Transparent objects are ubiquitous in industry, pharmaceuticals, and households. Grasping and manipulating these objects is a significant challenge for robots. Existing methods have difficulty reconstructing complete depth maps for challenging transparent objects, leaving holes in the depth reconstruction. Recent work has shown neural radiance fields (NeRFs) work well for depth perception in scenes with transparent objects, and these depth maps can be used to grasp transparent objects with high accuracy. NeRF-based depth reconstruction can still struggle with especially challenging transparent objects and lighting conditions. In this work, we propose Residual-NeRF, a method to improve depth perception and training speed for transparent objects. Robots often operate in the same area, such as a kitchen. By first learning a background NeRF of the scene without transparent objects to be manipulated, we reduce the ambiguity faced by learning the changes with the new object. We propose training two additional networks: a residual NeRF learns to infer residual RGB values and densities, and a Mixnet learns how to combine background and residual NeRFs. We contribute synthetic and real experiments that suggest Residual-NeRF improves depth perception of transparent objects. The results on synthetic data suggest Residual-NeRF outperforms the baselines with a 46.1% lower RMSE and a 29.5% lower MAE. Real-world qualitative experiments suggest Residual-NeRF leads to more robust depth maps with less noise and fewer holes. Website: https://residual-nerf.github.io
User-friendly 3D object editing is a challenging task that has attracted significant attention recently. The limitations of direct 3D object editing without 2D prior knowledge have prompted increased attention towards utilizing 2D generative models for 3D editing. While existing methods like Instruct NeRF-to-NeRF offer a solution, they often lack user-friendliness, particularly due to semantic guided editing. In the realm of 3D representation, 3D Gaussian Splatting emerges as a promising approach for its efficiency and natural explicit property, facilitating precise editing tasks. Building upon these insights, we propose DragGaussian, a 3D object drag-editing framework based on 3D Gaussian Splatting, leveraging diffusion models for interactive image editing with open-vocabulary input. This framework enables users to perform drag-based editing on pre-trained 3D Gaussian object models, producing modified 2D images through multi-view consistent editing. Our contributions include the introduction of a new task, the development of DragGaussian for interactive point-based 3D editing, and comprehensive validation of its effectiveness through qualitative and quantitative experiments.
Audio-driven talking head generation is advancing from 2D to 3D content. Notably, Neural Radiance Field (NeRF) is in the spotlight as a means to synthesize high-quality 3D talking head outputs. Unfortunately, this NeRF-based approach typically requires a large number of paired audio-visual data for each identity, thereby limiting the scalability of the method. Although there have been attempts to generate audio-driven 3D talking head animations with a single image, the results are often unsatisfactory due to insufficient information on obscured regions in the image. In this paper, we mainly focus on addressing the overlooked aspect of 3D consistency in the one-shot, audio-driven domain, where facial animations are synthesized primarily in front-facing perspectives. We propose a novel method, NeRFFaceSpeech, which enables to produce high-quality 3D-aware talking head. Using prior knowledge of generative models combined with NeRF, our method can craft a 3D-consistent facial feature space corresponding to a single image. Our spatial synchronization method employs audio-correlated vertex dynamics of a parametric face model to transform static image features into dynamic visuals through ray deformation, ensuring realistic 3D facial motion. Moreover, we introduce LipaintNet that can replenish the lacking information in the inner-mouth area, which can not be obtained from a given single image. The network is trained in a self-supervised manner by utilizing the generative capabilities without additional data. The comprehensive experiments demonstrate the superiority of our method in generating audio-driven talking heads from a single image with enhanced 3D consistency compared to previous approaches. In addition, we introduce a quantitative way of measuring the robustness of a model against pose changes for the first time, which has been possible only qualitatively.
Point-based representations have recently gained popularity in novel view synthesis, for their unique advantages, e.g., intuitive geometric representation, simple manipulation, and faster convergence. However, based on our observation, these point-based neural re-rendering methods are only expected to perform well under ideal conditions and suffer from noisy, patchy points and unbounded scenes, which are challenging to handle but defacto common in real applications. To this end, we revisit one such influential method, known as Neural Point-based Graphics (NPBG), as our baseline, and propose Robust Point-based Graphics (RPBG). We in-depth analyze the factors that prevent NPBG from achieving satisfactory renderings on generic datasets, and accordingly reform the pipeline to make it more robust to varying datasets in-the-wild. Inspired by the practices in image restoration, we greatly enhance the neural renderer to enable the attention-based correction of point visibility and the inpainting of incomplete rasterization, with only acceptable overheads. We also seek for a simple and lightweight alternative for environment modeling and an iterative method to alleviate the problem of poor geometry. By thorough evaluation on a wide range of datasets with different shooting conditions and camera trajectories, RPBG stably outperforms the baseline by a large margin, and exhibits its great robustness over state-of-the-art NeRF-based variants. Code available at https://github.com/QT-Zhu/RPBG.
Neural fields (NeRF) have emerged as a promising approach for representing continuous 3D scenes. Nevertheless, the lack of semantic encoding in NeRFs poses a significant challenge for scene decomposition. To address this challenge, we present a single model, Multi-Modal Decomposition NeRF (${M^2D}$NeRF), that is capable of both text-based and visual patch-based edits. Specifically, we use multi-modal feature distillation to integrate teacher features from pretrained visual and language models into 3D semantic feature volumes, thereby facilitating consistent 3D editing. To enforce consistency between the visual and language features in our 3D feature volumes, we introduce a multi-modal similarity constraint. We also introduce a patch-based joint contrastive loss that helps to encourage object-regions to coalesce in the 3D feature space, resulting in more precise boundaries. Experiments on various real-world scenes show superior performance in 3D scene decomposition tasks compared to prior NeRF-based methods.
Neural Radiance Field~(NeRF) achieves extremely high quality in object-scaled and indoor scene reconstruction. However, there exist some challenges when reconstructing large-scale scenes. MLP-based NeRFs suffer from limited network capacity, while volume-based NeRFs are heavily memory-consuming when the scene resolution increases. Recent approaches propose to geographically partition the scene and learn each sub-region using an individual NeRF. Such partitioning strategies help volume-based NeRF exceed the single GPU memory limit and scale to larger scenes. However, this approach requires multiple background NeRF to handle out-of-partition rays, which leads to redundancy of learning. Inspired by the fact that the background of current partition is the foreground of adjacent partition, we propose a scalable scene reconstruction method based on joint Multi-resolution Hash Grids, named DistGrid. In this method, the scene is divided into multiple closely-paved yet non-overlapped Axis-Aligned Bounding Boxes, and a novel segmented volume rendering method is proposed to handle cross-boundary rays, thereby eliminating the need for background NeRFs. The experiments demonstrate that our method outperforms existing methods on all evaluated large-scale scenes, and provides visually plausible scene reconstruction. The scalability of our method on reconstruction quality is further evaluated qualitatively and quantitatively.
Neural Radiance Fields (NeRFs) have become a rapidly growing research field with the potential to revolutionize typical photogrammetric workflows, such as those used for 3D scene reconstruction. As input, NeRFs require multi-view images with corresponding camera poses as well as the interior orientation. In the typical NeRF workflow, the camera poses and the interior orientation are estimated in advance with Structure from Motion (SfM). But the quality of the resulting novel views, which depends on different parameters such as the number and distribution of available images, as well as the accuracy of the related camera poses and interior orientation, is difficult to predict. In addition, SfM is a time-consuming pre-processing step, and its quality strongly depends on the image content. Furthermore, the undefined scaling factor of SfM hinders subsequent steps in which metric information is required. In this paper, we evaluate the potential of NeRFs for industrial robot applications. We propose an alternative to SfM pre-processing: we capture the input images with a calibrated camera that is attached to the end effector of an industrial robot and determine accurate camera poses with metric scale based on the robot kinematics. We then investigate the quality of the novel views by comparing them to ground truth, and by computing an internal quality measure based on ensemble methods. For evaluation purposes, we acquire multiple datasets that pose challenges for reconstruction typical of industrial applications, like reflective objects, poor texture, and fine structures. We show that the robot-based pose determination reaches similar accuracy as SfM in non-demanding cases, while having clear advantages in more challenging scenarios. Finally, we present first results of applying the ensemble method to estimate the quality of the synthetic novel view in the absence of a ground truth.
Due to the limited model capacity, leveraging distributed Neural Radiance Fields (NeRFs) for modeling extensive urban environments has become a necessity. However, current distributed NeRF registration approaches encounter aliasing artifacts, arising from discrepancies in rendering resolutions and suboptimal pose precision. These factors collectively deteriorate the fidelity of pose estimation within NeRF frameworks, resulting in occlusion artifacts during the NeRF blending stage. In this paper, we present a distributed NeRF system with tri-stage pose optimization. In the first stage, precise poses of images are achieved by bundle adjusting Mip-NeRF 360 with a coarse-to-fine strategy. In the second stage, we incorporate the inverting Mip-NeRF 360, coupled with the truncated dynamic low-pass filter, to enable the achievement of robust and precise poses, termed Frame2Model optimization. On top of this, we obtain a coarse transformation between NeRFs in different coordinate systems. In the third stage, we fine-tune the transformation between NeRFs by Model2Model pose optimization. After obtaining precise transformation parameters, we proceed to implement NeRF blending, showcasing superior performance metrics in both real-world and simulation scenarios. Codes and data will be publicly available at https://github.com/boilcy/Distributed-NeRF.
Despite the emergence of successful NeRF inpainting methods built upon explicit RGB and depth 2D inpainting supervisions, these methods are inherently constrained by the capabilities of their underlying 2D inpainters. This is due to two key reasons: (i) independently inpainting constituent images results in view-inconsistent imagery, and (ii) 2D inpainters struggle to ensure high-quality geometry completion and alignment with inpainted RGB images. To overcome these limitations, we propose a novel approach called MVIP-NeRF that harnesses the potential of diffusion priors for NeRF inpainting, addressing both appearance and geometry aspects. MVIP-NeRF performs joint inpainting across multiple views to reach a consistent solution, which is achieved via an iterative optimization process based on Score Distillation Sampling (SDS). Apart from recovering the rendered RGB images, we also extract normal maps as a geometric representation and define a normal SDS loss that motivates accurate geometry inpainting and alignment with the appearance. Additionally, we formulate a multi-view SDS score function to distill generative priors simultaneously from different view images, ensuring consistent visual completion when dealing with large view variations. Our experimental results show better appearance and geometry recovery than previous NeRF inpainting methods.
In this paper, we present a new approach to bridge the domain gap between synthetic and real-world data for unmanned aerial vehicle (UAV)-based perception. Our formulation is designed for dynamic scenes, consisting of small moving objects or human actions. We propose an extension of K-Planes Neural Radiance Field (NeRF), wherein our algorithm stores a set of tiered feature vectors. The tiered feature vectors are generated to effectively model conceptual information about a scene as well as an image decoder that transforms output feature maps into RGB images. Our technique leverages the information amongst both static and dynamic objects within a scene and is able to capture salient scene attributes of high altitude videos. We evaluate its performance on challenging datasets, including Okutama Action and UG2, and observe considerable improvement in accuracy over state of the art neural rendering methods.
Despite significant advancements in Neural Radiance Fields (NeRFs), the renderings may still suffer from aliasing and blurring artifacts, since it remains a fundamental challenge to effectively and efficiently characterize anisotropic areas induced by the cone-casting procedure. This paper introduces a Ripmap-Encoded Platonic Solid representation to precisely and efficiently featurize 3D anisotropic areas, achieving high-fidelity anti-aliasing renderings. Central to our approach are two key components: Platonic Solid Projection and Ripmap encoding. The Platonic Solid Projection factorizes the 3D space onto the unparalleled faces of a certain Platonic solid, such that the anisotropic 3D areas can be projected onto planes with distinguishable characterization. Meanwhile, each face of the Platonic solid is encoded by the Ripmap encoding, which is constructed by anisotropically pre-filtering a learnable feature grid, to enable featurzing the projected anisotropic areas both precisely and efficiently by the anisotropic area-sampling. Extensive experiments on both well-established synthetic datasets and a newly captured real-world dataset demonstrate that our Rip-NeRF attains state-of-the-art rendering quality, particularly excelling in the fine details of repetitive structures and textures, while maintaining relatively swift training times.
Active view selection in 3D scene reconstruction has been widely studied since training on informative views is critical for reconstruction. Recently, Neural Radiance Fields (NeRF) variants have shown promising results in active 3D reconstruction using uncertainty-guided view selection. They utilize uncertainties estimated with neural networks that encode scene geometry and appearance. However, the choice of uncertainty integration methods, either voxel-based or neural rendering, has conventionally depended on the types of scene uncertainty being estimated, whether geometric or appearance-related. In this paper, we introduce Colorized Surface Voxel (CSV)-based view selection, a new next-best view (NBV) selection method exploiting surface voxel-based measurement of uncertainty in scene appearance. CSV encapsulates the uncertainty of estimated scene appearance (e.g., color uncertainty) and estimated geometric information (e.g., surface). Using the geometry information, we interpret the uncertainty of scene appearance 3D-wise during the aggregation of the per-voxel uncertainty. Consequently, the uncertainty from occluded and complex regions is recognized under challenging scenarios with limited input data. Our method outperforms previous works on popular datasets, DTU and Blender, and our new dataset with imbalanced viewpoints, showing that the CSV-based view selection significantly improves performance by up to 30%.
The advances in the Neural Radiance Fields (NeRF) research offer extensive applications in diverse domains, but protecting their copyrights has not yet been researched in depth. Recently, NeRF watermarking has been considered one of the pivotal solutions for safely deploying NeRF-based 3D representations. However, existing methods are designed to apply only to implicit or explicit NeRF representations. In this work, we introduce an innovative watermarking method that can be employed in both representations of NeRF. This is achieved by fine-tuning NeRF to embed binary messages in the rendering process. In detail, we propose utilizing the discrete wavelet transform in the NeRF space for watermarking. Furthermore, we adopt a deferred back-propagation technique and introduce a combination with the patch-wise loss to improve rendering quality and bit accuracy with minimum trade-offs. We evaluate our method in three different aspects: capacity, invisibility, and robustness of the embedded watermarks in the 2D-rendered images. Our method achieves state-of-the-art performance with faster training speed over the compared state-of-the-art methods.
Photorealistic simulation plays a crucial role in applications such as autonomous driving, where advances in neural radiance fields (NeRFs) may allow better scalability through the automatic creation of digital 3D assets. However, reconstruction quality suffers on street scenes due to largely collinear camera motions and sparser samplings at higher speeds. On the other hand, the application often demands rendering from camera views that deviate from the inputs to accurately simulate behaviors like lane changes. In this paper, we propose several insights that allow a better utilization of Lidar data to improve NeRF quality on street scenes. First, our framework learns a geometric scene representation from Lidar, which is fused with the implicit grid-based representation for radiance decoding, thereby supplying stronger geometric information offered by explicit point cloud. Second, we put forth a robust occlusion-aware depth supervision scheme, which allows utilizing densified Lidar points by accumulation. Third, we generate augmented training views from Lidar points for further improvement. Our insights translate to largely improved novel view synthesis under real driving scenes.
Meticulous 3D environment representations have been a longstanding goal in computer vision and robotics fields. The recent emergence of neural implicit representations has introduced radical innovation to this field as implicit representations enable numerous capabilities. Among these, the Neural Radiance Field (NeRF) has sparked a trend because of the huge representational advantages, such as simplified mathematical models, compact environment storage, and continuous scene representations. Apart from computer vision, NeRF has also shown tremendous potential in the field of robotics. Thus, we create this survey to provide a comprehensive understanding of NeRF in the field of robotics. By exploring the advantages and limitations of NeRF, as well as its current applications and future potential, we hope to shed light on this promising area of research. Our survey is divided into two main sections: \textit{The Application of NeRF in Robotics} and \textit{The Advance of NeRF in Robotics}, from the perspective of how NeRF enters the field of robotics. In the first section, we introduce and analyze some works that have been or could be used in the field of robotics from the perception and interaction perspectives. In the second section, we show some works related to improving NeRF's own properties, which are essential for deploying NeRF in the field of robotics. In the discussion section of the review, we summarize the existing challenges and provide some valuable future research directions for reference.
Neural Radiance Fields (NeRF) have achieved impressive results in 3D reconstruction and novel view generation. A significant challenge within NeRF involves editing reconstructed 3D scenes, such as object removal, which demands consistency across multiple views and the synthesis of high-quality perspectives. Previous studies have integrated depth priors, typically sourced from LiDAR or sparse depth estimates from COLMAP, to enhance NeRF's performance in object removal. However, these methods are either expensive or time-consuming. This paper proposes a new pipeline that leverages SpinNeRF and monocular depth estimation models like ZoeDepth to enhance NeRF's performance in complex object removal with improved efficiency. A thorough evaluation of COLMAP's dense depth reconstruction on the KITTI dataset is conducted to demonstrate that COLMAP can be viewed as a cost-effective and scalable alternative for acquiring depth ground truth compared to traditional methods like LiDAR. This serves as the basis for evaluating the performance of monocular depth estimation models to determine the best one for generating depth priors for SpinNeRF. The new pipeline is tested in various scenarios involving 3D reconstruction and object removal, and the results indicate that our pipeline significantly reduces the time required for the acquisition of depth priors for object removal and enhances the fidelity of the synthesized views, suggesting substantial potential for building high-fidelity digital twin systems with increased efficiency in the future.
This paper focuses on training a robust RGB-D registration model without ground-truth pose supervision. Existing methods usually adopt a pairwise training strategy based on differentiable rendering, which enforces the photometric and the geometric consistency between the two registered frames as supervision. However, this frame-to-frame framework suffers from poor multi-view consistency due to factors such as lighting changes, geometry occlusion and reflective materials. In this paper, we present NeRF-UR, a novel frame-to-model optimization framework for unsupervised RGB-D registration. Instead of frame-to-frame consistency, we leverage the neural radiance field (NeRF) as a global model of the scene and use the consistency between the input and the NeRF-rerendered frames for pose optimization. This design can significantly improve the robustness in scenarios with poor multi-view consistency and provides better learning signal for the registration model. Furthermore, to bootstrap the NeRF optimization, we create a synthetic dataset, Sim-RGBD, through a photo-realistic simulator to warm up the registration model. By first training the registration model on Sim-RGBD and later unsupervisedly fine-tuning on real data, our framework enables distilling the capability of feature extraction and registration from simulation to reality. Our method outperforms the state-of-the-art counterparts on two popular indoor RGB-D datasets, ScanNet and 3DMatch. Code and models will be released for paper reproduction.
We present GStalker, a 3D audio-driven talking face generation model with Gaussian Splatting for both fast training (40 minutes) and real-time rendering (125 FPS) with a 3$\sim$5 minute video for training material, in comparison with previous 2D and 3D NeRF-based modeling frameworks which require hours of training and seconds of rendering per frame. Specifically, GSTalker learns an audio-driven Gaussian deformation field to translate and transform 3D Gaussians to synchronize with audio information, in which multi-resolution hashing grid-based tri-plane and temporal smooth module are incorporated to learn accurate deformation for fine-grained facial details. In addition, a pose-conditioned deformation field is designed to model the stabilized torso. To enable efficient optimization of the condition Gaussian deformation field, we initialize 3D Gaussians by learning a coarse static Gaussian representation. Extensive experiments in person-specific videos with audio tracks validate that GSTalker can generate high-fidelity and audio-lips synchronized results with fast training and real-time rendering speed.
Neural Radiance Fields (NeRF) show impressive performance in photo-realistic free-view rendering of scenes. Recent improvements on the NeRF such as TensoRF and ZipNeRF employ explicit models for faster optimization and rendering, as compared to the NeRF that employs an implicit representation. However, both implicit and explicit radiance fields require dense sampling of images in the given scene. Their performance degrades significantly when only a sparse set of views is available. Researchers find that supervising the depth estimated by a radiance field helps train it effectively with fewer views. The depth supervision is obtained either using classical approaches or neural networks pre-trained on a large dataset. While the former may provide only sparse supervision, the latter may suffer from generalization issues. As opposed to the earlier approaches, we seek to learn the depth supervision by designing augmented models and training them along with the main radiance field. Further, we aim to design a framework of regularizations that can work across different implicit and explicit radiance fields. We observe that certain features of these radiance field models overfit to the observed images in the sparse-input scenario. Our key finding is that reducing the capability of the radiance fields with respect to positional encoding, the number of decomposed tensor components or the size of the hash table, constrains the model to learn simpler solutions, which estimate better depth in certain regions. By designing augmented models based on such reduced capabilities, we obtain better depth supervision for the main radiance field. We achieve state-of-the-art view-synthesis performance with sparse input views on popular datasets containing forward-facing and 360$^\circ$ scenes by employing the above regularizations.
We present Real-time Gaussian SLAM (RTG-SLAM), a real-time 3D reconstruction system with an RGBD camera for large-scale environments using Gaussian splatting. The system features a compact Gaussian representation and a highly efficient on-the-fly Gaussian optimization scheme. We force each Gaussian to be either opaque or nearly transparent, with the opaque ones fitting the surface and dominant colors, and transparent ones fitting residual colors. By rendering depth in a different way from color rendering, we let a single opaque Gaussian well fit a local surface region without the need of multiple overlapping Gaussians, hence largely reducing the memory and computation cost. For on-the-fly Gaussian optimization, we explicitly add Gaussians for three types of pixels per frame: newly observed, with large color errors, and with large depth errors. We also categorize all Gaussians into stable and unstable ones, where the stable Gaussians are expected to well fit previously observed RGBD images and otherwise unstable. We only optimize the unstable Gaussians and only render the pixels occupied by unstable Gaussians. In this way, both the number of Gaussians to be optimized and pixels to be rendered are largely reduced, and the optimization can be done in real time. We show real-time reconstructions of a variety of large scenes. Compared with the state-of-the-art NeRF-based RGBD SLAM, our system achieves comparable high-quality reconstruction but with around twice the speed and half the memory cost, and shows superior performance in the realism of novel view synthesis and camera tracking accuracy.