Suryansh Kumar

Ce Liu, Suryansh Kumar, Shuhang Gu et al. · microsoft-research

Neural-network-based single image depth prediction (SIDP) is a challenging task where the goal is to predict the scene's per-pixel depth at test time. Since the problem, by definition, is ill-posed, the fundamental goal is to come up with an approach that can reliably model the scene depth from a set of training examples. In the pursuit of perfect depth estimation, most existing state-of-the-art learning techniques predict a single scalar depth value per-pixel. Yet, it is well-known that the trained model has accuracy limits and can predict imprecise depth. Therefore, an SIDP approach must be mindful of the expected depth variations in the model's prediction at test time. Accordingly, we introduce an approach that performs continuous modeling of per-pixel depth, where we can predict and reason about the per-pixel depth and its distribution. To this end, we model per-pixel scene depth using a multivariate Gaussian distribution. Moreover, contrary to the existing uncertainty modeling methods -- in the same spirit, where per-pixel depth is assumed to be independent, we introduce per-pixel covariance modeling that encodes its depth dependency w.r.t all the scene points. Unfortunately, per-pixel depth covariance modeling leads to a computationally expensive continuous loss function, which we solve efficiently using the learned low-rank approximation of the overall covariance matrix. Notably, when tested on benchmark datasets such as KITTI, NYU, and SUN-RGB-D, the SIDP model obtained by optimizing our loss function shows state-of-the-art results. Our method's accuracy (named MG) is among the top on the KITTI depth-prediction benchmark leaderboard.

Erik Sandström, Martin R. Oswald, Suryansh Kumar et al.

Many hand-held or mixed reality devices are used with a single sensor for 3D reconstruction, although they often comprise multiple sensors. Multi-sensor depth fusion is able to substantially improve the robustness and accuracy of 3D reconstruction methods, but existing techniques are not robust enough to handle sensors which operate with diverse value ranges as well as noise and outlier statistics. To this end, we introduce SenFuNet, a depth fusion approach that learns sensor-specific noise and outlier statistics and combines the data streams of depth frames from different sensors in an online fashion. Our method fuses multi-sensor depth streams regardless of time synchronization and calibration and generalizes well with little training data. We conduct experiments with various sensor combinations on the real-world CoRBS and Scene3D datasets, as well as the Replica dataset. Experiments demonstrate that our fusion strategy outperforms traditional and recent online depth fusion approaches. In addition, the combination of multiple sensors yields more robust outlier handling and more precise surface reconstruction than the use of a single sensor. The source code and data are available at https://github.com/tfy14esa/SenFuNet.

Ce Liu, Suryansh Kumar, Shuhang Gu et al.

We introduce VA-DepthNet, a simple, effective, and accurate deep neural network approach for the single-image depth prediction (SIDP) problem. The proposed approach advocates using classical first-order variational constraints for this problem. While state-of-the-art deep neural network methods for SIDP learn the scene depth from images in a supervised setting, they often overlook the invaluable invariances and priors in the rigid scene space, such as the regularity of the scene. The paper's main contribution is to reveal the benefit of classical and well-founded variational constraints in the neural network design for the SIDP task. It is shown that imposing first-order variational constraints in the scene space together with popular encoder-decoder-based network architecture design provides excellent results for the supervised SIDP task. The imposed first-order variational constraint makes the network aware of the depth gradient in the scene space, i.e., regularity. The paper demonstrates the usefulness of the proposed approach via extensive evaluation and ablation analysis over several benchmark datasets, such as KITTI, NYU Depth V2, and SUN RGB-D. The VA-DepthNet at test time shows considerable improvements in depth prediction accuracy compared to the prior art and is accurate also at high-frequency regions in the scene space. At the time of writing this paper, our method -- labeled as VA-DepthNet, when tested on the KITTI depth-prediction evaluation set benchmarks, shows state-of-the-art results, and is the top-performing published approach.

Yasaman Haghighi, Suryansh Kumar, Jean-Philippe Thiran et al.

Visual Simultaneous Localization and Mapping (vSLAM) is a widely used technique in robotics and computer vision that enables a robot to create a map of an unfamiliar environment using a camera sensor while simultaneously tracking its position over time. In this paper, we propose a novel RGBD vSLAM algorithm that can learn a memory-efficient, dense 3D geometry, and semantic segmentation of an indoor scene in an online manner. Our pipeline combines classical 3D vision-based tracking and loop closing with neural fields-based mapping. The mapping network learns the SDF of the scene as well as RGB, depth, and semantic maps of any novel view using only a set of keyframes. Additionally, we extend our pipeline to large scenes by using multiple local mapping networks. Extensive experiments on well-known benchmark datasets confirm that our approach provides robust tracking, mapping, and semantic labeling even with noisy, sparse, or no input depth. Overall, our proposed algorithm can greatly enhance scene perception and assist with a range of robot control problems.

Soomin Lee, Le Chen, Jiahao Wang et al.

In this paper, we tackle the problem of active robotic 3D reconstruction of an object. In particular, we study how a mobile robot with an arm-held camera can select a favorable number of views to recover an object's 3D shape efficiently. Contrary to the existing solution to this problem, we leverage the popular neural radiance fields-based object representation, which has recently shown impressive results for various computer vision tasks. However, it is not straightforward to directly reason about an object's explicit 3D geometric details using such a representation, making the next-best-view selection problem for dense 3D reconstruction challenging. This paper introduces a ray-based volumetric uncertainty estimator, which computes the entropy of the weight distribution of the color samples along each ray of the object's implicit neural representation. We show that it is possible to infer the uncertainty of the underlying 3D geometry given a novel view with the proposed estimator. We then present a next-best-view selection policy guided by the ray-based volumetric uncertainty in neural radiance fields-based representations. Encouraging experimental results on synthetic and real-world data suggest that the approach presented in this paper can enable a new research direction of using an implicit 3D object representation for the next-best-view problem in robot vision applications, distinguishing our approach from the existing approaches that rely on explicit 3D geometric modeling.

Berk Kaya, Suryansh Kumar, Carlos Oliveira et al.

Multi-view photometric stereo (MVPS) is a preferred method for detailed and precise 3D acquisition of an object from images. Although popular methods for MVPS can provide outstanding results, they are often complex to execute and limited to isotropic material objects. To address such limitations, we present a simple, practical approach to MVPS, which works well for isotropic as well as other object material types such as anisotropic and glossy. The proposed approach in this paper exploits the benefit of uncertainty modeling in a deep neural network for a reliable fusion of photometric stereo (PS) and multi-view stereo (MVS) network predictions. Yet, contrary to the recently proposed state-of-the-art, we introduce neural volume rendering methodology for a trustworthy fusion of MVS and PS measurements. The advantage of introducing neural volume rendering is that it helps in the reliable modeling of objects with diverse material types, where existing MVS methods, PS methods, or both may fail. Furthermore, it allows us to work on neural 3D shape representation, which has recently shown outstanding results for many geometric processing tasks. Our suggested new loss function aims to fits the zero level set of the implicit neural function using the most certain MVS and PS network predictions coupled with weighted neural volume rendering cost. The proposed approach shows state-of-the-art results when tested extensively on several benchmark datasets.

Suryansh Kumar, Luc Van Gool

This paper advocates the use of organic priors in classical non-rigid structure from motion (NRSfM). By organic priors, we mean invaluable intermediate prior information intrinsic to the NRSfM matrix factorization theory. It is shown that such priors reside in the factorized matrices, and quite surprisingly, existing methods generally disregard them. The paper's main contribution is to put forward a simple, methodical, and practical method that can effectively exploit such organic priors to solve NRSfM. The proposed method does not make assumptions other than the popular one on the low-rank shape and offers a reliable solution to NRSfM under orthographic projection. Our work reveals that the accessibility of organic priors is independent of the camera motion and shape deformation type. Besides that, the paper provides insights into the NRSfM factorization -- both in terms of shape and motion -- and is the first approach to show the benefit of single rotation averaging for NRSfM. Furthermore, we outline how to effectively recover motion and non-rigid 3D shape using the proposed organic prior based approach and demonstrate results that outperform prior-free NRSfM performance by a significant margin. Finally, we present the benefits of our method via extensive experiments and evaluations on several benchmark datasets.

Tobias Fischer, Yung-Hsu Yang, Suryansh Kumar et al.

To track the 3D locations and trajectories of the other traffic participants at any given time, modern autonomous vehicles are equipped with multiple cameras that cover the vehicle's full surroundings. Yet, camera-based 3D object tracking methods prioritize optimizing the single-camera setup and resort to post-hoc fusion in a multi-camera setup. In this paper, we propose a method for panoramic 3D object tracking, called CC-3DT, that associates and models object trajectories both temporally and across views, and improves the overall tracking consistency. In particular, our method fuses 3D detections from multiple cameras before association, reducing identity switches significantly and improving motion modeling. Our experiments on large-scale driving datasets show that fusion before association leads to a large margin of improvement over post-hoc fusion. We set a new state-of-the-art with 12.6% improvement in average multi-object tracking accuracy (AMOTA) among all camera-based methods on the competitive NuScenes 3D tracking benchmark, outperforming previously published methods by 6.5% in AMOTA with the same 3D detector.

Ce Liu, Suryansh Kumar, Shuhang Gu et al.

We introduce Stereo Risk, a new deep-learning approach to solve the classical stereo-matching problem in computer vision. As it is well-known that stereo matching boils down to a per-pixel disparity estimation problem, the popular state-of-the-art stereo-matching approaches widely rely on regressing the scene disparity values, yet via discretization of scene disparity values. Such discretization often fails to capture the nuanced, continuous nature of scene depth. Stereo Risk departs from the conventional discretization approach by formulating the scene disparity as an optimal solution to a continuous risk minimization problem, hence the name "stereo risk". We demonstrate that $L^1$ minimization of the proposed continuous risk function enhances stereo-matching performance for deep networks, particularly for disparities with multi-modal probability distributions. Furthermore, to enable the end-to-end network training of the non-differentiable $L^1$ risk optimization, we exploited the implicit function theorem, ensuring a fully differentiable network. A comprehensive analysis demonstrates our method's theoretical soundness and superior performance over the state-of-the-art methods across various benchmark datasets, including KITTI 2012, KITTI 2015, ETH3D, SceneFlow, and Middlebury 2014.

Han Yao Choong, Suryansh Kumar, Luc Van Gool

This paper proposes a quantum computing-based algorithm to solve the single image super-resolution (SISR) problem. One of the well-known classical approaches for SISR relies on the well-established patch-wise sparse modeling of the problem. Yet, this field's current state of affairs is that deep neural networks (DNNs) have demonstrated far superior results than traditional approaches. Nevertheless, quantum computing is expected to become increasingly prominent for machine learning problems soon. As a result, in this work, we take the privilege to perform an early exploration of applying a quantum computing algorithm to this important image enhancement problem, i.e., SISR. Among the two paradigms of quantum computing, namely universal gate quantum computing and adiabatic quantum computing (AQC), the latter has been successfully applied to practical computer vision problems, in which quantum parallelism has been exploited to solve combinatorial optimization efficiently. This work demonstrates formulating quantum SISR as a sparse coding optimization problem, which is solved using quantum annealers accessed via the D-Wave Leap platform. The proposed AQC-based algorithm is demonstrated to achieve improved speed-up over a classical analog while maintaining comparable SISR accuracy.

Nishant Jain, Suryansh Kumar, Luc Van Gool

We introduce an approach to enhance the novel view synthesis from images taken from a freely moving camera. The introduced approach focuses on outdoor scenes where recovering accurate geometric scaffold and camera pose is challenging, leading to inferior results using the state-of-the-art stable view synthesis (SVS) method. SVS and related methods fail for outdoor scenes primarily due to (i) over-relying on the multiview stereo (MVS) for geometric scaffold recovery and (ii) assuming COLMAP computed camera poses as the best possible estimates, despite it being well-studied that MVS 3D reconstruction accuracy is limited to scene disparity and camera-pose accuracy is sensitive to key-point correspondence selection. This work proposes a principled way to enhance novel view synthesis solutions drawing inspiration from the basics of multiple view geometry. By leveraging the complementary behavior of MVS and monocular depth, we arrive at a better scene depth per view for nearby and far points, respectively. Moreover, our approach jointly refines camera poses with image-based rendering via multiple rotation averaging graph optimization. The recovered scene depth and the camera-pose help better view-dependent on-surface feature aggregation of the entire scene. Extensive evaluation of our approach on the popular benchmark dataset, such as Tanks and Temples, shows substantial improvement in view synthesis results compared to the prior art. For instance, our method shows 1.5 dB of PSNR improvement on the Tank and Temples. Similar statistics are observed when tested on other benchmark datasets such as FVS, Mip-NeRF 360, and DTU.

Nishant Jain, Suryansh Kumar, Luc Van Gool

Neural Radiance Fields (NeRF) recently emerged as a new paradigm for object representation from multi-view (MV) images. Yet, it cannot handle multi-scale (MS) images and camera pose estimation errors, which generally is the case with multi-view images captured from a day-to-day commodity camera. Although recently proposed Mip-NeRF could handle multi-scale imaging problems with NeRF, it cannot handle camera pose estimation error. On the other hand, the newly proposed BARF can solve the camera pose problem with NeRF but fails if the images are multi-scale in nature. This paper presents a robust multi-scale neural radiance fields representation approach to simultaneously overcome both real-world imaging issues. Our method handles multi-scale imaging effects and camera-pose estimation problems with NeRF-inspired approaches by leveraging the fundamentals of scene rigidity. To reduce unpleasant aliasing artifacts due to multi-scale images in the ray space, we leverage Mip-NeRF multi-scale representation. For joint estimation of robust camera pose, we propose graph-neural network-based multiple motion averaging in the neural volume rendering framework. We demonstrate, with examples, that for an accurate neural representation of an object from day-to-day acquired multi-view images, it is crucial to have precise camera-pose estimates. Without considering robustness measures in the camera pose estimation, modeling for multi-scale aliasing artifacts via conical frustum can be counterproductive. We present extensive experiments on the benchmark datasets to demonstrate that our approach provides better results than the recent NeRF-inspired approaches for such realistic settings.

Weirong Chen, Suryansh Kumar, Fisher Yu

This work introduces an effective and practical solution to the dense two-view structure from motion (SfM) problem. One vital question addressed is how to mindfully use per-pixel optical flow correspondence between two frames for accurate pose estimation -- as perfect per-pixel correspondence between two images is difficult, if not impossible, to establish. With the carefully estimated camera pose and predicted per-pixel optical flow correspondences, a dense depth of the scene is computed. Later, an iterative refinement procedure is introduced to further improve optical flow matching confidence, camera pose, and depth, exploiting their inherent dependency in rigid SfM. The fundamental idea presented is to benefit from per-pixel uncertainty in the optical flow estimation and provide robustness to the dense SfM system via an online refinement. Concretely, we introduce our uncertainty-driven Dense Two-View SfM pipeline (DTV-SfM), consisting of an uncertainty-aware dense optical flow estimation approach that provides per-pixel correspondence with their confidence score of matching; a weighted dense bundle adjustment formulation that depends on optical flow uncertainty and bidirectional optical flow consistency to refine both pose and depth; a depth estimation network that considers its consistency with the estimated poses and optical flow respecting epipolar constraint. Extensive experiments show that the proposed approach achieves remarkable depth accuracy and state-of-the-art camera pose results superseding SuperPoint and SuperGlue accuracy when tested on benchmark datasets such as DeMoN, YFCC100M, and ScanNet. Code and more materials are available at http://vis.xyz/pub/dtv-sfm.

Nishant Jain, Suryansh Kumar, Luc Van Gool

We introduce an improved solution to the neural image-based rendering problem in computer vision. Given a set of images taken from a freely moving camera at train time, the proposed approach could synthesize a realistic image of the scene from a novel viewpoint at test time. The key ideas presented in this paper are (i) Recovering accurate camera parameters via a robust pipeline from unposed day-to-day images is equally crucial in neural novel view synthesis problem; (ii) It is rather more practical to model object's content at different resolutions since dramatic camera motion is highly likely in day-to-day unposed images. To incorporate the key ideas, we leverage the fundamentals of scene rigidity, multi-scale neural scene representation, and single-image depth prediction. Concretely, the proposed approach makes the camera parameters as learnable in a neural fields-based modeling framework. By assuming per view depth prediction is given up to scale, we constrain the relative pose between successive frames. From the relative poses, absolute camera pose estimation is modeled via a graph-neural network-based multiple motion averaging within the multi-scale neural-fields network, leading to a single loss function. Optimizing the introduced loss function provides camera intrinsic, extrinsic, and image rendering from unposed images. We demonstrate, with examples, that for a unified framework to accurately model multiscale neural scene representation from day-to-day acquired unposed multi-view images, it is equally essential to have precise camera-pose estimates within the scene representation framework. Without considering robustness measures in the camera pose estimation pipeline, modeling for multi-scale aliasing artifacts can be counterproductive. We present extensive experiments on several benchmark datasets to demonstrate the suitability of our approach.

Yeeun Park, Miqdad Naduthodi, Suryansh Kumar

Marker-based motion capture (MoCap) systems have long been the gold standard for accurate 4D human modeling, yet their reliance on specialized hardware and markers limits scalability and real-world deployment. Advancing reliable markerless 4D human motion capture requires datasets that reflect the complexity of real-world human interactions. Yet, existing benchmarks often lack realistic multi-person dynamics, severe occlusions, and challenging interaction patterns, leading to a persistent domain gap. In this work, we present a new dataset and evaluation for complex 4D markerless human motion capture. Our proposed MoCap dataset captures both single and multi-person scenarios with intricate motions, frequent inter-person occlusions, rapid position exchanges between similarly dressed subjects, and varying subject distances. It includes synchronized multi-view RGB and depth sequences, accurate camera calibration, ground-truth 3D motion capture from a Vicon system, and corresponding SMPL/SMPL-X parameters. This setup ensures precise alignment between visual observations and motion ground truth. Benchmarking state-of-the-art markerless MoCap models reveals substantial performance degradation under these realistic conditions, highlighting limitations of current approaches. We further demonstrate that targeted fine-tuning improves generalization, validating the dataset's realism and value for model development. Our evaluation exposes critical gaps in existing models and provides a rigorous foundation for advancing robust markerless 4D human motion capture.

Danilo Dordevic, Suryansh Kumar

We introduce the Evidential Transformer, an uncertainty-driven transformer model for improved and robust image retrieval. In this paper, we make several contributions to content-based image retrieval (CBIR). We incorporate probabilistic methods into image retrieval, achieving robust and reliable results, with evidential classification surpassing traditional training based on multiclass classification as a baseline for deep metric learning. Furthermore, we improve the state-of-the-art retrieval results on several datasets by leveraging the Global Context Vision Transformer (GC ViT) architecture. Our experimental results consistently demonstrate the reliability of our approach, setting a new benchmark in CBIR in all test settings on the Stanford Online Products (SOP) and CUB-200-2011 datasets.

Yunxiao Zhang, Suryansh Kumar

Retrospective novel view synthesis (NVS) of dynamic scenes is fundamental to applications such as sports. Recent dynamic 3D Gaussian Splatting (3DGS) approaches introduce temporally coupled formulations to enforce motion coherence across time. In this paper, we argue that, in a synchronized multi-view (MV) setting typical of sports, the dynamic scene at each time step is already strongly geometrically constrained. We posit that the availability of calibrated, synchronized viewpoints provides sufficient spatial consistency, and therefore, explicit temporal coupling, or complex multi-body constraints seems unnecessary for retrospective NVS. To this end, we propose an approach tailored for synchronized MV dynamic scene. By initializing the SfM-derived point cloud at the start time and propagating optimized Gaussians over time, we show that efficient retrospective NVS can be achieved without imposing a temporal deformation constraint. Complementing our methodological contribution, we introduce a Dynamic MV dataset framework built on Blender for reproducible NeRF and 3DGS research. The framework generates high-quality, synchronized camera rigs and exports training-ready datasets in standard formats, eliminating inconsistencies in coordinate conventions and data pipelines. Using the framework, we construct a dynamic benchmark suite and evaluate representative NeRF and 3DGS approaches under controlled conditions. Together, we show that, under a synchronized MV setup, efficient retrospective dynamic scene NVS can be achieved using 3DGS. At the same time, the dataset-generation framework enables reproducible and principled benchmarking of dynamic NVS methods.

Praroop Chanda, Suryansh Kumar

Recent progress in dense optical flow has been driven by increasingly complex architectures and multi-step refinement for test-time scaling. While these approaches achieve strong benchmark performance, they also require substantial computation during inference. This raises a fundamental question: Is scaling test-time computation the only way to improve dense optical flow accuracy? We argue that it is not. Instead, powerful visual semantic and geometric priors encoded in modern foundation models can reduce, if not overcome, the need for computationally expensive iterative refinement at test-time. In this paper, we present a framework that estimates dense optical flow in a single forward pass, leveraging pretrained foundation representations, while avoiding iterative refinement and additional inference-time computation, thus offering an alternative to test-time scaling. Our method extracts visual semantic features from a frozen DINO-v2 backbone and combines them with geometric cues from a monocular depth foundation model. We fuse these complementary priors into a unified representation and apply a global matching formulation to estimate dense correspondences without recurrent updates or test-time optimization. Despite avoiding iterative refinement, our approach achieves strong cross-dataset generalization across challenging benchmarks. On Sintel Final, we obtain 2.81 EPE without refinement, significantly improving over state-of-the-art (SOTA) SEA-RAFT under comparable training conditions and outperforming RAFT, GMFlow (without refinement), and recent FlowSeek in the same setting. These results suggest that strong foundation priors can substitute for test-time scaling, offering a computationally efficient alternative to refinement-heavy pipelines.

Yunxiao Zhang, William Stone, Suryansh Kumar

Camera virtualization -- an emerging solution to novel view synthesis -- holds transformative potential for visual entertainment, live performances, and sports broadcasting by enabling the generation of photorealistic images from novel viewpoints using images from a limited set of calibrated multiple static physical cameras. Despite recent advances, achieving spatially and temporally coherent and photorealistic rendering of dynamic scenes with efficient time-archival capabilities, particularly in fast-paced sports and stage performances, remains challenging for existing approaches. Recent methods based on 3D Gaussian Splatting (3DGS) for dynamic scenes could offer real-time view-synthesis results. Yet, they are hindered by their dependence on accurate 3D point clouds from the structure-from-motion method and their inability to handle large, non-rigid, rapid motions of different subjects (e.g., flips, jumps, articulations, sudden player-to-player transitions). Moreover, independent motions of multiple subjects can break the Gaussian-tracking assumptions commonly used in 4DGS, ST-GS, and other dynamic splatting variants. This paper advocates reconsidering a neural volume rendering formulation for camera virtualization and efficient time-archival capabilities, making it useful for sports broadcasting and related applications. By modeling a dynamic scene as rigid transformations across multiple synchronized camera views at a given time, our method performs neural representation learning, providing enhanced visual rendering quality at test time. A key contribution of our approach is its support for time-archival, i.e., users can revisit any past temporal instance of a dynamic scene and can perform novel view synthesis, enabling retrospective rendering for replay, analysis, and archival of live events, a functionality absent in existing neural rendering approaches and novel view synthesis...

Ishaan Rawal, Suryansh Kumar

Text-conditioned diffusion models have emerged as powerful tools for high-quality video generation. However, enabling Interactive Video Generation (IVG), where users control motion elements such as object trajectory, remains challenging. Recent training-free approaches introduce attention masking to guide trajectory, but this often degrades perceptual quality. We identify two key failure modes in these methods, both of which we interpret as domain shift problems, and propose solutions inspired by domain adaptation. First, we attribute the perceptual degradation to internal covariate shift induced by attention masking, as pretrained models are not trained to handle masked attention. To address this, we propose mask normalization, a pre-normalization layer designed to mitigate this shift via distribution matching. Second, we address initialization gap, where the randomly sampled initial noise does not align with IVG conditioning, by introducing a temporal intrinsic diffusion prior that enforces spatio-temporal consistency at each denoising step. Extensive qualitative and quantitative evaluations demonstrate that mask normalization and temporal intrinsic denoising improve both perceptual quality and trajectory control over the existing state-of-the-art IVG techniques.

Fikrican Özgür, René Zurbrügg, Suryansh Kumar

Hindsight Experience Replay (HER) is widely regarded as the state-of-the-art algorithm for achieving sample-efficient multi-goal reinforcement learning (RL) in robotic manipulation tasks with binary rewards. HER facilitates learning from failed attempts by replaying trajectories with redefined goals. However, it relies on a heuristic-based replay method that lacks a principled framework. To address this limitation, we introduce a novel replay strategy, "Next-Future", which focuses on rewarding single-step transitions. This approach significantly enhances sample efficiency and accuracy in learning multi-goal Markov decision processes (MDPs), particularly under stringent accuracy requirements -- a critical aspect for performing complex and precise robotic-arm tasks. We demonstrate the efficacy of our method by highlighting how single-step learning enables improved value approximation within the multi-goal RL framework. The performance of the proposed replay strategy is evaluated across eight challenging robotic manipulation tasks, using ten random seeds for training. Our results indicate substantial improvements in sample efficiency for seven out of eight tasks and higher success rates in six tasks. Furthermore, real-world experiments validate the practical feasibility of the learned policies, demonstrating the potential of "Next-Future" in solving complex robotic-arm tasks.

Suryansh Kumar

Multi-View Photometric Stereo (MVPS) is a popular method for fine-detailed 3D acquisition of an object from images. Despite its outstanding results on diverse material objects, a typical MVPS experimental setup requires a well-calibrated light source and a monocular camera installed on an immovable base. This restricts the use of MVPS on a movable platform, limiting us from taking MVPS benefits in 3D acquisition for mobile robotics applications. To this end, we introduce a new mobile robotic system for MVPS. While the proposed system brings advantages, it introduces additional algorithmic challenges. Addressing them, in this paper, we further propose an incremental approach for mobile robotic MVPS. Our approach leverages a supervised learning setup to predict per-view surface normal, object depth, and per-pixel uncertainty in model-predicted results. A refined depth map per view is obtained by solving an MVPS-driven optimization problem proposed in this paper. Later, we fuse the refined depth map while tracking the camera pose w.r.t the reference frame to recover globally consistent object 3D geometry. Experimental results show the advantages of our robotic system and algorithm, featuring the local high-frequency surface detail recovery with globally consistent object shape. Our work is beyond any MVPS system yet presented, providing encouraging results on objects with unknown reflectance properties using fewer frames without a tiring calibration and installation process, enabling computationally efficient robotic automation approach to photogrammetry. The proposed approach is nearly 100 times computationally faster than the state-of-the-art MVPS methods such as [1, 2] while maintaining the similar results when tested on subjects taken from the benchmark DiLiGenT MV dataset [3].

René Zurbrügg, Yifan Liu, Francis Engelmann et al.

Accurate grasping is the key to several robotic tasks including assembly and household robotics. Executing a successful grasp in a cluttered environment requires multiple levels of scene understanding: First, the robot needs to analyze the geometric properties of individual objects to find feasible grasps. These grasps need to be compliant with the local object geometry. Second, for each proposed grasp, the robot needs to reason about the interactions with other objects in the scene. Finally, the robot must compute a collision-free grasp trajectory while taking into account the geometry of the target object. Most grasp detection algorithms directly predict grasp poses in a monolithic fashion, which does not capture the composability of the environment. In this paper, we introduce an end-to-end architecture for object-centric grasping. The method uses pointcloud data from a single arbitrary viewing direction as an input and generates an instance-centric representation for each partially observed object in the scene. This representation is further used for object reconstruction and grasp detection in cluttered table-top scenes. We show the effectiveness of the proposed method by extensively evaluating it against state-of-the-art methods on synthetic datasets, indicating superior performance for grasping and reconstruction. Additionally, we demonstrate real-world applicability by decluttering scenes with varying numbers of objects.

Junting Chen, Guohao Li, Suryansh Kumar et al.

Object goal navigation is an important problem in Embodied AI that involves guiding the agent to navigate to an instance of the object category in an unknown environment -- typically an indoor scene. Unfortunately, current state-of-the-art methods for this problem rely heavily on data-driven approaches, \eg, end-to-end reinforcement learning, imitation learning, and others. Moreover, such methods are typically costly to train and difficult to debug, leading to a lack of transferability and explainability. Inspired by recent successes in combining classical and learning methods, we present a modular and training-free solution, which embraces more classic approaches, to tackle the object goal navigation problem. Our method builds a structured scene representation based on the classic visual simultaneous localization and mapping (V-SLAM) framework. We then inject semantics into geometric-based frontier exploration to reason about promising areas to search for a goal object. Our structured scene representation comprises a 2D occupancy map, semantic point cloud, and spatial scene graph. Our method propagates semantics on the scene graphs based on language priors and scene statistics to introduce semantic knowledge to the geometric frontiers. With injected semantic priors, the agent can reason about the most promising frontier to explore. The proposed pipeline shows strong experimental performance for object goal navigation on the Gibson benchmark dataset, outperforming the previous state-of-the-art. We also perform comprehensive ablation studies to identify the current bottleneck in the object navigation task.

Berk Kaya, Suryansh Kumar, Carlos Oliveira et al.

This paper presents a simple and effective solution to the longstanding classical multi-view photometric stereo (MVPS) problem. It is well-known that photometric stereo (PS) is excellent at recovering high-frequency surface details, whereas multi-view stereo (MVS) can help remove the low-frequency distortion due to PS and retain the global geometry of the shape. This paper proposes an approach that can effectively utilize such complementary strengths of PS and MVS. Our key idea is to combine them suitably while considering the per-pixel uncertainty of their estimates. To this end, we estimate per-pixel surface normals and depth using an uncertainty-aware deep-PS network and deep-MVS network, respectively. Uncertainty modeling helps select reliable surface normal and depth estimates at each pixel which then act as a true representative of the dense surface geometry. At each pixel, our approach either selects or discards deep-PS and deep-MVS network prediction depending on the prediction uncertainty measure. For dense, detailed, and precise inference of the object's surface profile, we propose to learn the implicit neural shape representation via a multilayer perceptron (MLP). Our approach encourages the MLP to converge to a natural zero-level set surface using the confident prediction from deep-PS and deep-MVS networks, providing superior dense surface reconstruction. Extensive experiments on the DiLiGenT-MV benchmark dataset show that our method provides high-quality shape recovery with a much lower memory footprint while outperforming almost all of the existing approaches.

Francesco Sarno, Suryansh Kumar, Berk Kaya et al.

We present an automated machine learning approach for uncalibrated photometric stereo (PS). Our work aims at discovering lightweight and computationally efficient PS neural networks with excellent surface normal accuracy. Unlike previous uncalibrated deep PS networks, which are handcrafted and carefully tuned, we leverage differentiable neural architecture search (NAS) strategy to find uncalibrated PS architecture automatically. We begin by defining a discrete search space for a light calibration network and a normal estimation network, respectively. We then perform a continuous relaxation of this search space and present a gradient-based optimization strategy to find an efficient light calibration and normal estimation network. Directly applying the NAS methodology to uncalibrated PS is not straightforward as certain task-specific constraints must be satisfied, which we impose explicitly. Moreover, we search for and train the two networks separately to account for the Generalized Bas-Relief (GBR) ambiguity. Extensive experiments on the DiLiGenT dataset show that the automatically searched neural architectures performance compares favorably with the state-of-the-art uncalibrated PS methods while having a lower memory footprint.

Berk Kaya, Suryansh Kumar, Francesco Sarno et al.

We present a modern solution to the multi-view photometric stereo problem (MVPS). Our work suitably exploits the image formation model in a MVPS experimental setup to recover the dense 3D reconstruction of an object from images. We procure the surface orientation using a photometric stereo (PS) image formation model and blend it with a multi-view neural radiance field representation to recover the object's surface geometry. Contrary to the previous multi-staged framework to MVPS, where the position, iso-depth contours, or orientation measurements are estimated independently and then fused later, our method is simple to implement and realize. Our method performs neural rendering of multi-view images while utilizing surface normals estimated by a deep photometric stereo network. We render the MVPS images by considering the object's surface normals for each 3D sample point along the viewing direction rather than explicitly using the density gradient in the volume space via 3D occupancy information. We optimize the proposed neural radiance field representation for the MVPS setup efficiently using a fully connected deep network to recover the 3D geometry of an object. Extensive evaluation on the DiLiGenT-MV benchmark dataset shows that our method performs better than the approaches that perform only PS or only multi-view stereo (MVS) and provides comparable results against the state-of-the-art multi-stage fusion methods.

Davide Menini, Suryansh Kumar, Martin R. Oswald et al.

This paper presents a real-time online vision framework to jointly recover an indoor scene's 3D structure and semantic label. Given noisy depth maps, a camera trajectory, and 2D semantic labels at train time, the proposed deep neural network based approach learns to fuse the depth over frames with suitable semantic labels in the scene space. Our approach exploits the joint volumetric representation of the depth and semantics in the scene feature space to solve this task. For a compelling online fusion of the semantic labels and geometry in real-time, we introduce an efficient vortex pooling block while dropping the use of routing network in online depth fusion to preserve high-frequency surface details. We show that the context information provided by the semantics of the scene helps the depth fusion network learn noise-resistant features. Not only that, it helps overcome the shortcomings of the current online depth fusion method in dealing with thin object structures, thickening artifacts, and false surfaces. Experimental evaluation on the Replica dataset shows that our approach can perform depth fusion at 37 and 10 frames per second with an average reconstruction F-score of 88% and 91%, respectively, depending on the depth map resolution. Moreover, our model shows an average IoU score of 0.515 on the ScanNet 3D semantic benchmark leaderboard.

Rhea Sanjay Sukthanker, Zhiwu Huang, Suryansh Kumar et al.

Flow-based generative models have shown an excellent ability to explicitly learn the probability density function of data via a sequence of invertible transformations. Yet, learning attentions in generative flows remains understudied, while it has made breakthroughs in other domains. To fill the gap, this paper introduces two types of invertible attention mechanisms, i.e., map-based and transformer-based attentions, for both unconditional and conditional generative flows. The key idea is to exploit a masked scheme of these two attentions to learn long-range data dependencies in the context of generative flows. The masked scheme allows for invertible attention modules with tractable Jacobian determinants, enabling its seamless integration at any positions of the flow-based models. The proposed attention mechanisms lead to more efficient generative flows, due to their capability of modeling the long-term data dependencies. Evaluation on multiple image synthesis tasks shows that the proposed attention flows result in efficient models and compare favorably against the state-of-the-art unconditional and conditional generative flows.

Yan Wu, Zhiwu Huang, Suryansh Kumar et al.

Modern solutions to the single image super-resolution (SISR) problem using deep neural networks aim not only at better performance accuracy but also at a lighter and computationally efficient model. To that end, recently, neural architecture search (NAS) approaches have shown some tremendous potential. Following the same underlying, in this paper, we suggest a novel trilevel NAS method that provides a better balance between different efficiency metrics and performance to solve SISR. Unlike available NAS, our search is more complete, and therefore it leads to an efficient, optimized, and compressed architecture. We innovatively introduce a trilevel search space modeling, i.e., hierarchical modeling on network-, cell-, and kernel-level structures. To make the search on trilevel spaces differentiable and efficient, we exploit a new sparsestmax technique that is excellent at generating sparse distributions of individual neural architecture candidates so that they can be better disentangled for the final selection from the enlarged search space. We further introduce the sorting technique to the sparsestmax relaxation for better network-level compression. The proposed NAS optimization additionally facilitates simultaneous search and training in a single phase, reducing search time and train time. Comprehensive evaluations on the benchmark datasets show our method's clear superiority over the state-of-the-art NAS in terms of a good trade-off between model size, performance, and efficiency.

Berk Kaya, Suryansh Kumar, Carlos Oliveira et al.

This paper presents an uncalibrated deep neural network framework for the photometric stereo problem. For training models to solve the problem, existing neural network-based methods either require exact light directions or ground-truth surface normals of the object or both. However, in practice, it is challenging to procure both of this information precisely, which restricts the broader adoption of photometric stereo algorithms for vision application. To bypass this difficulty, we propose an uncalibrated neural inverse rendering approach to this problem. Our method first estimates the light directions from the input images and then optimizes an image reconstruction loss to calculate the surface normals, bidirectional reflectance distribution function value, and depth. Additionally, our formulation explicitly models the concave and convex parts of a complex surface to consider the effects of interreflections in the image formation process. Extensive evaluation of the proposed method on the challenging subjects generally shows comparable or better results than the supervised and classical approaches.

Rhea Sanjay Sukthanker, Zhiwu Huang, Suryansh Kumar et al.

In this paper, we propose a new neural architecture search (NAS) problem of Symmetric Positive Definite (SPD) manifold networks, aiming to automate the design of SPD neural architectures. To address this problem, we first introduce a geometrically rich and diverse SPD neural architecture search space for an efficient SPD cell design. Further, we model our new NAS problem with a one-shot training process of a single supernet. Based on the supernet modeling, we exploit a differentiable NAS algorithm on our relaxed continuous search space for SPD neural architecture search. Statistical evaluation of our method on drone, action, and emotion recognition tasks mostly provides better results than the state-of-the-art SPD networks and traditional NAS algorithms. Empirical results show that our algorithm excels in discovering better performing SPD network design and provides models that are more than three times lighter than searched by the state-of-the-art NAS algorithms.

Suryansh Kumar, Luc Van Gool, Carlos E. P. de Oliveira et al.

Non-Rigid Structure-from-Motion (NRSfM) problem aims to recover 3D geometry of a deforming object from its 2D feature correspondences across multiple frames. Classical approaches to this problem assume a small number of feature points and, ignore the local non-linearities of the shape deformation, and therefore, struggles to reliably model non-linear deformations. Furthermore, available dense NRSfM algorithms are often hurdled by scalability, computations, noisy measurements and, restricted to model just global deformation. In this paper, we propose algorithms that can overcome these limitations with the previous methods and, at the same time, can recover a reliable dense 3D structure of a non-rigid object with higher accuracy. Assuming that a deforming shape is composed of a union of local linear subspace and, span a global low-rank space over multiple frames enables us to efficiently model complex non-rigid deformations. To that end, each local linear subspace is represented using Grassmannians and, the global 3D shape across multiple frames is represented using a low-rank representation. We show that our approach significantly improves accuracy, scalability, and robustness against noise. Also, our representation naturally allows for simultaneous reconstruction and clustering framework which in general is observed to be more suitable for NRSfM problems. Our method currently achieves leading performance on the standard benchmark datasets.

Suryansh Kumar, Yuchao Dai, Hongdong Li

This work addresses the task of dense 3D reconstruction of a complex dynamic scene from images. The prevailing idea to solve this task is composed of a sequence of steps and is dependent on the success of several pipelines in its execution. To overcome such limitations with the existing algorithm, we propose a unified approach to solve this problem. We assume that a dynamic scene can be approximated by numerous piecewise planar surfaces, where each planar surface enjoys its own rigid motion, and the global change in the scene between two frames is as-rigid-as-possible (ARAP). Consequently, our model of a dynamic scene reduces to a soup of planar structures and rigid motion of these local planar structures. Using planar over-segmentation of the scene, we reduce this task to solving a "3D jigsaw puzzle" problem. Hence, the task boils down to correctly assemble each rigid piece to construct a 3D shape that complies with the geometry of the scene under the ARAP assumption. Further, we show that our approach provides an effective solution to the inherent scale-ambiguity in structure-from-motion under perspective projection. We provide extensive experimental results and evaluation on several benchmark datasets. Quantitative comparison with competing approaches shows state-of-the-art performance.

Suryansh Kumar

A simple prior free factorization algorithm \cite{dai2014simple} is quite often cited work in the field of Non-Rigid Structure from Motion (NRSfM). The benefit of this work lies in its simplicity of implementation, strong theoretical justification to the motion and structure estimation, and its invincible originality. Despite this, the prevailing view is, that it performs exceedingly inferior to other methods on several benchmark datasets \cite{jensen2018benchmark,akhter2009nonrigid}. However, our subtle investigation provides some empirical statistics which made us think against such views. The statistical results we obtained supersedes Dai {\it{et al.}}\cite{dai2014simple} originally reported results on the benchmark datasets by a significant margin under some elementary changes in their core algorithmic idea \cite{dai2014simple}. Now, these results not only exposes some unrevealed areas for research in NRSfM but also give rise to new mathematical challenges for NRSfM researchers. We argue that by \textbf{properly} utilizing the well-established assumptions about a non-rigidly deforming shape i.e, it deforms smoothly over frames \cite{rabaud2008re} and it spans a low-rank space, the simple prior-free idea can provide results which is comparable to the best available algorithms. In this paper, we explore some of the hidden intricacies missed by Dai {\it{et. al.}} work \cite{dai2014simple} and how some elementary measures and modifications can enhance its performance, as high as approx. 18\% on the benchmark dataset. The improved performance is justified and empirically verified by extensive experiments on several datasets. We believe our work has both practical and theoretical importance for the development of better NRSfM algorithms.

Suryansh Kumar, Ram Srivatsav Ghorakavi, Yuchao Dai et al.

Recent geometric methods need reliable estimates of 3D motion parameters to procure accurate dense depth map of a complex dynamic scene from monocular images \cite{kumar2017monocular, ranftl2016dense}. Generally, to estimate \textbf{precise} measurements of relative 3D motion parameters and to validate its accuracy using image data is a challenging task. In this work, we propose an alternative approach that circumvents the 3D motion estimation requirement to obtain a dense depth map of a dynamic scene. Given per-pixel optical flow correspondences between two consecutive frames and, the sparse depth prior for the reference frame, we show that, we can effectively recover the dense depth map for the successive frames without solving for 3D motion parameters. Our method assumes a piece-wise planar model of a dynamic scene, which undergoes rigid transformation locally, and as-rigid-as-possible transformation globally between two successive frames. Under our assumption, we can avoid the explicit estimation of 3D rotation and translation to estimate scene depth. In essence, our formulation provides an unconventional way to think and recover the dense depth map of a complex dynamic scene which is incremental and motion free in nature. Our proposed method does not make object level or any other high-level prior assumption about the dynamic scene, as a result, it is applicable to a wide range of scenarios. Experimental results on the benchmarks dataset show the competence of our approach for multiple frames.

Suryansh Kumar

Given dense image feature correspondences of a non-rigidly moving object across multiple frames, this paper proposes an algorithm to estimate its 3D shape for each frame. To solve this problem accurately, the recent state-of-the-art algorithm reduces this task to set of local linear subspace reconstruction and clustering problem using Grassmann manifold representation \cite{kumar2018scalable}. Unfortunately, their method missed on some of the critical issues associated with the modeling of surface deformations, for e.g., the dependence of a local surface deformation on its neighbors. Furthermore, their representation to group high dimensional data points inevitably introduce the drawbacks of categorizing samples on the high-dimensional Grassmann manifold \cite{huang2015projection, harandi2014manifold}. Hence, to deal with such limitations with \cite{kumar2018scalable}, we propose an algorithm that jointly exploits the benefit of high-dimensional Grassmann manifold to perform reconstruction, and its equivalent lower-dimensional representation to infer suitable clusters. To accomplish this, we project each Grassmannians onto a lower-dimensional Grassmann manifold which preserves and respects the deformation of the structure w.r.t its neighbors. These Grassmann points in the lower-dimension then act as a representative for the selection of high-dimensional Grassmann samples to perform each local reconstruction. In practice, our algorithm provides a geometrically efficient way to solve dense NRSfM by switching between manifolds based on its benefit and usage. Experimental results show that the proposed algorithm is very effective in handling noise with reconstruction accuracy as good as or better than the competing methods.

Suryansh Kumar, Anoop Cherian, Yuchao Dai et al.

This paper addresses the task of dense non-rigid structure-from-motion (NRSfM) using multiple images. State-of-the-art methods to this problem are often hurdled by scalability, expensive computations, and noisy measurements. Further, recent methods to NRSfM usually either assume a small number of sparse feature points or ignore local non-linearities of shape deformations, and thus cannot reliably model complex non-rigid deformations. To address these issues, in this paper, we propose a new approach for dense NRSfM by modeling the problem on a Grassmann manifold. Specifically, we assume the complex non-rigid deformations lie on a union of local linear subspaces both spatially and temporally. This naturally allows for a compact representation of the complex non-rigid deformation over frames. We provide experimental results on several synthetic and real benchmark datasets. The procured results clearly demonstrate that our method, apart from being scalable and more accurate than state-of-the-art methods, is also more robust to noise and generalizes to highly non-linear deformations.

Suryansh Kumar, Yuchao Dai, Hongdong Li

This paper proposes a new approach for monocular dense 3D reconstruction of a complex dynamic scene from two perspective frames. By applying superpixel over-segmentation to the image, we model a generically dynamic (hence non-rigid) scene with a piecewise planar and rigid approximation. In this way, we reduce the dynamic reconstruction problem to a "3D jigsaw puzzle" problem which takes pieces from an unorganized "soup of superpixels". We show that our method provides an effective solution to the inherent relative scale ambiguity in structure-from-motion. Since our method does not assume a template prior, or per-object segmentation, or knowledge about the rigidity of the dynamic scene, it is applicable to a wide range of scenarios. Extensive experiments on both synthetic and real monocular sequences demonstrate the superiority of our method compared with the state-of-the-art methods.

Suryansh Kumar, Yuchao Dai, Hongdong Li

Non-rigid structure-from-motion (NRSfM) has so far been mostly studied for recovering 3D structure of a single non-rigid/deforming object. To handle the real world challenging multiple deforming objects scenarios, existing methods either pre-segment different objects in the scene or treat multiple non-rigid objects as a whole to obtain the 3D non-rigid reconstruction. However, these methods fail to exploit the inherent structure in the problem as the solution of segmentation and the solution of reconstruction could not benefit each other. In this paper, we propose a unified framework to jointly segment and reconstruct multiple non-rigid objects. To compactly represent complex multi-body non-rigid scenes, we propose to exploit the structure of the scenes along both temporal direction and spatial direction, thus achieving a spatio-temporal representation. Specifically, we represent the 3D non-rigid deformations as lying in a union of subspaces along the temporal direction and represent the 3D trajectories as lying in the union of subspaces along the spatial direction. This spatio-temporal representation not only provides competitive 3D reconstruction but also outputs robust segmentation of multiple non-rigid objects. The resultant optimization problem is solved efficiently using the Alternating Direction Method of Multipliers (ADMM). Extensive experimental results on both synthetic and real multi-body NRSfM datasets demonstrate the superior performance of our proposed framework compared with the state-of-the-art methods.

Suryansh Kumar, Yuchao Dai, Hongdong Li

Conventional structure-from-motion (SFM) research is primarily concerned with the 3D reconstruction of a single, rigidly moving object seen by a static camera, or a static and rigid scene observed by a moving camera --in both cases there are only one relative rigid motion involved. Recent progress have extended SFM to the areas of {multi-body SFM} (where there are {multiple rigid} relative motions in the scene), as well as {non-rigid SFM} (where there is a single non-rigid, deformable object or scene). Along this line of thinking, there is apparently a missing gap of "multi-body non-rigid SFM", in which the task would be to jointly reconstruct and segment multiple 3D structures of the multiple, non-rigid objects or deformable scenes from images. Such a multi-body non-rigid scenario is common in reality (e.g. two persons shaking hands, multi-person social event), and how to solve it represents a natural {next-step} in SFM research. By leveraging recent results of subspace clustering, this paper proposes, for the first time, an effective framework for multi-body NRSFM, which simultaneously reconstructs and segments each 3D trajectory into their respective low-dimensional subspace. Under our formulation, 3D trajectories for each non-rigid structure can be well approximated with a sparse affine combination of other 3D trajectories from the same structure (self-expressiveness). We solve the resultant optimization with the alternating direction method of multipliers (ADMM). We demonstrate the efficacy of the proposed framework through extensive experiments on both synthetic and real data sequences. Our method clearly outperforms other alternative methods, such as first clustering the 2D feature tracks to groups and then doing non-rigid reconstruction in each group or first conducting 3D reconstruction by using single subspace assumption and then clustering the 3D trajectories into groups.