior knowledge of a pre-trained StyleGAN model. We introduce disentangled encoders to separately embed structure and appearance information into the same latent code, governed by four tailored losses. Moreover, we develop a FaceBank aggregation method that leverages the generated data of the StyleGAN, anchoring the prediction to produce in-domain animes. To empower our model and promote the research of anime translation, we propose the first anime portrait parsing dataset, Danbooru-Parsing, containing 4,921 densely labeled images across 17 classes. This dataset connects the face semantics with appearances, enabling our new constrained translation setting. We further show the editability of our results, and extend our method to manga images, by generating the first manga parsing pseudo data. Extensive experiments demonstrate the values of our new dataset and method, resulting in the first feasible solution on anime translation.

Differentiable rendering is frequently used in gradient descent-based inverse rendering pipelines to solve for scene parameters – such as reflectance or lighting properties – from target image inputs. Efficient computation of accurate, low variance gradients is critical for rapid convergence. While many methods employ variance reduction strategies, they operate independently on each gradient descent iteration, requiring large sample counts and computation. Gradients may however vary slowly between iterations, leading to unexplored potential benefits when reusing sample information to exploit this coherence. We develop an algorithm to reuse Monte Carlo gradient samples between gradient iterations, motivated by reservoir-based temporal importance resampling in forward rendering. Direct application of this method is not feasible, as we are computing many derivative estimates (i.e., one per optimization parameter) instead of a single pixel intensity estimate; moreover, each of these gradient estimates can affect multiple pixels, and gradients can take on negative values. We address these challenges by reformulating differential rendering integrals in parameter space, developing a new resampling estimator that treats negative functions, and combining these ideas into a reuse algorithm for inverse texture optimization. We significantly reduce gradient error compared to baselines, and demonstrate faster inverse rendering convergence in settings involving complex direct lighting and material textures.

We present a method to animate a character incorporating multiple part-wise motion priors (PMP). While previous works allow creating realistic articulated motions from reference data, the range of motion is largely limited by the available samples. Especially for the interaction-rich scenarios, it is impractical to attempt acquiring every possible interacting motion, as the combination of physical parameters increases exponentially. The proposed PMP allows us to assemble multiple part skills to animate a character, creating a diverse set of motions with different combinations of existing data. In our pipeline, we can train an agent with a wide range of part-wise priors. Therefore, each body part can obtain a kinematic insight of the style from the motion captures, or at the same time extract dynamics-related information from the additional part-specific simulation. For example, we can first train a general interaction skill, e.g. grasping, only for the dexterous part, and then combine the expert trajectories from the pre-trained agent with the kinematic priors of other limbs. Eventually, our whole-body agent learns a novel physical interaction skill even with the absence of the object trajectories in the reference motion sequence.

We propose a computational design approach for covering a surface with individually addressable RGB LEDs, effectively forming a low-resolution surface screen. To achieve a low-cost and scalable approach, we propose creating designs from flat PCB panels bent in-place along the surface of a 3D printed core. Working with standard rigid PCBs enables the use of established PCB manufacturing services, allowing the fabrication of designs with several hundred LEDs. Our approach optimizes the PCB geometry for folding, and then jointly optimizes the LED packing, circuit and routing, solving a challenging layout problem under strict manufacturing requirements. Unlike paper, PCBs cannot bend beyond a certain point without breaking. Therefore, we introduce parametric cut patterns acting as hinges, designed to allow bending while remaining compact. To tackle the joint optimization of placement, circuit and routing, we propose a specialized algorithm that splits the global problem into one sub-problem per triangle, which is then individually solved. Our technique generates PCB blueprints in a completely automated way. After being fabricated by a PCB manufacturing service, the boards are bent and glued by the user onto the 3D printed support. We demonstrate our technique on a range of physical models and virtual examples, creating intricate surface light patterns from hundreds of LEDs. The code and data for this paper are available at https://github.com/mfremer/pcbend.