Fabian Manhardt

Gu Wang, Fabian Manhardt, Xingyu Liu et al. · tsinghua

6D object pose estimation is a fundamental yet challenging problem in computer vision. Convolutional Neural Networks (CNNs) have recently proven to be capable of predicting reliable 6D pose estimates even under monocular settings. Nonetheless, CNNs are identified as being extremely data-driven, and acquiring adequate annotations is oftentimes very time-consuming and labor intensive. To overcome this limitation, we propose a novel monocular 6D pose estimation approach by means of self-supervised learning, removing the need for real annotations. After training our proposed network fully supervised with synthetic RGB data, we leverage current trends in noisy student training and differentiable rendering to further self-supervise the model on these unsupervised real RGB(-D) samples, seeking for a visually and geometrically optimal alignment. Moreover, employing both visible and amodal mask information, our self-supervision becomes very robust towards challenging scenarios such as occlusion. Extensive evaluations demonstrate that our proposed self-supervision outperforms all other methods relying on synthetic data or employing elaborate techniques from the domain adaptation realm. Noteworthy, our self-supervised approach consistently improves over its synthetically trained baseline and often almost closes the gap towards its fully supervised counterpart. The code and models are publicly available at https://github.com/THU-DA-6D-Pose-Group/self6dpp.git.

Guangyao Zhai, Dianye Huang, Shun-Cheng Wu et al.

6-DoF robotic grasping is a long-lasting but unsolved problem. Recent methods utilize strong 3D networks to extract geometric grasping representations from depth sensors, demonstrating superior accuracy on common objects but perform unsatisfactorily on photometrically challenging objects, e.g., objects in transparent or reflective materials. The bottleneck lies in that the surface of these objects can not reflect back accurate depth due to the absorption or refraction of light. In this paper, in contrast to exploiting the inaccurate depth data, we propose the first RGB-only 6-DoF grasping pipeline called MonoGraspNet that utilizes stable 2D features to simultaneously handle arbitrary object grasping and overcome the problems induced by photometrically challenging objects. MonoGraspNet leverages keypoint heatmap and normal map to recover the 6-DoF grasping poses represented by our novel representation parameterized with 2D keypoints with corresponding depth, grasping direction, grasping width, and angle. Extensive experiments in real scenes demonstrate that our method can achieve competitive results in grasping common objects and surpass the depth-based competitor by a large margin in grasping photometrically challenging objects. To further stimulate robotic manipulation research, we additionally annotate and open-source a multi-view and multi-scene real-world grasping dataset, containing 120 objects of mixed photometric complexity with 20M accurate grasping labels.

Chenyangguang Zhang, Yan Di, Ruida Zhang et al.

Reconstructing hand-held objects from a single RGB image is an important and challenging problem. Existing works utilizing Signed Distance Fields (SDF) reveal limitations in comprehensively capturing the complex hand-object interactions, since SDF is only reliable within the proximity of the target, and hence, infeasible to simultaneously encode local hand and object cues. To address this issue, we propose DDF-HO, a novel approach leveraging Directed Distance Field (DDF) as the shape representation. Unlike SDF, DDF maps a ray in 3D space, consisting of an origin and a direction, to corresponding DDF values, including a binary visibility signal determining whether the ray intersects the objects and a distance value measuring the distance from origin to target in the given direction. We randomly sample multiple rays and collect local to global geometric features for them by introducing a novel 2D ray-based feature aggregation scheme and a 3D intersection-aware hand pose embedding, combining 2D-3D features to model hand-object interactions. Extensive experiments on synthetic and real-world datasets demonstrate that DDF-HO consistently outperforms all baseline methods by a large margin, especially under Chamfer Distance, with about 80% leap forward. Codes are available at https://github.com/ZhangCYG/DDFHO.

Yan Di, Ruida Zhang, Zhiqiang Lou et al.

While 6D object pose estimation has recently made a huge leap forward, most methods can still only handle a single or a handful of different objects, which limits their applications. To circumvent this problem, category-level object pose estimation has recently been revamped, which aims at predicting the 6D pose as well as the 3D metric size for previously unseen instances from a given set of object classes. This is, however, a much more challenging task due to severe intra-class shape variations. To address this issue, we propose GPV-Pose, a novel framework for robust category-level pose estimation, harnessing geometric insights to enhance the learning of category-level pose-sensitive features. First, we introduce a decoupled confidence-driven rotation representation, which allows geometry-aware recovery of the associated rotation matrix. Second, we propose a novel geometry-guided point-wise voting paradigm for robust retrieval of the 3D object bounding box. Finally, leveraging these different output streams, we can enforce several geometric consistency terms, further increasing performance, especially for non-symmetric categories. GPV-Pose produces superior results to state-of-the-art competitors on common public benchmarks, whilst almost achieving real-time inference speed at 20 FPS.

Ruida Zhang, Yan Di, Zhiqiang Lou et al.

Category-level object pose estimation aims to predict the 6D pose as well as the 3D metric size of arbitrary objects from a known set of categories. Recent methods harness shape prior adaptation to map the observed point cloud into the canonical space and apply Umeyama algorithm to recover the pose and size. However, their shape prior integration strategy boosts pose estimation indirectly, which leads to insufficient pose-sensitive feature extraction and slow inference speed. To tackle this problem, in this paper, we propose a novel geometry-guided Residual Object Bounding Box Projection network RBP-Pose that jointly predicts object pose and residual vectors describing the displacements from the shape-prior-indicated object surface projections on the bounding box towards the real surface projections. Such definition of residual vectors is inherently zero-mean and relatively small, and explicitly encapsulates spatial cues of the 3D object for robust and accurate pose regression. We enforce geometry-aware consistency terms to align the predicted pose and residual vectors to further boost performance.

Yamei Chen, Yan Di, Guangyao Zhai et al.

Category-level object pose estimation, aiming to predict the 6D pose and 3D size of objects from known categories, typically struggles with large intra-class shape variation. Existing works utilizing mean shapes often fall short of capturing this variation. To address this issue, we present SecondPose, a novel approach integrating object-specific geometric features with semantic category priors from DINOv2. Leveraging the advantage of DINOv2 in providing SE(3)-consistent semantic features, we hierarchically extract two types of SE(3)-invariant geometric features to further encapsulate local-to-global object-specific information. These geometric features are then point-aligned with DINOv2 features to establish a consistent object representation under SE(3) transformations, facilitating the mapping from camera space to the pre-defined canonical space, thus further enhancing pose estimation. Extensive experiments on NOCS-REAL275 demonstrate that SecondPose achieves a 12.4% leap forward over the state-of-the-art. Moreover, on a more complex dataset HouseCat6D which provides photometrically challenging objects, SecondPose still surpasses other competitors by a large margin.

Hanzhi Chen, Fabian Manhardt, Nassir Navab et al.

In this paper, we introduce neural texture learning for 6D object pose estimation from synthetic data and a few unlabelled real images. Our major contribution is a novel learning scheme which removes the drawbacks of previous works, namely the strong dependency on co-modalities or additional refinement. These have been previously necessary to provide training signals for convergence. We formulate such a scheme as two sub-optimisation problems on texture learning and pose learning. We separately learn to predict realistic texture of objects from real image collections and learn pose estimation from pixel-perfect synthetic data. Combining these two capabilities allows then to synthesise photorealistic novel views to supervise the pose estimator with accurate geometry. To alleviate pose noise and segmentation imperfection present during the texture learning phase, we propose a surfel-based adversarial training loss together with texture regularisation from synthetic data. We demonstrate that the proposed approach significantly outperforms the recent state-of-the-art methods without ground-truth pose annotations and demonstrates substantial generalisation improvements towards unseen scenes. Remarkably, our scheme improves the adopted pose estimators substantially even when initialised with much inferior performance.

Guangyao Zhai, Xiaoni Cai, Dianye Huang et al.

Object rearrangement is pivotal in robotic-environment interactions, representing a significant capability in embodied AI. In this paper, we present SG-Bot, a novel rearrangement framework that utilizes a coarse-to-fine scheme with a scene graph as the scene representation. Unlike previous methods that rely on either known goal priors or zero-shot large models, SG-Bot exemplifies lightweight, real-time, and user-controllable characteristics, seamlessly blending the consideration of commonsense knowledge with automatic generation capabilities. SG-Bot employs a three-fold procedure--observation, imagination, and execution--to adeptly address the task. Initially, objects are discerned and extracted from a cluttered scene during the observation. These objects are first coarsely organized and depicted within a scene graph, guided by either commonsense or user-defined criteria. Then, this scene graph subsequently informs a generative model, which forms a fine-grained goal scene considering the shape information from the initial scene and object semantics. Finally, for execution, the initial and envisioned goal scenes are matched to formulate robotic action policies. Experimental results demonstrate that SG-Bot outperforms competitors by a large margin.

Prune Truong, Marie-Julie Rakotosaona, Fabian Manhardt et al.

Neural Radiance Field (NeRF) has recently emerged as a powerful representation to synthesize photorealistic novel views. While showing impressive performance, it relies on the availability of dense input views with highly accurate camera poses, thus limiting its application in real-world scenarios. In this work, we introduce Sparse Pose Adjusting Radiance Field (SPARF), to address the challenge of novel-view synthesis given only few wide-baseline input images (as low as 3) with noisy camera poses. Our approach exploits multi-view geometry constraints in order to jointly learn the NeRF and refine the camera poses. By relying on pixel matches extracted between the input views, our multi-view correspondence objective enforces the optimized scene and camera poses to converge to a global and geometrically accurate solution. Our depth consistency loss further encourages the reconstructed scene to be consistent from any viewpoint. Our approach sets a new state of the art in the sparse-view regime on multiple challenging datasets.

Yongzhi Su, Yan Di, Fabian Manhardt et al.

Despite monocular 3D object detection having recently made a significant leap forward thanks to the use of pre-trained depth estimators for pseudo-LiDAR recovery, such two-stage methods typically suffer from overfitting and are incapable of explicitly encapsulating the geometric relation between depth and object bounding box. To overcome this limitation, we instead propose OPA-3D, a single-stage, end-to-end, Occlusion-Aware Pixel-Wise Aggregation network that to jointly estimate dense scene depth with depth-bounding box residuals and object bounding boxes, allowing a two-stream detection of 3D objects, leading to significantly more robust detections. Thereby, the geometry stream denoted as the Geometry Stream, combines visible depth and depth-bounding box residuals to recover the object bounding box via explicit occlusion-aware optimization. In addition, a bounding box based geometry projection scheme is employed in an effort to enhance distance perception. The second stream, named as the Context Stream, directly regresses 3D object location and size. This novel two-stream representation further enables us to enforce cross-stream consistency terms which aligns the outputs of both streams, improving the overall performance. Extensive experiments on the public benchmark demonstrate that OPA-3D outperforms state-of-the-art methods on the main Car category, whilst keeping a real-time inference speed. We plan to release all codes and trained models soon.

Dekai Zhu, Guangyao Zhai, Yan Di et al.

Reliable multi-agent trajectory prediction is crucial for the safe planning and control of autonomous systems. Compared with single-agent cases, the major challenge in simultaneously processing multiple agents lies in modeling complex social interactions caused by various driving intentions and road conditions. Previous methods typically leverage graph-based message propagation or attention mechanism to encapsulate such interactions in the format of marginal probabilistic distributions. However, it is inherently sub-optimal. In this paper, we propose IPCC-TP, a novel relevance-aware module based on Incremental Pearson Correlation Coefficient to improve multi-agent interaction modeling. IPCC-TP learns pairwise joint Gaussian Distributions through the tightly-coupled estimation of the means and covariances according to interactive incremental movements. Our module can be conveniently embedded into existing multi-agent prediction methods to extend original motion distribution decoders. Extensive experiments on nuScenes and Argoverse 2 datasets demonstrate that IPCC-TP improves the performance of baselines by a large margin.

Christina Tsalicoglou, Fabian Manhardt, Alessio Tonioni et al.

The ability to generate highly realistic 2D images from mere text prompts has recently made huge progress in terms of speed and quality, thanks to the advent of image diffusion models. Naturally, the question arises if this can be also achieved in the generation of 3D content from such text prompts. To this end, a new line of methods recently emerged trying to harness diffusion models, trained on 2D images, for supervision of 3D model generation using view dependent prompts. While achieving impressive results, these methods, however, have two major drawbacks. First, rather than commonly used 3D meshes, they instead generate neural radiance fields (NeRFs), making them impractical for most real applications. Second, these approaches tend to produce over-saturated models, giving the output a cartoonish looking effect. Therefore, in this work we propose a novel method for generation of highly realistic-looking 3D meshes. To this end, we extend NeRF to employ an SDF backbone, leading to improved 3D mesh extraction. In addition, we propose a novel way to finetune the mesh texture, removing the effect of high saturation and improving the details of the output 3D mesh.

Chenyangguang Zhang, Guanlong Jiao, Yan Di et al. · tsinghua

Previous works concerning single-view hand-held object reconstruction typically rely on supervision from 3D ground-truth models, which are hard to collect in real world. In contrast, readily accessible hand-object videos offer a promising training data source, but they only give heavily occluded object observations. In this paper, we present a novel synthetic-to-real framework to exploit Multi-view Occlusion-aware supervision from hand-object videos for Hand-held Object reconstruction (MOHO) from a single image, tackling two predominant challenges in such setting: hand-induced occlusion and object's self-occlusion. First, in the synthetic pre-training stage, we render a large-scaled synthetic dataset SOMVideo with hand-object images and multi-view occlusion-free supervisions, adopted to address hand-induced occlusion in both 2D and 3D spaces. Second, in the real-world finetuning stage, MOHO leverages the amodal-mask-weighted geometric supervision to mitigate the unfaithful guidance caused by the hand-occluded supervising views in real world. Moreover, domain-consistent occlusion-aware features are amalgamated in MOHO to resist object's self-occlusion for inferring the complete object shape. Extensive experiments on HO3D and DexYCB datasets demonstrate 2D-supervised MOHO gains superior results against 3D-supervised methods by a large margin.

Yan Di, Chenyangguang Zhang, Ruida Zhang et al.

In this paper, we propose U-RED, an Unsupervised shape REtrieval and Deformation pipeline that takes an arbitrary object observation as input, typically captured by RGB images or scans, and jointly retrieves and deforms the geometrically similar CAD models from a pre-established database to tightly match the target. Considering existing methods typically fail to handle noisy partial observations, U-RED is designed to address this issue from two aspects. First, since one partial shape may correspond to multiple potential full shapes, the retrieval method must allow such an ambiguous one-to-many relationship. Thereby U-RED learns to project all possible full shapes of a partial target onto the surface of a unit sphere. Then during inference, each sampling on the sphere will yield a feasible retrieval. Second, since real-world partial observations usually contain noticeable noise, a reliable learned metric that measures the similarity between shapes is necessary for stable retrieval. In U-RED, we design a novel point-wise residual-guided metric that allows noise-robust comparison. Extensive experiments on the synthetic datasets PartNet, ComplementMe and the real-world dataset Scan2CAD demonstrate that U-RED surpasses existing state-of-the-art approaches by 47.3%, 16.7% and 31.6% respectively under Chamfer Distance.

Marie-Julie Rakotosaona, Fabian Manhardt, Diego Martin Arroyo et al.

With the introduction of Neural Radiance Fields (NeRFs), novel view synthesis has recently made a big leap forward. At the core, NeRF proposes that each 3D point can emit radiance, allowing to conduct view synthesis using differentiable volumetric rendering. While neural radiance fields can accurately represent 3D scenes for computing the image rendering, 3D meshes are still the main scene representation supported by most computer graphics and simulation pipelines, enabling tasks such as real time rendering and physics-based simulations. Obtaining 3D meshes from neural radiance fields still remains an open challenge since NeRFs are optimized for view synthesis, not enforcing an accurate underlying geometry on the radiance field. We thus propose a novel compact and flexible architecture that enables easy 3D surface reconstruction from any NeRF-driven approach. Upon having trained the radiance field, we distill the volumetric 3D representation into a Signed Surface Approximation Network, allowing easy extraction of the 3D mesh and appearance. Our final 3D mesh is physically accurate and can be rendered in real time on an array of devices.

Bowen Fu, Gu Wang, Chenyangguang Zhang et al. · tsinghua

Reconstructing hand-held objects from a single RGB image is a challenging task in computer vision. In contrast to prior works that utilize deterministic modeling paradigms, we employ a point cloud denoising diffusion model to account for the probabilistic nature of this problem. In the core, we introduce centroid-fixed dual-stream conditional diffusion for monocular hand-held object reconstruction (D-SCo), tackling two predominant challenges. First, to avoid the object centroid from deviating, we utilize a novel hand-constrained centroid fixing paradigm, enhancing the stability of diffusion and reverse processes and the precision of feature projection. Second, we introduce a dual-stream denoiser to semantically and geometrically model hand-object interactions with a novel unified hand-object semantic embedding, enhancing the reconstruction performance of the hand-occluded region of the object. Experiments on the synthetic ObMan dataset and three real-world datasets HO3D, MOW and DexYCB demonstrate that our approach can surpass all other state-of-the-art methods.

Yan Di, Chenyangguang Zhang, Pengyuan Wang et al.

In this paper, we present a novel shape reconstruction method leveraging diffusion model to generate 3D sparse point cloud for the object captured in a single RGB image. Recent methods typically leverage global embedding or local projection-based features as the condition to guide the diffusion model. However, such strategies fail to consistently align the denoised point cloud with the given image, leading to unstable conditioning and inferior performance. In this paper, we present CCD-3DR, which exploits a novel centered diffusion probabilistic model for consistent local feature conditioning. We constrain the noise and sampled point cloud from the diffusion model into a subspace where the point cloud center remains unchanged during the forward diffusion process and reverse process. The stable point cloud center further serves as an anchor to align each point with its corresponding local projection-based features. Extensive experiments on synthetic benchmark ShapeNet-R2N2 demonstrate that CCD-3DR outperforms all competitors by a large margin, with over 40% improvement. We also provide results on real-world dataset Pix3D to thoroughly demonstrate the potential of CCD-3DR in real-world applications. Codes will be released soon

Ruida Zhang, Yan Di, Fabian Manhardt et al.

Category-level pose estimation is a challenging problem due to intra-class shape variations. Recent methods deform pre-computed shape priors to map the observed point cloud into the normalized object coordinate space and then retrieve the pose via post-processing, i.e., Umeyama's Algorithm. The shortcomings of this two-stage strategy lie in two aspects: 1) The surrogate supervision on the intermediate results can not directly guide the learning of pose, resulting in large pose error after post-processing. 2) The inference speed is limited by the post-processing step. In this paper, to handle these shortcomings, we propose an end-to-end trainable network SSP-Pose for category-level pose estimation, which integrates shape priors into a direct pose regression network. SSP-Pose stacks four individual branches on a shared feature extractor, where two branches are designed to deform and match the prior model with the observed instance, and the other two branches are applied for directly regressing the totally 9 degrees-of-freedom pose and performing symmetry reconstruction and point-wise inlier mask prediction respectively. Consistency loss terms are then naturally exploited to align the outputs of different branches and promote the performance. During inference, only the direct pose regression branch is needed. In this manner, SSP-Pose not only learns category-level pose-sensitive characteristics to boost performance but also keeps a real-time inference speed. Moreover, we utilize the symmetry information of each category to guide the shape prior deformation, and propose a novel symmetry-aware loss to mitigate the matching ambiguity. Extensive experiments on public datasets demonstrate that SSP-Pose produces superior performance compared with competitors with a real-time inference speed at about 25Hz.

Yongliang Lin, Yongzhi Su, Praveen Nathan et al.

In this work, we present a novel dense-correspondence method for 6DoF object pose estimation from a single RGB-D image. While many existing data-driven methods achieve impressive performance, they tend to be time-consuming due to their reliance on rendering-based refinement approaches. To circumvent this limitation, we present HiPose, which establishes 3D-3D correspondences in a coarse-to-fine manner with a hierarchical binary surface encoding. Unlike previous dense-correspondence methods, we estimate the correspondence surface by employing point-to-surface matching and iteratively constricting the surface until it becomes a correspondence point while gradually removing outliers. Extensive experiments on public benchmarks LM-O, YCB-V, and T-Less demonstrate that our method surpasses all refinement-free methods and is even on par with expensive refinement-based approaches. Crucially, our approach is computationally efficient and enables real-time critical applications with high accuracy requirements.

Issa Mouawad, Nikolas Brasch, Fabian Manhardt et al.

Monocular 3D object detection continues to attract attention due to the cost benefits and wider availability of RGB cameras. Despite the recent advances and the ability to acquire data at scale, annotation cost and complexity still limit the size of 3D object detection datasets in the supervised settings. Self-supervised methods, on the other hand, aim at training deep networks relying on pretext tasks or various consistency constraints. Moreover, other 3D perception tasks (such as depth estimation) have shown the benefits of temporal priors as a self-supervision signal. In this work, we argue that the temporal consistency on the level of object poses, provides an important supervision signal given the strong prior on physical motion. Specifically, we propose a self-supervised loss which uses this consistency, in addition to render-and-compare losses, to refine noisy pose predictions and derive high-quality pseudo labels. To assess the effectiveness of the proposed method, we finetune a synthetically trained monocular 3D object detection model using the pseudo-labels that we generated on real data. Evaluation on the standard KITTI3D benchmark demonstrates that our method reaches competitive performance compared to other monocular self-supervised and supervised methods.

Felix Wimbauer, Fabian Manhardt, Michael Oechsle et al.

The synthesis of immersive 3D scenes from text is rapidly maturing, driven by novel video generative models and feed-forward 3D reconstruction, with vast potential in AR/VR and world modeling. While panoramic images have proven effective for scene initialization, existing approaches suffer from a trade-off between visual fidelity and explorability: autoregressive expansion suffers from context drift, while panoramic video generation is limited to low resolution. We present Stepper, a unified framework for text-driven immersive 3D scene synthesis that circumvents these limitations via stepwise panoramic scene expansion. Stepper leverages a novel multi-view 360° diffusion model that enables consistent, high-resolution expansion, coupled with a geometry reconstruction pipeline that enforces geometric coherence. Trained on a new large-scale, multi-view panorama dataset, Stepper achieves state-of-the-art fidelity and structural consistency, outperforming prior approaches, thereby setting a new standard for immersive scene generation.

Ruida Zhang, Chenyangguang Zhang, Yan Di et al.

In this paper, we present KP-RED, a unified KeyPoint-driven REtrieval and Deformation framework that takes object scans as input and jointly retrieves and deforms the most geometrically similar CAD models from a pre-processed database to tightly match the target. Unlike existing dense matching based methods that typically struggle with noisy partial scans, we propose to leverage category-consistent sparse keypoints to naturally handle both full and partial object scans. Specifically, we first employ a lightweight retrieval module to establish a keypoint-based embedding space, measuring the similarity among objects by dynamically aggregating deformation-aware local-global features around extracted keypoints. Objects that are close in the embedding space are considered similar in geometry. Then we introduce the neural cage-based deformation module that estimates the influence vector of each keypoint upon cage vertices inside its local support region to control the deformation of the retrieved shape. Extensive experiments on the synthetic dataset PartNet and the real-world dataset Scan2CAD demonstrate that KP-RED surpasses existing state-of-the-art approaches by a large margin. Codes and trained models are released on https://github.com/lolrudy/KP-RED.

Michael Niemeyer, Fabian Manhardt, Marie-Julie Rakotosaona et al.

Recent advances in view synthesis and real-time rendering have achieved photorealistic quality at impressive rendering speeds. While Radiance Field-based methods achieve state-of-the-art quality in challenging scenarios such as in-the-wild captures and large-scale scenes, they often suffer from excessively high compute requirements linked to volumetric rendering. Gaussian Splatting-based methods, on the other hand, rely on rasterization and naturally achieve real-time rendering but suffer from brittle optimization heuristics that underperform on more challenging scenes. In this work, we present RadSplat, a lightweight method for robust real-time rendering of complex scenes. Our main contributions are threefold. First, we use radiance fields as a prior and supervision signal for optimizing point-based scene representations, leading to improved quality and more robust optimization. Next, we develop a novel pruning technique reducing the overall point count while maintaining high quality, leading to smaller and more compact scene representations with faster inference speeds. Finally, we propose a novel test-time filtering approach that further accelerates rendering and allows to scale to larger, house-sized scenes. We find that our method enables state-of-the-art synthesis of complex captures at 900+ FPS.

Francis Engelmann, Fabian Manhardt, Michael Niemeyer et al.

Large visual-language models (VLMs), like CLIP, enable open-set image segmentation to segment arbitrary concepts from an image in a zero-shot manner. This goes beyond the traditional closed-set assumption, i.e., where models can only segment classes from a pre-defined training set. More recently, first works on open-set segmentation in 3D scenes have appeared in the literature. These methods are heavily influenced by closed-set 3D convolutional approaches that process point clouds or polygon meshes. However, these 3D scene representations do not align well with the image-based nature of the visual-language models. Indeed, point cloud and 3D meshes typically have a lower resolution than images and the reconstructed 3D scene geometry might not project well to the underlying 2D image sequences used to compute pixel-aligned CLIP features. To address these challenges, we propose OpenNeRF which naturally operates on posed images and directly encodes the VLM features within the NeRF. This is similar in spirit to LERF, however our work shows that using pixel-wise VLM features (instead of global CLIP features) results in an overall less complex architecture without the need for additional DINO regularization. Our OpenNeRF further leverages NeRF's ability to render novel views and extract open-set VLM features from areas that are not well observed in the initial posed images. For 3D point cloud segmentation on the Replica dataset, OpenNeRF outperforms recent open-vocabulary methods such as LERF and OpenScene by at least +4.9 mIoU.

Titas Anciukevičius, Fabian Manhardt, Federico Tombari et al.

Generating 3D scenes is a challenging open problem, which requires synthesizing plausible content that is fully consistent in 3D space. While recent methods such as neural radiance fields excel at view synthesis and 3D reconstruction, they cannot synthesize plausible details in unobserved regions since they lack a generative capability. Conversely, existing generative methods are typically not capable of reconstructing detailed, large-scale scenes in the wild, as they use limited-capacity 3D scene representations, require aligned camera poses, or rely on additional regularizers. In this work, we introduce the first diffusion model able to perform fast, detailed reconstruction and generation of real-world 3D scenes. To achieve this, we make three contributions. First, we introduce a new neural scene representation, IB-planes, that can efficiently and accurately represent large 3D scenes, dynamically allocating more capacity as needed to capture details visible in each image. Second, we propose a denoising-diffusion framework to learn a prior over this novel 3D scene representation, using only 2D images without the need for any additional supervision signal such as masks or depths. This supports 3D reconstruction and generation in a unified architecture. Third, we develop a principled approach to avoid trivial 3D solutions when integrating the image-based rendering with the diffusion model, by dropping out representations of some images. We evaluate the model on several challenging datasets of real and synthetic images, and demonstrate superior results on generation, novel view synthesis and 3D reconstruction.

Nikolai Kalischek, Michael Oechsle, Fabian Manhardt et al.

We introduce a novel method for generating 360° panoramas from text prompts or images. Our approach leverages recent advances in 3D generation by employing multi-view diffusion models to jointly synthesize the six faces of a cubemap. Unlike previous methods that rely on processing equirectangular projections or autoregressive generation, our method treats each face as a standard perspective image, simplifying the generation process and enabling the use of existing multi-view diffusion models. We demonstrate that these models can be adapted to produce high-quality cubemaps without requiring correspondence-aware attention layers. Our model allows for fine-grained text control, generates high resolution panorama images and generalizes well beyond its training set, whilst achieving state-of-the-art results, both qualitatively and quantitatively. Project page: https://cubediff.github.io/

Jonas Kulhanek, Marie-Julie Rakotosaona, Fabian Manhardt et al.

In this work, we present a novel level-of-detail (LOD) method for 3D Gaussian Splatting that enables real-time rendering of large-scale scenes on memory-constrained devices. Our approach introduces a hierarchical LOD representation that iteratively selects optimal subsets of Gaussians based on camera distance, thus largely reducing both rendering time and GPU memory usage. We construct each LOD level by applying a depth-aware 3D smoothing filter, followed by importance-based pruning and fine-tuning to maintain visual fidelity. To further reduce memory overhead, we partition the scene into spatial chunks and dynamically load only relevant Gaussians during rendering, employing an opacity-blending mechanism to avoid visual artifacts at chunk boundaries. Our method achieves state-of-the-art performance on both outdoor (Hierarchical 3DGS) and indoor (Zip-NeRF) datasets, delivering high-quality renderings with reduced latency and memory requirements.

Orest Kupyn, Fabian Manhardt, Federico Tombari et al.

Video generation models have progressed tremendously through large latent diffusion transformers trained with rectified flow techniques. Yet these models still struggle with geometric inconsistencies, unstable motion, and visual artifacts that break the illusion of realistic 3D scenes. 3D-consistent video generation could significantly impact numerous downstream applications in generation and reconstruction tasks. We explore how epipolar geometry constraints improve modern video diffusion models. Despite massive training data, these models fail to capture fundamental geometric principles underlying visual content. We align diffusion models using pairwise epipolar geometry constraints via preference-based optimization, directly addressing unstable camera trajectories and geometric artifacts through mathematically principled geometric enforcement. Our approach efficiently enforces geometric principles without requiring end-to-end differentiability. Evaluation demonstrates that classical geometric constraints provide more stable optimization signals than modern learned metrics, which produce noisy targets that compromise alignment quality. Training on static scenes with dynamic cameras ensures high-quality measurements while the model generalizes effectively to diverse dynamic content. By bridging data-driven deep learning with classical geometric computer vision, we present a practical method for generating spatially consistent videos without compromising visual quality.

Michael Niemeyer, Fabian Manhardt, Marie-Julie Rakotosaona et al.

Recent advances in neural scene representations have led to unprecedented quality in 3D reconstruction and view synthesis. Despite achieving high-quality results for common benchmarks with curated data, outputs often degrade for data that contain per image variations such as strong exposure changes, present, e.g., in most scenes with indoor and outdoor areas or rooms with windows. In this paper, we introduce Neural Exposure Fields (NExF), a novel technique for robustly reconstructing 3D scenes with high quality and 3D-consistent appearance from challenging real-world captures. In the core, we propose to learn a neural field predicting an optimal exposure value per 3D point, enabling us to optimize exposure along with the neural scene representation. While capture devices such as cameras select optimal exposure per image/pixel, we generalize this concept and perform optimization in 3D instead. This enables accurate view synthesis in high dynamic range scenarios, bypassing the need of post-processing steps or multi-exposure captures. Our contributions include a novel neural representation for exposure prediction, a system for joint optimization of the scene representation and the exposure field via a novel neural conditioning mechanism, and demonstrated superior performance on challenging real-world data. We find that our approach trains faster than prior works and produces state-of-the-art results on several benchmarks improving by over 55% over best-performing baselines.

Pietro Bonazzi, Mengqi Wang, Diego Martin Arroyo et al.

Synthesizing realistic and diverse indoor 3D scene layouts in a controllable fashion opens up applications in simulated navigation and virtual reality. As concise and robust representations of a scene, scene graphs have proven to be well-suited as the semantic control on the generated layout. We present a variant of the conditional variational autoencoder (cVAE) model to synthesize 3D scenes from scene graphs and floor plans. We exploit the properties of self-attention layers to capture high-level relationships between objects in a scene, and use these as the building blocks of our model. Our model, leverages graph transformers to estimate the size, dimension and orientation of the objects in a room while satisfying relationships in the given scene graph. Our experiments shows self-attention layers leads to sparser (7.9x compared to Graphto3D) and more diverse scenes (16%).

Issa Mouawad, Nikolas Brasch, Fabian Manhardt et al.

For autonomous vehicles, driving safely is highly dependent on the capability to correctly perceive the environment in 3D space, hence the task of 3D object detection represents a fundamental aspect of perception. While 3D sensors deliver accurate metric perception, monocular approaches enjoy cost and availability advantages that are valuable in a wide range of applications. Unfortunately, training monocular methods requires a vast amount of annotated data. Interestingly, self-supervised approaches have recently been successfully applied to ease the training process and unlock access to widely available unlabelled data. While related research leverages different priors including LIDAR scans and stereo images, such priors again limit usability. Therefore, in this work, we propose a novel approach to self-supervise 3D object detection purely from RGB sequences alone, leveraging multi-view constraints and weak labels. Our experiments on KITTI 3D dataset demonstrate performance on par with state-of-the-art self-supervised methods using LIDAR scans or stereo images.

Pengyuan Wang, Fabian Manhardt, Luca Minciullo et al.

The ability to successfully grasp objects is crucial in robotics, as it enables several interactive downstream applications. To this end, most approaches either compute the full 6D pose for the object of interest or learn to predict a set of grasping points. While the former approaches do not scale well to multiple object instances or classes yet, the latter require large annotated datasets and are hampered by their poor generalization capabilities to new geometries. To overcome these shortcomings, we propose to teach a robot how to grasp an object with a simple and short human demonstration. Hence, our approach neither requires many annotated images nor is it restricted to a specific geometry. We first present a small sequence of RGB-D images displaying a human-object interaction. This sequence is then leveraged to build associated hand and object meshes that represent the depicted interaction. Subsequently, we complete missing parts of the reconstructed object shape and estimate the relative transformation between the reconstruction and the visible object in the scene. Finally, we transfer the a-priori knowledge from the relative pose between object and human hand with the estimate of the current object pose in the scene into necessary grasping instructions for the robot. Exhaustive evaluations with Toyota's Human Support Robot (HSR) in real and synthetic environments demonstrate the applicability of our proposed methodology and its advantage in comparison to previous approaches.

Enis Simsar, Evin Pınar Örnek, Fabian Manhardt et al.

With the advent of deep learning, estimating depth from a single RGB image has recently received a lot of attention, being capable of empowering many different applications ranging from path planning for robotics to computational cinematography. Nevertheless, while the depth maps are in their entirety fairly reliable, the estimates around object discontinuities are still far from satisfactory. This can be contributed to the fact that the convolutional operator naturally aggregates features across object discontinuities, resulting in smooth transitions rather than clear boundaries. Therefore, in order to circumvent this issue, we propose a novel convolutional operator which is explicitly tailored to avoid feature aggregation of different object parts. In particular, our method is based on estimating per-part depth values by means of superpixels. The proposed convolutional operator, which we dub "Instance Convolution", then only considers each object part individually on the basis of the estimated superpixels. Our evaluation with respect to the NYUv2 as well as the iBims dataset clearly demonstrates the superiority of Instance Convolutions over the classical convolution at estimating depth around occlusion boundaries, while producing comparable results elsewhere. Code will be made publicly available upon acceptance.

Helisa Dhamo, Fabian Manhardt, Nassir Navab et al.

Controllable scene synthesis consists of generating 3D information that satisfy underlying specifications. Thereby, these specifications should be abstract, i.e. allowing easy user interaction, whilst providing enough interface for detailed control. Scene graphs are representations of a scene, composed of objects (nodes) and inter-object relationships (edges), proven to be particularly suited for this task, as they allow for semantic control on the generated content. Previous works tackling this task often rely on synthetic data, and retrieve object meshes, which naturally limits the generation capabilities. To circumvent this issue, we instead propose the first work that directly generates shapes from a scene graph in an end-to-end manner. In addition, we show that the same model supports scene modification, using the respective scene graph as interface. Leveraging Graph Convolutional Networks (GCN) we train a variational Auto-Encoder on top of the object and edge categories, as well as 3D shapes and scene layouts, allowing latter sampling of new scenes and shapes.

Yan Di, Fabian Manhardt, Gu Wang et al.

Directly regressing all 6 degrees-of-freedom (6DoF) for the object pose (e.g. the 3D rotation and translation) in a cluttered environment from a single RGB image is a challenging problem. While end-to-end methods have recently demonstrated promising results at high efficiency, they are still inferior when compared with elaborate P$n$P/RANSAC-based approaches in terms of pose accuracy. In this work, we address this shortcoming by means of a novel reasoning about self-occlusion, in order to establish a two-layer representation for 3D objects which considerably enhances the accuracy of end-to-end 6D pose estimation. Our framework, named SO-Pose, takes a single RGB image as input and respectively generates 2D-3D correspondences as well as self-occlusion information harnessing a shared encoder and two separate decoders. Both outputs are then fused to directly regress the 6DoF pose parameters. Incorporating cross-layer consistencies that align correspondences, self-occlusion and 6D pose, we can further improve accuracy and robustness, surpassing or rivaling all other state-of-the-art approaches on various challenging datasets.

Gu Wang, Fabian Manhardt, Jianzhun Shao et al.

6D object pose estimation is a fundamental problem in computer vision. Convolutional Neural Networks (CNNs) have recently proven to be capable of predicting reliable 6D pose estimates even from monocular images. Nonetheless, CNNs are identified as being extremely data-driven, and acquiring adequate annotations is oftentimes very time-consuming and labor intensive. To overcome this shortcoming, we propose the idea of monocular 6D pose estimation by means of self-supervised learning, removing the need for real annotations. After training our proposed network fully supervised with synthetic RGB data, we leverage recent advances in neural rendering to further self-supervise the model on unannotated real RGB-D data, seeking for a visually and geometrically optimal alignment. Extensive evaluations demonstrate that our proposed self-supervision is able to significantly enhance the model's original performance, outperforming all other methods relying on synthetic data or employing elaborate techniques from the domain adaptation realm.

Fabian Manhardt, Gu Wang, Benjamin Busam et al.

Contemporary monocular 6D pose estimation methods can only cope with a handful of object instances. This naturally hampers possible applications as, for instance, robots seamlessly integrated in everyday processes necessarily require the ability to work with hundreds of different objects. To tackle this problem of immanent practical relevance, we propose a novel method for class-level monocular 6D pose estimation, coupled with metric shape retrieval. Unfortunately, acquiring adequate annotations is very time-consuming and labor intensive. This is especially true for class-level 6D pose estimation, as one is required to create a highly detailed reconstruction for all objects and then annotate each object and scene using these models. To overcome this shortcoming, we additionally propose the idea of synthetic-to-real domain transfer for class-level 6D poses by means of self-supervised learning, which removes the burden of collecting numerous manual annotations. In essence, after training our proposed method fully supervised with synthetic data, we leverage recent advances in differentiable rendering to self-supervise the model with unannotated real RGB-D data to improve latter inference. We experimentally demonstrate that we can retrieve precise 6D poses and metric shapes from a single RGB image.

Fabian Manhardt, Wadim Kehl, Adrien Gaidon

We present a deep learning method for end-to-end monocular 3D object detection and metric shape retrieval. We propose a novel loss formulation by lifting 2D detection, orientation, and scale estimation into 3D space. Instead of optimizing these quantities separately, the 3D instantiation allows to properly measure the metric misalignment of boxes. We experimentally show that our 10D lifting of sparse 2D Regions of Interests (RoIs) achieves great results both for 6D pose and recovery of the textured metric geometry of instances. This further enables 3D synthetic data augmentation via inpainting recovered meshes directly onto the 2D scenes. We evaluate on KITTI3D against other strong monocular methods and demonstrate that our approach doubles the AP on the 3D pose metrics on the official test set, defining the new state of the art.

Fabian Manhardt, Diego Martin Arroyo, Christian Rupprecht et al.

3D object detection and pose estimation from a single image are two inherently ambiguous problems. Oftentimes, objects appear similar from different viewpoints due to shape symmetries, occlusion and repetitive textures. This ambiguity in both detection and pose estimation means that an object instance can be perfectly described by several different poses and even classes. In this work we propose to explicitly deal with this uncertainty. For each object instance we predict multiple pose and class outcomes to estimate the specific pose distribution generated by symmetries and repetitive textures. The distribution collapses to a single outcome when the visual appearance uniquely identifies just one valid pose. We show the benefits of our approach which provides not only a better explanation for pose ambiguity, but also a higher accuracy in terms of pose estimation.

Fabian Manhardt, Wadim Kehl, Nassir Navab et al.

We present a novel approach for model-based 6D pose refinement in color data. Building on the established idea of contour-based pose tracking, we teach a deep neural network to predict a translational and rotational update. At the core, we propose a new visual loss that drives the pose update by aligning object contours, thus avoiding the definition of any explicit appearance model. In contrast to previous work our method is correspondence-free, segmentation-free, can handle occlusion and is agnostic to geometrical symmetry as well as visual ambiguities. Additionally, we observe a strong robustness towards rough initialization. The approach can run in real-time and produces pose accuracies that come close to 3D ICP without the need for depth data. Furthermore, our networks are trained from purely synthetic data and will be published together with the refinement code to ensure reproducibility.

Tomas Hodan, Frank Michel, Eric Brachmann et al.

We propose a benchmark for 6D pose estimation of a rigid object from a single RGB-D input image. The training data consists of a texture-mapped 3D object model or images of the object in known 6D poses. The benchmark comprises of: i) eight datasets in a unified format that cover different practical scenarios, including two new datasets focusing on varying lighting conditions, ii) an evaluation methodology with a pose-error function that deals with pose ambiguities, iii) a comprehensive evaluation of 15 diverse recent methods that captures the status quo of the field, and iv) an online evaluation system that is open for continuous submission of new results. The evaluation shows that methods based on point-pair features currently perform best, outperforming template matching methods, learning-based methods and methods based on 3D local features. The project website is available at bop.felk.cvut.cz.

Wadim Kehl, Fabian Manhardt, Federico Tombari et al.

We present a novel method for detecting 3D model instances and estimating their 6D poses from RGB data in a single shot. To this end, we extend the popular SSD paradigm to cover the full 6D pose space and train on synthetic model data only. Our approach competes or surpasses current state-of-the-art methods that leverage RGB-D data on multiple challenging datasets. Furthermore, our method produces these results at around 10Hz, which is many times faster than the related methods. For the sake of reproducibility, we make our trained networks and detection code publicly available.