We introduce a novel geometry-guided online video view synthesis method with enhanced view and temporal consistency. Traditional approaches achieve high-quality synthesis from dense multi-view camera setups but require significant computational resources. In contrast, selective-input methods reduce this cost but often compromise quality, leading to multi-view and temporal inconsistencies such as flickering artifacts. Our method addresses this challenge to deliver efficient, high-quality novel-view synthesis with view and temporal consistency. The key innovation of our approach lies in using global geometry to guide an image-based rendering pipeline. To accomplish this, we progressively refine depth maps using color difference masks across time. These depth maps are then accumulated through truncated signed distance fields in the synthesized view’s image space. This depth representation is view and temporally consistent, and is used to guide a pre-trained blending network that fuses multiple forward-rendered input-view images. Thus, the network is encouraged to output geometrically consistent synthesis results across multiple views and time. Our approach achieves consistent, high-quality video synthesis, while running efficiently in an online manner.

We present Large Inverse Rendering Model (LIRM), a transformer architecture that jointly reconstructs high-quality shape, materials, and radiance fields with view-dependent effects in less than a second. Our model builds upon the recent Large Reconstruction Models (LRMs) that achieve state-of-the-art sparse-view reconstruction quality. However, existing LRMs struggle to reconstruct unseen parts accurately and cannot recover glossy appearance or generate relightable 3D contents that can be consumed by standard Graphics engines. To address these limitations, we make three key technical contributions to build a more practical multi-view 3D reconstruction framework. First, we introduce an update model that allows us to progressively add more input views to improve our reconstruction. Second, we propose a hexa-plane neural SDF representation to better recover detailed textures, geometry and material parameters. Third, we develop a novel neural directional-embedding mechanism to handle view-dependent effects. Trained on a large-scale shape and material dataset with a tailored coarse-to-fine training scheme, our model achieves compelling results. It compares favorably to optimization-based dense-view inverse rendering methods in terms of geometry and relighting accuracy, while requiring only a fraction of the inference time.

We introduce the Deformable Gaussian Splats Large Reconstruction Model (DGS-LRM), the first feed-forward method predicting deformable 3D Gaussian splats from a monocular posed video of any dynamic scene. Feed-forward scene reconstruction has gained significant attention for its ability to rapidly create digital replicas of real-world environments. However, most existing models are limited to static scenes and fail to reconstruct the motion of moving objects. Developing a feed-forward model for dynamic scene reconstruction poses significant challenges, including the scarcity of training data and the need for appropriate 3D representations and training paradigms. To address these challenges, we introduce several key technical contributions: an enhanced large-scale synthetic dataset with ground-truth multi-view videos and dense 3D scene flow supervision; a per-pixel deformable 3D Gaussian representation that is easy to learn, supports high-quality dynamic view synthesis, and enables long-range 3D tracking; and a large transformer network that achieves real-time, generalizable dynamic scene reconstruction. Extensive qualitative and quantitative experiments demonstrate that DGS-LRM achieves dynamic scene reconstruction quality comparable to optimization-based methods, while significantly outperforming the state-of-the-art predictive dynamic reconstruction method on real-world examples. Its predicted physically grounded 3D deformation is accurate and can be readily adapted for long-range 3D tracking tasks, achieving performance on par with state-of-the-art monocular video 3D tracking methods.

We propose a method that achieves state-of-the-art rendering quality and efficiency on monocular dynamic scene reconstruction using deformable 3D Gaussians. Implicit deformable representations commonly model motion with a canonical space and time-dependent backward-warping deformation field. Our method, GauFRe, uses a forward-warping deformation to explicitly model non-rigid transformations of scene geometry. Specifically, we propose a template set of 3D Gaussians residing in a canonical space, and a time-dependent forward-warping deformation field to model dynamic objects. Additionally, we tailor a 3D Gaussian-specific static component supported by an inductive bias-aware initialization approach which allows the deformation field to focus on moving scene regions, improving the rendering of complex real-world motion. The differentiable pipeline is optimized end-to-end with a self-supervised rendering loss. Experiments show our method achieves competitive results and higher efficiency than both previous state-of-the-art NeRF and Gaussian-based methods. For real-world scenes, GauFRe can train in ≈20 mins and offer 96 FPS real-time rendering on an RTX 3090 GPU.

We present TextureDreamer, a novel image-guided texture synthesis method to transfer relightable textures from a small number of input images (3 to 5) to target 3D shapes across arbitrary categories. Texture creation is a pivotal challenge in vision and graphics. Industrial companies hire experienced artists to manually craft textures for 3D assets. Classical methods require densely sampled views and ac-curately aligned geometry, while learning-based methods are confined to category-specific shapes within the dataset. In contrast, TextureDreamer can transfer highly detailed, intricate textures from real-world environments to arbi-trary objects with only a few casually captured images, po-tentially significantly democratizing texture creation. Our core idea, personalized geometry-aware score distillation (PGSD), draws inspiration from recent advancements in diffuse models, including personalized modeling for texture information extraction, score distillation for detailed appearance synthesis, and explicit geometry guidance with ControlNet. Our integration and several essential modifications substantially improve the texture quality. Experiments on real images spanning different categories show that TextureDreamer can successfully transfer highly realistic, se-mantic meaningful texture to arbitrary objects, surpassing the visual quality of previous state-of-the-art. Project page: https://texturedreamer.github.io

We introduce ReplaceAnything3D model RAM3D, a novel method for 3D object replacement in 3D scenes based on users' text description. Given multi-view images of a scene, a text prompt describing the object to replace, and another describing the new object, our Erase-and-Replace approach can effectively swap objects in 3D scenes with newly generated content while maintaining 3D consistency across multiple viewpoints. We demonstrate the versatility of RAM3D by applying it to various realistic 3D scene types, showcasing results of modified objects that blend in seamlessly with the scene without impacting its overall integrity.

Depth estimation is an important step in many computer vision problems such as 3D reconstruction, novel view synthesis, and computational photography. Most existing work focuses on depth estimation from single frames. When applied to videos, the result lacks temporal consistency, showing flickering and swimming artifacts. In this paper we aim to estimate temporally consistent depth maps of video streams in an online setting. This is a difficult problem as future frames are not available and the method must choose between enforcing consistency and correcting errors from previous estimations. The presence of dynamic objects further complicates the problem. We propose to address these challenges by using a global point cloud that is dynamically updated each frame, along with a learned fusion approach in image space. Our approach encourages consistency while simultaneously allowing updates to handle errors and dynamic objects. Qualitative and quantitative results show that our method achieves state-of-the-art quality for consistent video depth estimation.

The task of synthesizing novel views from a single image has useful applications in virtual reality and mobile computing, and a number of approaches to the problem have been proposed in recent years. A Multiplane Image (MPI) estimates the scene as a stack of RGBA layers, and can model complex appearance effects, anti-alias depth errors and synthesize soft edges better than methods that use textured meshes or layered depth images. And unlike neural radiance fields, an MPI can be efficiently rendered on graphics hardware. However, MPIs are highly redundant and require a large number of depth layers to achieve plausible results. Based on the observation that the depth complexity in local image regions is lower than that over the entire image, we split an MPI into many small, tiled regions, each with only a few depth planes. We call this representation a Tiled Multiplane Image (TMPI). We propose a method for generating a TMPI with adaptive depth planes for single-view 3D photography in the wild. Our synthesized results are comparable to state-of-the-art single-view MPI methods while having lower computational overhead.

Image view synthesis has seen great success in reconstructing photorealistic visuals, thanks to deep learning and various novel representations. The next key step in immersive virtual experiences is view synthesis of dynamic scenes. However, several challenges exist due to the lack of high-quality training datasets, and the additional time dimension for videos of dynamic scenes. To address this issue, we introduce a multi-view video dataset, captured with a custom 10-camera rig in 120FPS. The dataset contains 96 high-quality scenes showing various visual effects and human interactions in outdoor scenes. We develop a new algorithm, Deep 3D Mask Volume, which enables temporally-stable view extrapolation from binocular videos of dynamic scenes, captured by static cameras. Our algorithm addresses the temporal inconsistency of disocclusions by identifying the error-prone areas with a 3D mask volume, and replaces them with static background observed throughout the video. Our method enables manipulation in 3D space as opposed to simple 2D masks, We demonstrate better temporal stability than frame-by-frame static view synthesis methods, or those that use 2D masks. The resulting view synthesis videos show minimal flickering artifacts and allow for larger translational movements.

This paper presents a stylized novel view synthesis method. Applying state-of-the-art stylization methods to novel views frame by frame often causes jittering artifacts due to the lack of cross-view consistency. Therefore, this paper investigates 3D scene stylization that provides a strong inductive bias for consistent novel view synthesis. Specifically, we adopt the emerging neural radiance fields (NeRF) as our choice of 3D scene representation for their capability to render high-quality novel views for a variety of scenes. However, as rendering a novel view from a NeRF requires a large number of samples, training a stylized NeRF requires a large amount of GPU memory that goes beyond an off-the-shelf GPU capacity. We introduce a new training method to address this problem by alternating the NeRF and stylization optimization steps. Such a method enables us to make full use of our hardware memory capacity to both generate images at higher resolution and adopt more expressive image style transfer methods. Our experiments show that our method produces stylized NeRFs for a wide range of content, including indoor, outdoor and dynamic scenes, and synthesizes high-quality novel views with cross-view consistency.

Image view synthesis has seen great success in reconstructing photorealistic visuals, thanks to deep learning and various novel representations. The next key step in immersive virtual experiences is view synthesis of dynamic scenes. However, several challenges exist due to the lack of high-quality training datasets, and the additional time dimension for videos of dynamic scenes. To address this issue, we introduce a multi-view video dataset, captured with a custom 10-camera rig in 120FPS. The dataset contains 96 high-quality scenes showing various visual effects and human interactions in outdoor scenes. We develop a new algorithm, Deep 3D Mask Volume, which enables temporally-stable view extrapolation from binocular videos of dynamic scenes, captured by static cameras. Our algorithm addresses the temporal inconsistency of disocclusions by identifying the error-prone areas with a 3D mask volume, and replaces them with static background observed throughout the video. Our method enables manipulation in 3D space as opposed to simple 2D masks, We demonstrate better temporal stability than frame-by-frame static view synthesis methods, or those that use 2D masks. The resulting view synthesis videos show minimal flickering artifacts and allow for larger translational movements.

Due to higher resolutions and refresh rates, as well as more photorealistic effects, real-time rendering has become increasingly challenging for video games and emerging virtual reality headsets. To meet this demand, modern graphics hardware and game engines often reduce the computational cost by rendering at a lower resolution and then upsampling to the native resolution. Following the recent advances in image and video superresolution in computer vision, we propose a machine learning approach that is specifically tailored for high-quality upsampling of rendered content in real-time applications. The main insight of our work is that in rendered content, the image pixels are point-sampled, but precise temporal dynamics are available. Our method combines this specific information that is typically available in modern renderers (i.e., depth and dense motion vectors) with a novel temporal network design that takes into account such specifics and is aimed at maximizing video quality while delivering real-time performance. By training on a large synthetic dataset rendered from multiple 3D scenes with recorded camera motion, we demonstrate high fidelity and temporally stable results in real-time, even in the highly challenging 4 × 4 upsampling scenario, significantly outperforming existing superresolution and temporal antialiasing work.

Recently, several discriminative learning approaches have been proposed for effective image restoration, achieving convincing trade-off between image quality and computational efficiency. However, these methods require separate training for each restoration task (e.g., denoising, deblurring, demosaicing) and problem condition (e.g., noise level of input images). This makes it time-consuming and difficult to encompass all tasks and conditions during training. In this paper, we propose a discriminative transfer learning method that incorporates formal proximal optimization and discriminative learning for general image restoration. The method requires a single-pass discriminative training and allows for reuse across various problems and conditions while achieving an efficiency comparable to previous discriminative approaches. Furthermore, after being trained, our model can be easily transferred to new likelihood terms to solve untrained tasks, or be combined with existing priors to further improve image restoration quality.

Addressing vergence-accommodation conflict in head-mounted displays (HMDs) requires resolving two interrelated problems. First, the hardware must support viewing sharp imagery over the full accommodation range of the user. Second, HMDs should accurately reproduce retinal defocus blur to correctly drive accommodation. A multitude of accommodation-supporting HMDs have been proposed, with three architectures receiving particular attention: varifocal, multifocal, and light field displays. These designs all extend depth of focus, but rely on computationally expensive rendering and optimization algorithms to reproduce accurate defocus blur (often limiting content complexity and interactive applications). To date, no unified framework has been proposed to support driving these emerging HMDs using commodity content. In this paper, we introduce DeepFocus , a generic, end-to-end convolutional neural network designed to efficiently solve the full range of computational tasks for accommodation-supporting HMDs. This network is demonstrated to accurately synthesize defocus blur, focal stacks, multilayer decompositions, and multiview imagery using only commonly available RGB-D images, enabling real-time, near-correct depictions of retinal blur with a broad set of accommodation-supporting HMDs.

Continuous-wave time-of-flight (ToF) cameras show great promise as low-cost depth image sensors in mobile applications. However, they also suffer from several challenges, including limited illumination intensity, which mandates the use of large numerical aperture lenses, and thus results in a shallow depth of field, making it difficult to capture scenes with large variations in depth. Another shortcoming is the limited spatial resolution of currently available ToF sensors. In this paper we analyze the image formation model for blurred ToF images. By directly working with raw sensor measurements but regularizing the recovered depth and amplitude images, we are able to simultaneously deblur and super-resolve the output of ToF cameras. Our method outperforms existing methods on both synthetic and real datasets. In the future our algorithm should extend easily to cameras that do not follow the cosine model of continuous-wave sensors, as well as to multi-frequency or multi-phase imaging employed in more recent ToF cameras.

Blind motion deblurring from a single image is a highly under-constrained problem with many degenerate solutions. A good approximation of the intrinsic image can, therefore, only be obtained with the help of prior information in the form of (often nonconvex) regularization terms for both the intrinsic image and the kernel. While the best choice of image priors is still a topic of ongoing investigation, this research is made more complicated by the fact that historically each new prior requires the development of a custom optimization method. In this paper, we develop a stochastic optimization method for blind deconvolution. Since this stochastic solver does not require the explicit computation of the gradient of the objective function and uses only efficient local evaluation of the objective, new priors can be implemented and tested very quickly. We demonstrate that this framework, in combination with different image priors produces results with Peak Signal-to-Noise Ratio (PSNR) values that match or exceed the results obtained by much more complex state-of-the-art blind motion deblurring algorithms.

The functional difference between a diffuse wall and a mirror is well understood: one scatters back into all directions, and the other one preserves the directionality of reflected light. The temporal structure of the light, however, is left intact by both: assuming simple surface reflection, photons that arrive first are reflected first. In this paper, we exploit this insight to recover objects outside the line of sight from second-order diffuse reflections, effectively turning walls into mirrors. We formulate the reconstruction task as a linear inverse problem on the transient response of a scene, which we acquire using an affordable setup consisting of a modulated light source and a time-of-flight image sensor. By exploiting sparsity in the reconstruction domain, we achieve resolutions in the order of a few centimeters for object shape (depth and laterally) and albedo. Our method is robust to ambient light and works for large room-sized scenes. It is drastically faster and less expensive than previous approaches using femtosecond lasers and streak cameras, and does not require any moving parts.

We analyze light propagation in an unknown scene using projectors and cameras that operate at transient timescales. In this new photography regime, the projector emits a spatio-temporal 3D signal and the camera receives a transformed version of it, determined by the set of all light transport paths through the scene and the time delays they induce. The underlying 3D-to-3D transformation encodes scene geometry and global transport in great detail, but individual transport components ( e.g ., direct reflections, inter-reflections, caustics, etc .) are coupled nontrivially in both space and time. To overcome this complexity, we observe that transient light transport is always separable in the temporal frequency domain . This makes it possible to analyze transient transport one temporal frequency at a time by trivially adapting techniques from conventional projector-to-camera transport. We use this idea in a prototype that offers three never-seen-before abilities: (1) acquiring time-of-flight depth images that are robust to general indirect transport, such as interreflections and caustics; (2) distinguishing between direct views of objects and their mirror reflection; and (3) using a photonic mixer device to capture sharp, evolving wavefronts of "light-in-flight".