Dynamic Neural Radiance Field (NeRF) from monocular videos has recently been explored for space-time novel view synthesis and achieved excellent results. However, defocus blur caused by depth variation often occurs in video capture, compromising the quality of dynamic reconstruction because the lack of sharp details interferes with modeling temporal consistency between input views. To tackle this issue, we propose D2RF, the first dynamic NeRF method designed to restore sharp novel views from defocused monocular videos. We introduce layered Depth-of-Field (DoF) volume rendering to model the defocus blur and reconstruct a sharp NeRF supervised by defocused views. The blur model is inspired by the connection between DoF rendering and volume rendering. The opacity in volume rendering aligns with the layer visibility in DoF rendering.To execute the blurring, we modify the layered blur kernel to the ray-based kernel and employ an optimized sparse kernel to gather the input rays efficiently and render the optimized rays with our layered DoF volume rendering. We synthesize a dataset with defocused dynamic scenes for our task, and extensive experiments on our dataset show that our method outperforms existing approaches in synthesizing all-in-focus novel views from defocus blur while maintaining spatial-temporal consistency in the scene.
We propose MomentsNeRF, a novel framework for one- and few-shot neural rendering that predicts a neural representation of a 3D scene using Orthogonal Moments. Our architecture offers a new transfer learning method to train on multi-scenes and incorporate a per-scene optimization using one or a few images at test time. Our approach is the first to successfully harness features extracted from Gabor and Zernike moments, seamlessly integrating them into the NeRF architecture. We show that MomentsNeRF performs better in synthesizing images with complex textures and shapes, achieving a significant noise reduction, artifact elimination, and completing the missing parts compared to the recent one- and few-shot neural rendering frameworks. Extensive experiments on the DTU and Shapenet datasets show that MomentsNeRF improves the state-of-the-art by {3.39\;dB\;PSNR}, 11.1% SSIM, 17.9% LPIPS, and 8.3% DISTS metrics. Moreover, it outperforms state-of-the-art performance for both novel view synthesis and single-image 3D view reconstruction. The source code is accessible at: https://amughrabi.github.io/momentsnerf/.
One of the core activities of an active observer involves moving to secure a "better" view of the scene, where the definition of "better" is task-dependent. This paper focuses on the task of human pose estimation from videos capturing a person's activity. Self-occlusions within the scene can complicate or even prevent accurate human pose estimation. To address this, relocating the camera to a new vantage point is necessary to clarify the view, thereby improving 2D human pose estimation. This paper formalizes the process of achieving an improved viewpoint. Our proposed solution to this challenge comprises three main components: a NeRF-based Drone-View Data Generation Framework, an On-Drone Network for Camera View Error Estimation, and a Combined Planner for devising a feasible motion plan to reposition the camera based on the predicted errors for camera views. The Data Generation Framework utilizes NeRF-based methods to generate a comprehensive dataset of human poses and activities, enhancing the drone's adaptability in various scenarios. The Camera View Error Estimation Network is designed to evaluate the current human pose and identify the most promising next viewing angles for the drone, ensuring a reliable and precise pose estimation from those angles. Finally, the combined planner incorporates these angles while considering the drone's physical and environmental limitations, employing efficient algorithms to navigate safe and effective flight paths. This system represents a significant advancement in active 2D human pose estimation for an autonomous UAV agent, offering substantial potential for applications in aerial cinematography by improving the performance of autonomous human pose estimation and maintaining the operational safety and efficiency of UAVs.
3D building reconstruction from imaging data is an important task for many applications ranging from urban planning to reconnaissance. Modern Novel View synthesis (NVS) methods like NeRF and Gaussian Splatting offer powerful techniques for developing 3D models from natural 2D imagery in an unsupervised fashion. These algorithms generally require input training views surrounding the scene of interest, which, in the case of large buildings, is typically not available across all camera elevations. In particular, the most readily available camera viewpoints at scale across most buildings are at near-ground (e.g., with mobile phones) and aerial (drones) elevations. However, due to the significant difference in viewpoint between drone and ground image sets, camera registration - a necessary step for NVS algorithms - fails. In this work we propose a method, DRAGON, that can take drone and ground building imagery as input and produce a 3D NVS model. The key insight of DRAGON is that intermediate elevation imagery may be extrapolated by an NVS algorithm itself in an iterative procedure with perceptual regularization, thereby bridging the visual feature gap between the two elevations and enabling registration. We compiled a semi-synthetic dataset of 9 large building scenes using Google Earth Studio, and quantitatively and qualitatively demonstrate that DRAGON can generate compelling renderings on this dataset compared to baseline strategies.
Understanding 3D scenes is a crucial challenge in computer vision research with applications spanning multiple domains. Recent advancements in distilling 2D vision-language foundation models into neural fields, like NeRF and 3DGS, enable open-vocabulary segmentation of 3D scenes from 2D multi-view images without the need for precise 3D annotations. However, while effective, these methods typically rely on the per-pixel distillation of high-dimensional CLIP features, introducing ambiguity and necessitating complex regularization strategies, which adds inefficiency during training. This paper presents MaskField, which enables efficient 3D open-vocabulary segmentation with neural fields from a novel perspective. Unlike previous methods, MaskField decomposes the distillation of mask and semantic features from foundation models by formulating a mask feature field and queries. MaskField overcomes ambiguous object boundaries by naturally introducing SAM segmented object shapes without extra regularization during training. By circumventing the direct handling of dense high-dimensional CLIP features during training, MaskField is particularly compatible with explicit scene representations like 3DGS. Our extensive experiments show that MaskField not only surpasses prior state-of-the-art methods but also achieves remarkably fast convergence. We hope that MaskField will inspire further exploration into how neural fields can be trained to comprehend 3D scenes from 2D models.
Neural implicit representation of visual scenes has attracted a lot of attention in recent research of computer vision and graphics. Most prior methods focus on how to reconstruct 3D scene representation from a set of images. In this work, we demonstrate the possibility to recover the neural radiance fields (NeRF) from a single blurry image and its corresponding event stream. We model the camera motion with a cubic B-Spline in SE(3) space. Both the blurry image and the brightness change within a time interval, can then be synthesized from the 3D scene representation given the 6-DoF poses interpolated from the cubic B-Spline. Our method can jointly learn both the implicit neural scene representation and recover the camera motion by minimizing the differences between the synthesized data and the real measurements without pre-computed camera poses from COLMAP. We evaluate the proposed method with both synthetic and real datasets. The experimental results demonstrate that we are able to render view-consistent latent sharp images from the learned NeRF and bring a blurry image alive in high quality. Code and data are available at https://github.com/wu-cvgl/BeNeRF.
NeRF-based methods reconstruct 3D scenes by building a radiance field with implicit or explicit representations. While NeRF-based methods can perform novel view synthesis (NVS) at arbitrary scale, the performance in high-resolution novel view synthesis (HRNVS) with low-resolution (LR) optimization often results in oversmoothing. On the other hand, single-image super-resolution (SR) aims to enhance LR images to HR counterparts but lacks multi-view consistency. To address these challenges, we propose Arbitrary-Scale Super-Resolution NeRF (ASSR-NeRF), a novel framework for super-resolution novel view synthesis (SRNVS). We propose an attention-based VoxelGridSR model to directly perform 3D super-resolution (SR) on the optimized volume. Our model is trained on diverse scenes to ensure generalizability. For unseen scenes trained with LR views, we then can directly apply our VoxelGridSR to further refine the volume and achieve multi-view consistent SR. We demonstrate quantitative and qualitatively that the proposed method achieves significant performance in SRNVS.
Purpose: This study addresses the challenge of extended SPECT imaging duration under low-count conditions, as encountered in Lu-177 SPECT imaging, by developing a self-supervised learning approach to synthesize skipped SPECT projection views, thus shortening scan times in clinical settings. Methods: We employed a self-supervised coordinate-based learning technique, adapting the neural radiance field (NeRF) concept in computer vision to synthesize under-sampled SPECT projection views. For each single scan, we used self-supervised coordinate learning to estimate skipped SPECT projection views. The method was tested with various down-sampling factors (DFs=2, 4, 8) on both Lu-177 phantom SPECT/CT measurements and clinical SPECT/CT datasets, from 11 patients undergoing Lu-177 DOTATATE and 6 patients undergoing Lu-177 PSMA-617 radiopharmaceutical therapy. Results: For SPECT reconstructions, our method outperformed the use of linearly interpolated projections and partial projection views in relative contrast-to-noise-ratios (RCNR) averaged across different downsampling factors: 1) DOTATATE: 83% vs. 65% vs. 67% for lesions and 86% vs. 70% vs. 67% for kidney, 2) PSMA: 76% vs. 69% vs. 68% for lesions and 75% vs. 55% vs. 66% for organs, including kidneys, lacrimal glands, parotid glands, and submandibular glands. Conclusion: The proposed method enables reduction in acquisition time (by factors of 2, 4, or 8) while maintaining quantitative accuracy in clinical SPECT protocols by allowing for the collection of fewer projections. Importantly, the self-supervised nature of this NeRF-based approach eliminates the need for extensive training data, instead learning from each patient's projection data alone. The reduction in acquisition time is particularly relevant for imaging under low-count conditions and for protocols that require multiple-bed positions such as whole-body imaging.
3D Gaussian Splatting (3DGS) has shown a powerful capability for novel view synthesis due to its detailed expressive ability and highly efficient rendering speed. Unfortunately, creating relightable 3D assets and reconstructing faithful geometry with 3DGS is still problematic, particularly for reflective objects, as its discontinuous representation raises difficulties in constraining geometries. Volumetric signed distance field (SDF) methods provide robust geometry reconstruction, while the expensive ray marching hinders its real-time application and slows the training. Besides, these methods struggle to capture sharp geometric details. To this end, we propose to guide 3DGS and SDF bidirectionally in a complementary manner, including an SDF-aided Gaussian splatting for efficient optimization of the relighting model and a GS-guided SDF enhancement for high-quality geometry reconstruction. At the core of our SDF-aided Gaussian splatting is the mutual supervision of the depth and normal between blended Gaussians and SDF, which avoids the expensive volume rendering of SDF. Thanks to this mutual supervision, the learned blended Gaussians are well-constrained with a minimal time cost. As the Gaussians are rendered in a deferred shading mode, the alpha-blended Gaussians are smooth, while individual Gaussians may still be outliers, yielding floater artifacts. Therefore, we introduce an SDF-aware pruning strategy to remove Gaussian outliers located distant from the surface defined by SDF, avoiding floater issue. This way, our GS framework provides reasonable normal and achieves realistic relighting, while the mesh from depth is still problematic. Therefore, we design a GS-guided SDF refinement, which utilizes the blended normal from Gaussians to finetune SDF. With this enhancement, our method can further provide high-quality meshes for reflective objects at the cost of 17% extra training time.
Neural implicit functions have demonstrated significant importance in various areas such as computer vision, graphics. Their advantages include the ability to represent complex shapes and scenes with high fidelity, smooth interpolation capabilities, and continuous representations. Despite these benefits, the development and analysis of implicit functions have been limited by the lack of comprehensive datasets and the substantial computational resources required for their implementation and evaluation. To address these challenges, we introduce "Implicit-Zoo": a large-scale dataset requiring thousands of GPU training days designed to facilitate research and development in this field. Our dataset includes diverse 2D and 3D scenes, such as CIFAR-10, ImageNet-1K, and Cityscapes for 2D image tasks, and the OmniObject3D dataset for 3D vision tasks. We ensure high quality through strict checks, refining or filtering out low-quality data. Using Implicit-Zoo, we showcase two immediate benefits as it enables to: (1) learn token locations for transformer models; (2) directly regress 3D cameras poses of 2D images with respect to NeRF models. This in turn leads to an improved performance in all three task of image classification, semantic segmentation, and 3D pose regression, thereby unlocking new avenues for research.
Recently, Neural Radiance Fields (NeRF) achieved impressive results in novel view synthesis. Block-NeRF showed the capability of leveraging NeRF to build large city-scale models. For large-scale modeling, a mass of image data is necessary. Collecting images from specially designed data-collection vehicles can not support large-scale applications. How to acquire massive high-quality data remains an opening problem. Noting that the automotive industry has a huge amount of image data, crowd-sourcing is a convenient way for large-scale data collection. In this paper, we present a crowd-sourced framework, which utilizes substantial data captured by production vehicles to reconstruct the scene with the NeRF model. This approach solves the key problem of large-scale reconstruction, that is where the data comes from and how to use them. Firstly, the crowd-sourced massive data is filtered to remove redundancy and keep a balanced distribution in terms of time and space. Then a structure-from-motion module is performed to refine camera poses. Finally, images, as well as poses, are used to train the NeRF model in a certain block. We highlight that we present a comprehensive framework that integrates multiple modules, including data selection, sparse 3D reconstruction, sequence appearance embedding, depth supervision of ground surface, and occlusion completion. The complete system is capable of effectively processing and reconstructing high-quality 3D scenes from crowd-sourced data. Extensive quantitative and qualitative experiments were conducted to validate the performance of our system. Moreover, we proposed an application, named first-view navigation, which leveraged the NeRF model to generate 3D street view and guide the driver with a synthesized video.
Most current NeRF variants for satellites are designed for one specific scene and fall short of generalization to new geometry. Additionally, the RGB images require pan-sharpening as an independent preprocessing step. This paper introduces psPRF, a Planar Neural Radiance Field designed for paired low-resolution RGB (LR-RGB) and high-resolution panchromatic (HR-PAN) images from satellite sensors with Rational Polynomial Cameras (RPC). To capture the cross-modal prior from both of the LR-RGB and HR-PAN images, for the Unet-shaped architecture, we adapt the encoder with explicit spectral-to-spatial convolution (SSConv) to enhance the multimodal representation ability. To support the generalization ability of psRPF across scenes, we adopt projection loss to ensure strong geometry self-supervision. The proposed method is evaluated with the multi-scene WorldView-3 LR-RGB and HR-PAN pairs, and achieves state-of-the-art performance.
Embodied agents require robust navigation systems to operate in unstructured environments, making the robustness of Simultaneous Localization and Mapping (SLAM) models critical to embodied agent autonomy. While real-world datasets are invaluable, simulation-based benchmarks offer a scalable approach for robustness evaluations. However, the creation of a challenging and controllable noisy world with diverse perturbations remains under-explored. To this end, we propose a novel, customizable pipeline for noisy data synthesis, aimed at assessing the resilience of multi-modal SLAM models against various perturbations. The pipeline comprises a comprehensive taxonomy of sensor and motion perturbations for embodied multi-modal (specifically RGB-D) sensing, categorized by their sources and propagation order, allowing for procedural composition. We also provide a toolbox for synthesizing these perturbations, enabling the transformation of clean environments into challenging noisy simulations. Utilizing the pipeline, we instantiate the large-scale Noisy-Replica benchmark, which includes diverse perturbation types, to evaluate the risk tolerance of existing advanced RGB-D SLAM models. Our extensive analysis uncovers the susceptibilities of both neural (NeRF and Gaussian Splatting -based) and non-neural SLAM models to disturbances, despite their demonstrated accuracy in standard benchmarks. Our code is publicly available at https://github.com/Xiaohao-Xu/SLAM-under-Perturbation.
We propose a novel unsupervised method to learn the pose and part-segmentation of articulated objects with rigid parts. Given two observations of an object in different articulation states, our method learns the geometry and appearance of object parts by using an implicit model from the first observation, distils the part segmentation and articulation from the second observation while rendering the latter observation. Additionally, to tackle the complexities in the joint optimization of part segmentation and articulation, we propose a voxel grid-based initialization strategy and a decoupled optimization procedure. Compared to the prior unsupervised work, our model obtains significantly better performance, and generalizes to objects with multiple parts while it can be efficiently from few views for the latter observation.
Deep neural networks has been highly successful in data-intense computer vision applications, while such success relies heavily on the massive and clean data. In real-world scenarios, clean data sometimes is difficult to obtain. For example, in image classification and segmentation tasks, precise annotations of millions samples are generally very expensive and time-consuming. In 3D static scene reconstruction task, most NeRF related methods require the foundational assumption of the static scene (e.g. consistent lighting condition and persistent object positions), which is often violated in real-world scenarios. To address these problem, learning with noisy ground truth (LNGT) has emerged as an effective learning method and shows great potential. In this short survey, we propose a formal definition unify the analysis of LNGT LNGT in the context of different machine learning tasks (classification and regression). Based on this definition, we propose a novel taxonomy to classify the existing work according to the error decomposition with the fundamental definition of machine learning. Further, we provide in-depth analysis on memorization effect and insightful discussion about potential future research opportunities from 2D classification to 3D reconstruction, in the hope of providing guidance to follow-up research.
Neural Radiance Fields (NeRF) is an emerging technique to synthesize 3D objects from 2D images with a wide range of potential applications. However, rendering existing NeRF models is extremely computation intensive, making it challenging to support real-time interaction on mobile devices. In this paper, we take the first initiative to examine the state-of-the-art real-time NeRF rendering technique from a system perspective. We first define the entire working pipeline of the NeRF serving system. We then identify possible control knobs that are critical to the system from the communication, computation, and visual performance perspective. Furthermore, an extensive measurement study is conducted to reveal the effects of these control knobs on system performance. Our measurement results reveal that different control knobs contribute differently towards improving the system performance, with the mesh granularity being the most effective knob and the quantization being the least effective knob. In addition, diverse hardware device settings and network conditions have to be considered to fully unleash the benefit of operating under the appropriate knobs
The insertion of objects into a scene and relighting are commonly utilized applications in augmented reality (AR). Previous methods focused on inserting virtual objects using CAD models or real objects from single-view images, resulting in highly limited AR application scenarios. We propose a novel NeRF-based pipeline for inserting object NeRFs into scene NeRFs, enabling novel view synthesis and realistic relighting, supporting physical interactions like casting shadows onto each other, from two sets of images depicting the object and scene. The lighting environment is in a hybrid representation of Spherical Harmonics and Spherical Gaussians, representing both high- and low-frequency lighting components very well, and supporting non-Lambertian surfaces. Specifically, we leverage the benefits of volume rendering and introduce an innovative approach for efficient shadow rendering by comparing the depth maps between the camera view and the light source view and generating vivid soft shadows. The proposed method achieves realistic relighting effects in extensive experimental evaluations.
Event cameras, known for their high dynamic range, absence of motion blur, and low energy usage, have recently found a wide range of applications thanks to these attributes. In the past few years, the field of event-based 3D reconstruction saw remarkable progress, with the Neural Radiance Field (NeRF) based approach demonstrating photorealistic view synthesis results. However, the volume rendering paradigm of NeRF necessitates extensive training and rendering times. In this paper, we introduce Event Enhanced Gaussian Splatting (E2GS), a novel method that incorporates event data into Gaussian Splatting, which has recently made significant advances in the field of novel view synthesis. Our E2GS effectively utilizes both blurry images and event data, significantly improving image deblurring and producing high-quality novel view synthesis. Our comprehensive experiments on both synthetic and real-world datasets demonstrate our E2GS can generate visually appealing renderings while offering faster training and rendering speed (140 FPS). Our code is available at https://github.com/deguchihiroyuki/E2GS.
Object Pose Estimation is a crucial component in robotic grasping and augmented reality. Learning based approaches typically require training data from a highly accurate CAD model or labeled training data acquired using a complex setup. We address this by learning to estimate pose from weakly labeled data without a known CAD model. We propose to use a NeRF to learn object shape implicitly which is later used to learn view-invariant features in conjunction with CNN using a contrastive loss. While NeRF helps in learning features that are view-consistent, CNN ensures that the learned features respect symmetry. During inference, CNN is used to predict view-invariant features which can be used to establish correspondences with the implicit 3d model in NeRF. The correspondences are then used to estimate the pose in the reference frame of NeRF. Our approach can also handle symmetric objects unlike other approaches using a similar training setup. Specifically, we learn viewpoint invariant, discriminative features using NeRF which are later used for pose estimation. We evaluated our approach on LM, LM-Occlusion, and T-Less dataset and achieved benchmark accuracy despite using weakly labeled data.
Neural Radiance Fields (NeRF) have shown remarkable success in representing 3D scenes and generating novel views. However, they often struggle with aliasing artifacts, especially when rendering images from different camera distances from the training views. To address the issue, Mip-NeRF proposed using volumetric frustums to render a pixel and suggested integrated positional encoding (IPE). While effective, this approach requires long training times due to its reliance on MLP architecture. In this work, we propose a novel anti-aliasing technique that utilizes grid-based representations, usually showing significantly faster training time. In addition, we exploit frequency-domain representation to handle the aliasing problem inspired by the sampling theorem. The proposed method, FreqMipAA, utilizes scale-specific low-pass filtering (LPF) and learnable frequency masks. Scale-specific low-pass filters (LPF) prevent aliasing and prioritize important image details, and learnable masks effectively remove problematic high-frequency elements while retaining essential information. By employing a scale-specific LPF and trainable masks, FreqMipAA can effectively eliminate the aliasing factor while retaining important details. We validated the proposed technique by incorporating it into a widely used grid-based method. The experimental results have shown that the FreqMipAA effectively resolved the aliasing issues and achieved state-of-the-art results in the multi-scale Blender dataset. Our code is available at https://github.com/yi0109/FreqMipAA .