3D scene reconstruction and understanding have gained increasing popularity, yet existing methods struggle to capture fine-grained, language-aware 3D representations from$2 D$images. In this paper, we present GALA, a novel framework for open-vocabulary 3D scene understanding with 3D Gaussian Splatting (3DGS). GALA distills a scenespecific 3D instance feature field via self-supervised contrastive learning. To further extend this to generalized language feature fields, we introduce a core contribution of GALA, a cross-attention module with two learnable codebooks that encode view-independent semantic embeddings. This design not only ensures intra-instance feature similarity but also supports seamless 2D and 3D open-vocabulary queries. It reduces memory consumption by avoiding perGaussian high-dimensional feature learning. Extensive experiments on real-world datasets demonstrate GALA's remarkable open-vocabulary performance on both$2 D$and 3D.

3D Gaussian Splatting has recently gained traction for its efficient training and real-time rendering. While its vanilla representation is mainly designed for view synthesis, recent works extended it to scene understanding with language features. However, storing additional high-dimensional features per Gaussian for semantic information is memoryintensive, which limits their ability to segment and interpret challenging scenes. To this end, we introduce SuperGSeg, a novel approach that fosters cohesive, context-aware hierarchical scene representation by disentangling segmentation and language field distillation. SuperGSeg first employs neural 3D Gaussians to learn geometry, instance and hierarchical segmentation features from multi-view images with the aid of off-the-shelf$2 D$masks. These features are then leveraged to create a sparse set of Super-Gaussians. Super-Gaussians facilitate the lifting and distillation of$2 D$language features into 3D space. They enable hierarchical scene understanding with high-dimensional language feature rendering at moderate GPU memory costs. Extensive experiments demonstrate that SuperGSeg achieves remarkable performance on both open-vocabulary object selection and semantic segmentation tasks. More results at supergseg.github.io.

We tackle the problem of monocular-to-stereo video conversion and propose a novel architecture for inpainting and refinement of the warped right view obtained by depth-based reprojection of the input left view. We extend the Stable Video Diffusion (SVD) model to utilize the input left video, the warped right video, and the disocclusion masks as conditioning input to generate a high-quality right camera view. In order to effectively exploit information from neighboring frames for inpainting, we modify the attention layers in SVD to compute full attention for discoccluded pixels. Our model is trained to generate the right view video in an end-to-end manner without iterative diffusion steps by minimizing image space losses to ensure high-quality generation. Our approach outperforms previous state-of-theart methods, being ranked best$2.6 \times$more often than the second-place method in a user study, while being$6 \times$faster.

Minimal parametrization of 3D lines plays a critical role in camera localization and structural mapping. Existing representations in robotics and computer vision predominantly handle independent lines, overlooking structural regularities such as sets of parallel lines that are pervasive in man-made environments. This paper introduces RiemanLine, a unified minimal representation for 3D lines formulated on Riemannian manifolds that jointly accommodates both individual lines and parallel-line groups. Our key idea is to decouple each line landmark into global and local components: a shared vanishing direction optimized on the unit sphere, and scaled normal vectors constrained on orthogonal subspaces, enabling compact encoding of structural regularities. For n parallel lines, the proposed representation reduces the parameter space from 4n (orthonormal form) to 2n+2, naturally embedding parallelism without explicit constraints. We further integrate this parameterization into a factor graph framework, allowing global direction alignment and local reprojection optimization within a unified manifold-based bundle adjustment. Extensive experiments on ICL-NUIM, TartanAir, and synthetic benchmarks demonstrate that our method achieves significantly more accurate pose estimation and line reconstruction, while reducing parameter dimensionality and improving convergence stability.

Generating realistic 3D scenes is an area of growing interest in computer vision and robotics. However, creating high-quality, diverse synthetic 3D content often requires expert intervention, making it costly and complex. Recently, efforts to automate this process with learning techniques, particularly diffusion models, have shown significant improvements in tasks like furniture rearrangement. However, applying diffusion models to floor-conditioned indoor scene synthesis remains under-explored. This task is especially challenging as it requires arranging objects in continuous space while selecting from discrete object categories, posing unique difficulties for conventional diffusion methods. To bridge this gap, we present MiDiffusion, a novel mixed discrete-continuous diffusion model designed to synthesize plausible 3D indoor scenes given a floor plan and pre-arranged objects. We represent a scene layout by a 2D floor plan and a set of objects, each defined by category, location, size, and orientation. Our approach uniquely applies structured corruption across mixed discrete semantic and continuous geometric domains, resulting in a better-conditioned problem for denoising. Evaluated on the 3D-FRONT dataset, MiDiffusion outperforms state-of-the-art autoregressive and diffusion models in floor-conditioned 3D scene synthesis. Additionally, it effectively handles partial object constraints via a corruption-and-masking strategy without task-specific training, demonstrating advantages in scene completion and furniture arrangement tasks.

Sparse-view novel view synthesis is fundamentally ill-posed due to severe geometric ambiguity. Current methods are caught in a trade-off: regressive models are geometrically faithful but incomplete, whereas generative models can complete scenes but often introduce structural inconsistencies. We propose OracleGS, a novel framework that reconciles generative completeness with regressive fidelity for sparse view Gaussian Splatting. Instead of using generative models to patch incomplete reconstructions, our "propose-and-validate" framework first leverages a pre-trained 3D-aware diffusion model to synthesize novel views to propose a complete scene. We then repurpose a multi-view stereo (MVS) model as a 3D-aware oracle to validate the 3D uncertainties of generated views, using its attention maps to reveal regions where the generated views are well-supported by multi-view evidence versus where they fall into regions of high uncertainty due to occlusion, lack of texture, or direct inconsistency. This uncertainty signal directly guides the optimization of a 3D Gaussian Splatting model via an uncertainty-weighted loss. Our approach conditions the powerful generative prior on multi-view geometric evidence, filtering hallucinatory artifacts while preserving plausible completions in under-constrained regions, outperforming state-of-the-art methods on datasets including Mip-NeRF 360 and NeRF Synthetic.

Recent advances in view synthesis and real-time rendering have achieved photorealistic quality at impressive ren-dering 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 suf-fer 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 opti-mization 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 radi-ance 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.

State-of-the-art novel view synthesis methods achieve impressive results for multi-view captures of static 3D scenes. However, the reconstructed scenes still lack “liveliness,“ a key component for creating engaging 3D experiences. Recently, novel video diffusion models generate realistic videos with complex motion and enable animations of 2D images, however they cannot naively be used to animate 3D scenes as they lack multi-view consistency. To breathe life into the static world, we propose Gaussians2Life, a method for animating parts of high-quality 3D scenes in a Gaussian Splatting representation. Our key idea is to leverage powerful video diffusion models as the generative component of our model and to combine these with a robust technique to lift 2D videos into meaningful 3D motion. We find that, in contrast to prior work, this enables realistic animations of complex, pre-existing 3D scenes and further enables the animation of a large variety of object classes, while related work is mostly focused on prior-based character animation, or single 3D objects. Our model enables the creation of consistent, immersive 3D experiences for arbitrary scenes.

Foundational vision-language models like CLIP are emerging as a promising paradigm in vision due to their excellent generalization. However, adapting these models for downstream tasks while maintaining their generalization remains challenging. In literature, one branch of methods adapts CLIP by learning prompts using images. While effective, these methods often rely on image-label data, which is not always practical, and struggle to generalize to new datasets due to overfitting on few-shot source data. Another approach explores training-free methods by generating class captions from large language models (LLMs) and performing prompt ensembling, but these methods often produce static, class-specific prompts that cannot be transferred to new classes and incur additional costs by generating LLM descriptions for each class separately. In this work, we aim to combine the strengths of both approaches by learning prompts using only text data derived from LLMs. As supervised training of prompts in the image-free setup is non-trivial, we develop a language-only efficient training approach that enables prompts to distill rich contextual knowledge from LLM data. Furthermore, by mapping the LLM contextual text data within the learned prompts, our approach enables zero-shot transfer of prompts to new classes and datasets, potentially reducing the LLM prompt engineering cost. To the best of our knowledge, this is the first work that learns generalized and transferable prompts for image tasks using only text data. We perform evaluations on 4 benchmarks, where ProText improves over ensembling methods while being competitive with those using labeled images.

Shape completion, a crucial task in 3D computer vision, involves predicting and filling the missing regions of scanned or partially observed objects. Current methods expect known pose or canonical coordinates and do not perform well under varying rotations, limiting their real-world applicability. We introduce ESCAPE (Equivariant Shape Completion via Anchor Point Encoding), a novel framework designed to achieve rotation-equivariant shape completion. Our approach employs a distinctive encoding strategy by selecting anchor points from a shape and representing all points as a distance to all anchor points. This enables the model to capture a consistent, rotation-equivariant understanding of the object’s geometry. ESCAPE leverages a transformer architecture to encode and decode the distance transformations, ensuring that generated shape completions remain accurate and equivariant under rotational transformations. Subsequently, we perform optimization to calculate the predicted shapes from the encodings. Experimental evaluations demonstrate that ESCAPE achieves robust, high-quality reconstructions across arbitrary rotations and translations, showcasing its effectiveness in real-world applications without additional pose estimation modules. Our code is publicly available.1

Neural radiance fields are an emerging 3D scene representation and recently even been extended to learn features for scene understanding by distilling open-vocabulary features from vision-language models. However, current method primarily focus on object-centric representations, supporting object segmentation or detection, while understanding semantic relationships between objects remains largely unexplored. To address this gap, we propose RelationField, the first method to extract inter-object relationships directly from neural radiance fields. RelationField represents relationships between objects as pairs of rays within a neural radiance field, effectively extending its formulation to include implicit relationship queries. To teach RelationField complex, open-vocabulary relationships, relationship knowledge is distilled from multi-modal LLMs. To evaluate RelationField, we solve open-vocabulary 3D scene graph generation tasks and relationship-guided instance segmentation, achieving state-of-the-art performance in both tasks. See the project website at relationfield.github.io.

6D object pose estimation remains challenging for many applications due to dependencies on complete 3D models, multi-view images, or training limited to specific object categories. These requirements make generalization to novel objects difficult for which neither 3D models nor multi-view images may be available. To address this, we propose a novel method One2Any that estimates the relative 6-degrees of freedom (DOF) object pose using only a single reference-single query RGB-D image, without prior knowledge of its 3D model, multi-view data, or category constraints. We treat object pose estimation as an encoding-decoding process: first, we obtain a comprehensive Reference Object Pose Embedding (ROPE) that encodes an object’s shape, orientation, and texture from a single reference view. Using this embedding, a U-Net-based pose decoding module produces Reference Object Coordinate (ROC) for new views, enabling fast and accurate pose estimation. This simple encoding-decoding framework allows our model to be trained on any pair-wise pose data, enabling large-scale training and demonstrating great scalability. Experiments on multiple benchmark datasets demonstrate that our model generalizes well to novel objects, achieving state-of-the-art accuracy and robustness even rivaling methods that require multi-view or CAD inputs, at a fraction of compute. Code is available at https://github.com/lmy1001/One2Any.

Unseen object pose estimation methods often rely on CAD models or multiple reference views, making the onboarding stage costly. To simplify reference acquisition, we aim to estimate the unseen object’s pose through a single unposed RGB-D reference image. While previous works leverage reference images as pose anchors to limit the range of relative pose, our scenario presents significant challenges since the relative transformation could vary across the entire SE(3) space. Moreover, factors like occlusion, sensor noise, and extreme geometry could result in low viewpoint overlap. To address these challenges, we present a novel approach and benchmark, termed UNOPose1, for UNseen One-reference-based object Pose estimation. Building upon a coarse-to-fine paradigm, UNOPose constructs an SE(3)-invariant reference frame to standardize object representation despite pose and size variations. To alleviate small overlap across viewpoints, we recalibrate the weight of each correspondence based on its predicted likelihood of being within the overlapping region. Evaluated on our proposed benchmark based on the BOP Challenge, UNOPose demonstrates superior performance, significantly outperforming traditional and learning-based methods in the one-reference setting and remaining competitive with CAD-model-based methods. The code and dataset are available at github.com/shanice-l/UNOPose.

Understanding fine-grained temporal dynamics is crucial in egocentric videos, where continuous streams capture frequent, close-up interactions with objects. In this work, we bring to light that current egocentric video question-answering datasets often include questions that can be answered using only few frames or commonsense reasoning, without being necessarily grounded in the actual video. Our analysis shows that state-of-the-art Multi-Modal Large Language Models (MLLMs) on these benchmarks achieve remarkably high performance using just text or a single frame as input. To address these limitations, we introduce EgoTempo, a dataset specifically designed to evaluate temporal understanding in the egocentric domain. EgoTempo emphasizes tasks that require integrating information across the entire video, ensuring that models would need to rely on temporal patterns rather than static cues or pre-existing knowledge. Extensive experiments on EgoTempo show that current MLLMs still fall short in temporal reasoning on egocentric videos, and thus we hope EgoTempo will catalyze new research in the field and inspire models that better capture the complexity of temporal dynamics. Dataset and code are available at https://github.com/google-research-datasets/egotempo.git.

Open-vocabulary semantic segmentation models associate vision and text to label pixels from an undefined set of classes using textual queries, providing versatile performance on novel datasets. However, large shifts between training and test domains degrade their performance, requiring fine-tuning for effective real-world applications. We introduce Semantic Library Adaptation (SemLA), a novel framework for training-free, test-time domain adaptation. SemLA leverages a library of LoRA-based adapters indexed with CLIP embeddings, dynamically merging the most relevant adapters based on proximity to the target domain in the embedding space. This approach constructs an ad-hoc model tailored to each specific input without additional training. Our method scales efficiently, enhances explainability by tracking adapter contributions, and inherently protects data privacy, making it ideal for sensitive applications. Comprehensive experiments on a 20-domain benchmark built over 10 standard datasets demonstrate SemLA ’s superior adaptability and performance across diverse settings, establishing a new standard in domain adaptation for open-vocabulary semantic segmentation.

Recent advances in text-to-image customization have enabled high-fidelity, context-rich generation of personalized images, allowing specific concepts to appear in a variety of scenarios. However, current methods struggle with combining multiple personalized models, often leading to attribute entanglement or requiring separate training to preserve concept distinctiveness. We present LoRACLR, a novel approach for multi-concept image generation that merges multiple LoRA models, each fine-tuned for a distinct concept, into a single, unified model without additional individual fine-tuning. LoRACLR uses a contrastive objective to align and merge the weight spaces of these models, ensuring compatibility while minimizing interference. By enforcing distinct yet cohesive representations for each concept, LoRACLR enables efficient, scalable model composition for high-quality, multi-concept image synthesis. Our results highlight the effectiveness of LoRACLR in accurately merging multiple concepts, advancing the capabilities of personalized image generation.

Spatial reasoning is a fundamental aspect of human cognition, enabling intuitive understanding and manipulation of objects in three-dimensional space. While foundation models demonstrate remarkable performance on some benchmarks, they still struggle with 3D reasoning tasks like arranging objects in space according to open-ended language instructions, particularly in dense and physically constrained environments. We introduce LayoutVLM, a framework and scene layout representation that exploits the semantic knowledge of Vision-Language Models (VLMs) and supports differentiable optimization to ensure physical plausibility. LayoutVLM employs VLMs to generate two mutually reinforcing representations from visually marked images, and a self-consistent decoding process to improve VLMs spatial planning. Our experiments show that LayoutVLM addresses the limitations of existing LLM and constraint-based approaches, producing physically plausible 3D layouts better aligned with the semantic intent of input language instructions. We also demonstrate that fine-tuning VLMs with the proposed scene layout representation extracted from existing scene datasets can improve their reasoning performance.

Knowledge distillation (KD) is the de facto standard for compressing large-scale multimodal models into smaller ones. Prior works have explored ever more complex KD strategies involving different objectives, teacher-ensembles, and weight inheritance. In this work, we explore an alternative, yet simple approach—active data curation as effective distillation for contrastive multimodal pretraining. Our simple online batch selection method, ACID, outperforms strong KD baselines across various model-,data-and compute-configurations. Further, we find such an active curation strategy to in fact be complementary to standard KD, and can be effectively combined to train highly performant inference-efficient models. Our simple and scalable pretraining framework, ACED, achieves state-of-the-art results across 27 zero-shot classification and image-text retrieval tasks with upto 11% less inference FLOPs. We further demonstrate that ACED yields strong vision-encoders for training generative multimodal models, outperforming larger vision encoders on image-captioning and visual question-answering tasks.

Visual in-context learning (VICL), as a new paradigm in computer vision, allows the model to rapidly adapt to various tasks with only a handful of prompts and examples. While effective, the existing VICL paradigm exhibits poor generalizability under distribution shifts. In this work, we propose test-time Visual In-Context Tuning (VICT), a method that can adapt VICL models on the fly with a single test sample. Specifically, we flip the role between the task prompts and the test sample and use a cycle consistency loss to reconstruct the original task prompt output. Our key insight is that a model should be aware of a new test distribution if it can successfully recover the original task prompts. Extensive experiments on six representative vision tasks ranging from high-level visual understanding to low-level image processing, with 15 common corruptions, demonstrate that our VICT can improve the generalizability of VICL to unseen new domains. In addition, we show the potential of applying VICT for unseen tasks at test time. Code: https://github.com/Jiahao000/VICT.

Simultaneously localizing camera poses and constructing Gaussian radiance fields in dynamic scenes establish a crucial bridge between 2D images and the 4D real world. Instead of removing dynamic objects as distractors and reconstructing only static environments, this paper proposes an efficient architecture that incrementally tracks camera poses and establishes the 4D Gaussian radiance fields in unknown scenarios by using a sequence of RGB-D images. First, by generating motion masks, we obtain static and dynamic priors for each pixel. To eliminate the influence of static scenes and improve the efficiency on learning the motion of dynamic objects, we classify the Gaussian primitives into static and dynamic Gaussian sets, while the sparse control points along with an MLP is utilized to model the transformation fields of the dynamic Gaussians. To more accurately learn the motion of dynamic Gaussians, a novel 2D optical flow map reconstruction algorithm is designed to render optical flows of dynamic objects between neighbor images, which are further used to supervise the 4D Gaussian radiance fields along with traditional photometric and geometric constraints. In experiments, qualitative and quantitative evaluation results show that the proposed method achieves robust tracking and high-quality view synthesis performance in real-world environments.

In panoptic segmentation, individual instances must be separated within semantic classes. As state-of-the-art methods rely on a pre-defined set of classes, they struggle with novel categories and out-of-distribution (OOD) data. This is particularly problematic in safety-critical applications, such as autonomous driving, where reliability in unseen scenarios is essential. We address the gap between outstanding benchmark performance and reliability by proposing Prior2Former (P2F), the first approach for segmentation vision transformers rooted in evidential learning. P2F extends the mask vision transformer architecture by incorporating a Beta prior for computing model uncertainty in pixel-wise binary mask assignments. This design enables high-quality uncertainty estimation that effectively detects novel and OOD objects enabling state-of-the-art anomaly instance segmentation and open-world panoptic segmentation. Unlike most segmentation models addressing unknown classes, P2F operates without access to OOD data samples or contrastive training on void (i.e., unlabeled) classes, making it highly applicable in real-world scenarios where such prior information is unavailable. Additionally, P2F can be flexibly applied to anomaly instance and panoptic segmentation. Through comprehensive experiments on the Cityscapes, COCO, SegmentMeIfYouCan, and OoDIS datasets, P2F demonstrates state-of-the-art performance across the board.

Scaling up model size and training data has advanced foundation models for instance-level perception, achieving state-of-the-art in-domain and zero-shot performance across object detection and segmentation. However, their high computational cost limits adoption on resource-constrained platforms. We first examine the limitations of existing architectures in enabling efficient edge deployment without compromising performance. We then introduce MOBIUS, a family of foundation models for universal instance segmentation, designed for Pareto-optimal downscaling to support deployment across devices ranging from high-end accelerators to mobile hardware. To reduce training and inference demands, we propose: (i) a bottleneck pixel decoder for efficient multi-scale and multi-modal fusion, (ii) a language-guided uncertainty calibration loss for adaptive decoder pruning, and (iii) a streamlined, unified training strategy. Unlike efficient baselines that trade accuracy for reduced complexity, MOBIUS reduces pixel and transformer decoder FLOPs by up to 55% and 75%, respectively, while maintaining state-of-the-art performance in just a third of the training iterations. MOBIUS establishes a new benchmark for efficient segmentation on both high-performance computing platforms and mobile devices.

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.

Transformers have become the predominant architecture in foundation models due to their excellent performance across various domains. However, the substantial cost of scaling these models remains a significant concern. This problem arises primarily from their dependence on a fixed number of parameters within linear projections. When architectural modifications (e.g., channel dimensions) are introduced, the entire model typically requires retraining from scratch. As model sizes continue growing, this strategy results in increasingly high computational costs and becomes unsustainable. To overcome this problem, we introduce Tokenformer, a natively scalable architecture that leverages the attention mechanism not only for computations among input tokens but also for interactions between tokens and model parameters, thereby enhancing architectural flexibility. By treating model parameters as tokens, we replace all the linear projections in Transformers with our token-parameter attention layer, where input tokens act as queries and model parameters as keys and values. This reformulation allows for progressive and efficient scaling without necessitating retraining from scratch. Our model scales from 124M to 1.4B parameters by incrementally adding new key-value parameter pairs, achieving performance comparable to Transformers trained from scratch while greatly reducing training costs. Code and models are available at {\color{red}\url{https://github.com/Haiyang-W/TokenFormer.git}}

This paper proposes a versatile graph-based lifelong localization framework using LiDAR, LiLoc, which enhances its timeliness by maintaining a single central session while improves the accuracy through multi-modal factors between the central and subsidiary sessions. First, an adaptive submap joining strategy is employed to generate prior submaps (keyframes and poses) for the central session, and to provide priors for subsidiaries when constraints are needed for robust localization. Next, a coarse-to-fine pose initialization for subsidiary sessions is performed using vertical recognition and ICP refinement in the global coordinate frame. To elevate the accuracy of subsequent localization, we propose an egocentric factor graph (EFG) module that integrates the IMU preintegration, LiDAR odometry and scan match factors in a joint optimization manner. Specifically, the scan match factors are constructed by a novel propagation model that efficiently distributes the prior constrains as edges to the relevant prior pose nodes, weighted by noises based on keyframe registration errors. Additionally, the framework supports flexible switching between two modes: relocalization (RLM) and incremental localization (ILM) based on the proposed overlap-based mechanism to select or update the prior submaps from central session. The proposed LiLoc is tested on public and custom datasets, demonstrating accurate localization performance against state-of-the-art methods. Our codes will be publicly available on https://github.com/Yixin-F/LiLoc.

Automating the expert-dependent and labor-intensive task of 3D scene synthesis would significantly benefit fields such as architectural design, robotics simulation, and virtual reality. Recent approaches to 3D scene synthesis often rely on the commonsense reasoning of large language models (LLMs) or strong visual priors from image generation models. However, current LLMs exhibit limited 3D spatial reasoning, undermining the realism and global coherence of synthesized scenes, while image-generation-based methods often constrain viewpoint control and introduce multi-view inconsistencies. In this work, we present Video Perception models for 3D Scene synthesis (VIPScene), a novel framework that exploits the encoded commonsense knowledge of the 3D physical world in video generation models to ensure coherent scene layouts and consistent object placements across views. VIPScene accepts both text and image prompts and seamlessly integrates video generation, feedforward 3D reconstruction, and open-vocabulary perception models to semantically and geometrically analyze each object in a scene. This enables flexible scene synthesis with high realism and structural consistency. For a more sufficient evaluation on coherence and plausibility, we further introduce First-Person View Score (FPVScore), utilizing a continuous first-person perspective to capitalize on the reasoning ability of multimodal large language models. Extensive experiments show that VIPScene significantly outperforms existing methods and generalizes well across diverse scenarios.

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.

Learning effective multi-modal 3D representations of objects is essential for numerous applications, such as augmented reality and robotics. Existing methods often rely on task-specific embeddings that are tailored either for semantic understanding or geometric reconstruction. As a result, these embeddings typically cannot be decoded into explicit geometry and simultaneously reused across tasks. In this paper, we propose Object-X, a versatile multi-modal object representation framework capable of encoding rich object embeddings (e.g., images, point cloud, text) and decoding them back into detailed geometric and visual reconstructions. Object-X operates by geometrically grounding the captured modalities in a 3D voxel grid and learning an unstructured embedding fusing the information from the voxels with the object attributes. The learned embedding enables 3D Gaussian Splatting-based object reconstruction, while also supporting a range of downstream tasks, including scene alignment, single-image 3D object reconstruction, and localization. Evaluations on two challenging real-world datasets demonstrate that Object-X produces high-fidelity novel-view synthesis comparable to standard 3D Gaussian Splatting, while significantly improving geometric accuracy. Moreover, Object-X achieves competitive performance with specialized methods in scene alignment and localization. Critically, our object-centric descriptors require 3-4 orders of magnitude less storage compared to traditional image- or point cloud-based approaches, establishing Object-X as a scalable and highly practical solution for multi-modal 3D scene representation.

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.

Large-scale machine learning models deliver strong performance across a wide range of tasks but come with significant computational and resource constraints. To mitigate these challenges, local smaller models are often deployed alongside larger models, relying on routing and deferral mechanisms to offload complex tasks. However, existing approaches inadequately balance the capabilities of these models, often resulting in unnecessary deferrals or sub-optimal resource usage. In this work, we introduce a novel loss function called Gatekeeper for calibrating smaller models in cascade setups. Our approach fine-tunes the smaller model to confidently handle tasks it can perform correctly while deferring complex tasks to the larger model. Moreover, it incorporates a mechanism for managing the trade-off between model performance and deferral accuracy and is broadly applicable across various tasks and domains without any architectural changes. We evaluated our method on encoder-only, decoder-only, and encoder-decoder architectures. Experiments across image classification, language modeling, and vision-language tasks show that our approach substantially improves deferral performance.

Reconstructing the geometry and appearance of a given scene is a fundamental task in 3D computer graphics and computer vision. Recently, radiance fields have emerged as a representation of light transport in the scene, allowing, as a byproduct, also to extract 3D geometry solely from multi-view imagery. Initially designed for RGB captures, existing approaches have been extended to other sensor modalities. Among these, transient imaging — measuring the time-of-flight of light at picosecond resolution — has emerged as a promising alternative, offering rich spatio-temporal information to improve reconstruction quality from limited viewpoints and obstructed views. However, its applicability to outdoor scenarios has been highly problematic due to interference from ambient light and the different sensor behavior under high-photon-flux conditions typical of outdoor settings. Addressing this gap, we introduce Transient LASSO, a neural scene reconstruction method operating on raw transient measures of outdoor in-the-wild captures to accurately reconstruct the underlying scene geometry and properties. We demonstrate the effectiveness of our method across a variety of outdoor environments, including complex urban scenes with dense traffic and infrastructure. Finally, we also show the potential use cases of our method for downstream applications such as sensor parameter optimization.

For autonomous vehicles, driving safely is highly dependent on the capability to correctly perceive the environment in the 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. To compensate for this need, we propose a novel approach to self-supervise 3D object detection purely from RGB video sequences, leveraging geometric constraints and weak labels. Unlike other approaches that exploit additional sensors during training, our method relies on the temporal continuity of video sequences. A supervised pre-training on synthetic data produces initial plausible 3D boxes, then our geometric and photometrically grounded losses provide a strong self-supervision signal that allows the model to be fine-tuned on real data without labels. Our experiments on Autonomous Driving benchmark datasets showcase the effectiveness and generality of our approach and the competitive performance compared to other self-supervised approaches. • Self-supervised 3D object detection from RGB videos using geometric constraints. • Temporal continuity repaces additional sensors in training. • Self-supervision enables fine-tuning on real data without labels.

Open-vocabulary semantic segmentation enables models to identify novel object categories beyond their training data. While this flexibility represents a significant advancement, current approaches still rely on manually specified class names as input, creating an inherent bottleneck in real-world applications. This work proposes a Vocabulary-Free Semantic Segmentation pipeline, eliminating the need for predefined class vocabularies. Specifically, we address the chicken-and-egg problem where users need knowledge of all potential objects within a scene to identify them, yet the purpose of segmentation is often to discover these objects. The proposed approach leverages Vision–Language Models to automatically recognize objects and generate appropriate class names, aiming to solve the challenge of class specification and naming quality. Through extensive experiments on several public datasets, we highlight the crucial role of the text encoder in model performance, particularly when the image text classes are paired with generated descriptions. Despite the challenges introduced by the sensitivity of the segmentation text encoder to false negatives within the class tagging process, which adds complexity to the task, we demonstrate that our fully automated pipeline significantly enhances vocabulary-free segmentation accuracy across diverse real-world scenarios. Code is available at https://github.com/klarareichard/open-vocab2free-seg . • Propose a novel two-stage pipeline using an image tagger and a class-specific decoder. • Setting a new benchmark for Vocabulary-Free Semantic Segmentation. • Show the impact of enriched text inputs on the encoder assuming a perfect tagger. • Analyze the influence of undetected objects and false detections on the segmentation.

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.

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.

Our objective is to leverage a differentiable radiance field e.g. NeRF to reconstruct detailed 3D surfaces in addition to producing the standard novel view renderings. There have been related methods that perform such tasks, usually by utilizing a signed distance field (SDF). However, the state-of-the-art approaches still fail to correctly reconstruct the small-scale details, such as the leaves, ropes, and textile surfaces. Considering that different methods formulate and optimize the projection from SDF to radiance field with a globally constant Eikonal regularization, we improve with a ray-wise weighting factor to prioritize the rendering and zero-crossing surface fitting on top of establishing a perfect SDF. We propose to adaptively adjust the regularization on the signed distance field so that unsatisfying rendering rays won’t enforce strong Eikonal regularization which is ineffective, and allow the gradients from regions with well-learned radiance to effectively back-propagated to the SDF. Consequently, balancing the two objectives in order to generate accurate and detailed surfaces. Additionally, concerning whether there is a geometric bias between the zero-crossing surface in SDF and rendering points in the radiance field, the projection becomes adjustable as well depending on different 3D locations during optimization. Our proposed RaNeuS1are extensively evaluated on both synthetic and real datasets, achieving state-of-the-art results on both novel view synthesis and geometric reconstruction.1Codes are released at https://github.com/wangyida/ra-neus.

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 cap-turing this variation. To address this issue, we present Sec-ondPose, a novel approach integrating object-specific ge-ometric features with semantic category priors from DI-NOv2. 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 represen-tation under SE(3) transformations, facilitating the map-ping from camera space to the pre-defined canonical space, thus further enhancing pose estimation. Extensive exper-iments 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.

Existing 3D scene understanding methods are heavily focused on 3D semantic and instance segmentation. However, identifying objects and their parts only constitutes an intermediate step towards a more fine-grained goal, which is effectively interacting with the functional interactive elements (e.g., handles, knobs, buttons) in the scene to accomplish diverse tasks. To this end, we introduce SceneFun3D, a large-scale dataset with more than 14. Bk highly accurate interaction annotations for 710 high-resolution real-world 3D indoor scenes. We accompany the annotations with motion parameter information, describing how to interact with these elements, and a diverse set of natural language descriptions of tasks that involve manipulating them in the scene context. To showcase the value of our dataset, we introduce three novel tasks, namely functionality segmentation, task-driven affordance grounding and 3D motion estimation, and adapt existing state-of-the-art methods to tackle them. Our experiments show that solving these tasks in real 3D scenes remains challenging despite recent progress in closed-set and open-set 3D scene understanding methods.

3D shape generation from text is a fundamental task in 3D representation learning. The text-shape pairs exhibit a hierarchical structure, where a general text like “chair” covers all 3D shapes of the chair, while more detailed prompts refer to more specific shapes. Furthermore, both text and 3D shapes are inherently hierarchical structures. However, existing Text2Shape methods, such as SDFusion, do not exploit that. In this work, we propose HyperSD-Fusion, a dual-branch diffusion model that generates 3D shapes from a given text. Since hyperbolic space is suitable for handling hierarchical data, we propose to learn the hierarchical representations of text and 3D shapes in hyperbolic space. First, we introduce a hyperbolic text-image encoder to learn the sequential and multi-modal hierarchical features of text in hyperbolic space. In addition, we design a hyperbolic text-graph convolution module to learn the hierarchical features of text in hyperbolic space. In order to fully utilize these text features, we introduce a dual-branch structure to embed text features in 3D feature space. At last, to endow the generated 3D shapes with a hierarchical structure, we devise a hyperbolic hierarchical loss. Our method is the first to explore the hyperbolic hierarchical representation for text-to-shape generation. Experimental results on the existing text-to-shape paired dataset, Text2Shape, achieved state-of-the-art results. We release our implementation under HyperSDFusion.github.io.

Recovering the 3D scene geometry from a single view is a fundamental yet ill-posed problem in computer vision. While classical depth estimation methods infer only a 2.5D scene representation limited to the image plane, recent approaches based on radiance fields reconstruct a full 3D representation. However, these methods still struggle with occluded regions since inferring geometry without visual observation requires (i) semantic knowledge of the surroundings, and (ii) reasoning about spatial context. We propose KYN, a novel method for single-view scene reconstruction that reasons about semantic and spatial context to predict each point's density. We introduce a vision-language modulation module to enrich point features with fine-grained semantic information. We aggregate point representations across the scene through a language-guided spatial attention mechanism to yield per-point density predictions aware of the 3D semantic context. We show that KYN improves 3D shape recovery compared to predicting density for each 3D point in isolation. We achieve state-of-the-art results in scene and object reconstruction on KITTI-360, and show improved zero-shot generalization compared to prior work. Project page: https://ruili3.github.io/kyn.

Images produced by text-to-image diffusion models might not always faithfully represent the semantic intent of the provided text prompt, where the model might overlook or entirely fail to produce certain objects. Existing solutions often require customly tailored functions for each of these problems, leading to sub-optimal results, especially for complex prompts. Our work introduces a novel perspective by tackling this challenge in a contrastive context. Our approach intuitively promotes the segregation of objects in attention maps while also maintaining that pairs of related attributes are kept close to each other. We conduct extensive experiments across a wide variety of scenarios, each involving unique combinations of objects, attributes, and scenes. These experiments effectively showcase the versatil-ity, efficiency, and flexibility of our method in working with both latent and pixel-based diffusion models, including Sta-ble Diffusion and Imagen. Moreover, we publicly share our source code to facilitate further research.

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 pretraining 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. Sec-ond, in the real-world finetuning stage, MOHO leverages the amodal-mask-weighted geometric supervision to mitigate the unfaithful guidance caused by the hand-occluded su-pervising 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.

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 ex-isting 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 ob-jects by dynamically aggregating deformation-aware local-global features around extracted keypoints. Objects that are close in the embedding space are considered similar in ge-ometry. Then we introduce the neural cage-based defor-mation 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. Exten-sive 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 will be released in https://github.com/lolrudy/KP-RED.

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.

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.

Vision-Language Models (VLMs) have made remarkable progress in document-based Visual Question Answering (i.e., responding to queries about the contents of an input document provided as an image). In this work, we show these models can memorize responses for training samples and regurgitate them even when the relevant visual information has been removed. This includes Personal Identifiable Information (PII) repeated once in the training set, indicating these models could divulge memorised sensitive information and therefore pose a privacy risk. We quantitatively measure the extractability of information in controlled experiments and differentiate between cases where it arises from generalization capabilities or from memorization. We further investigate the factors that influence memorization across multiple state-of-the-art models and propose an effective heuristic countermeasure that empirically prevents the extractability of PII.

In robotics, it’s crucial to understand object deformation during tactile interactions. A precise understanding of deformation can elevate robotic simulations and have broad implications across different industries. We introduce a method using Physics-Encoded Graph Neural Networks (GNNs) for such predictions. Similar to robotic grasping and manipulation scenarios, we focus on modeling the dynamics between a rigid mesh contacting a deformable mesh under external forces. Our approach represents both the soft body and the rigid body within graph structures, where nodes hold the physical states of the meshes. We also incorporate cross-attention mechanisms to capture the interplay between the objects. By jointly learning geometry and physics, our model reconstructs consistent and detailed deformations. We’ve made our code and dataset public to advance research in robotic simulation and grasping.†

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.

Estimating the pose of objects through vision is essential to make robotic platforms interact with the environment. Yet, it presents many challenges, often related to the lack of flexibility and generalizability of state-of-the-art solutions. Diffusion models are a cutting-edge neural architecture transforming 2D and 3D computer vision, outlining remarkable performances in zero-shot novel-view synthesis. Such a use case is particularly intriguing for reconstructing 3D objects. However, localizing objects in unstructured environments is rather unexplored. To this end, this work presents Zero123-6D, the first work to demonstrate the utility of Diffusion Model-based novel-view-synthesizers in enhancing RGB 6D pose estimation at category-level, by integrating them with feature extraction techniques. Novel View Synthesis allows to obtain a coarse pose that is refined through an online optimization method introduced in this work to deal with intra-category geometric differences. In such a way, the outlined method shows reduction in data requirements, removal of the necessity of depth information in zero-shot category-level 6D pose estimation task, and increased performance, quantitatively demonstrated through experiments on the CO3D dataset.

Autonomous vehicles demand detailed maps to maneuver reliably through traffic, which need to be kept up-to-date to ensure a safe operation. A promising way to adapt the maps to the ever-changing road-network is to use crowd-sourced data from a fleet of vehicles. In this work, we present a mapping system that fuses local submaps gathered from a fleet of vehicles at a central instance to produce a coherent map of the road environment including drivable area, lane markings, poles, obstacles and more as a 3D mesh. Each vehicle contributes locally reconstructed submaps as lightweight meshes, making our method applicable to a wide range of reconstruction methods and sensor modalities. Our method jointly aligns and merges the noisy and incomplete local submaps using a scene-specific Neural Signed Distance Field, which is supervised using the submap meshes to predict a fused environment representation. We leverage memory-efficient sparse feature-grids to scale to large areas and introduce a confidence score to model uncertainty in scene reconstruction. Our approach is evaluated on two datasets with different local mapping methods, showing improved pose alignment and reconstruction over existing methods. Additionally, we demonstrate the benefit of multi-session mapping and examine the required amount of data to enable high-fidelity map learning for autonomous vehicles.

In recent years, coordinate-based neural implicit representations have shown promising results for the task of Simultaneous Localization and Mapping (SLAM). While achieving impressive performance on small synthetic scenes, these methods often suffer from losing details, especially for complex real-world scenes. In this work, we introduce DNS SLAM, a novel neural RGB-D semantic SLAM approach featuring a hybrid representation. Relying only on 2D semantic priors, we propose the first semantic neural SLAM method that trains class-wise scene representations while providing stable camera tracking at the same time. Our method integrates multi-view geometry constraints with image-based feature extraction to improve appearance details and to output color, occupancy, and semantic class information, enabling many downstream applications. To further enable fast tracking, we introduce a lightweight coarse scene representation which is trained in a self-supervised manner in latent space. Our experimental results achieve state-of-the-art performance on both synthetic data and real-world data tracking while maintaining a commendable operational speed on off-the-shelf hardware. Further, our method outputs class-wise decomposed reconstructions with better texture, capturing appearance and geometric details.

Neural 3D scene representations have shown great potential for 3D reconstruction from 2D images. However, reconstructing real-world captures of complex scenes still remains a challenge. Existing generic 3D reconstruction methods often struggle to represent fine geometric details and do not adequately model reflective surfaces of large-scale scenes. Techniques that explicitly focus on reflective surfaces can model complex and detailed reflections by exploiting better reflection parameterizations. However, we observe that these methods are often not robust in real scenarios where non-reflective as well as reflective components are present. In this work, we propose UniSDF, a general purpose 3D reconstruction method that can reconstruct large complex scenes with reflections. We investigate both camera view as well as reflected view-based color parameterization techniques and find that explicitly blending these representations in 3D space enables reconstruction of surfaces that are more geometrically accurate, especially for reflective surfaces. We further combine this representation with a multi-resolution grid backbone that is trained in a coarse-to-fine manner, enabling faster reconstructions than prior methods. Extensive experiments on object-level datasets DTU, Shiny Blender as well as unbounded datasets Mip-NeRF 360 and Ref-NeRF real demonstrate that our method is able to robustly reconstruct complex large-scale scenes with fine details and reflective surfaces, leading to the best overall performance. Project page: https://fangjinhuawang.github.io/UniSDF.

Text-to-image models are becoming increasingly popular, revolutionizing the landscape of digital art creation by enabling highly detailed and creative visual content generation. These models have been widely employed across various domains, particularly in art generation, where they facilitate a broad spectrum of creative expression and democratize access to artistic creation. In this paper, we introduce STYLEBREEDER, a comprehensive dataset of 6.8M images and 1.8M prompts generated by 95K users on Artbreeder, a platform that has emerged as a significant hub for creative exploration with over 13M users. We introduce a series of tasks with this dataset aimed at identifying diverse artistic styles, generating personalized content, and recommending styles based on user interests. By documenting unique, user-generated styles that transcend conventional categories like 'cyberpunk' or 'Picasso,' we explore the potential for unique, crowd-sourced styles that could provide deep insights into the collective creative psyche of users worldwide. We also evaluate different personalization methods to enhance artistic expression and introduce a style atlas, making these models available in LoRA format for public use. Our research demonstrates the potential of text-to-image diffusion models to uncover and promote unique artistic expressions, further democratizing AI in art and fostering a more diverse and inclusive artistic community. The dataset, code, and models are available at https://stylebreeder.github.io under a Public Domain (CC0) license.

Gaussian splatting has become a popular representation for novel-view synthesis, exhibiting clear strengths in efficiency, photometric quality, and compositional edibility. Following its success, many works have extended Gaussians to 4D, showing that dynamic Gaussians maintain these benefits while also tracking scene geometry far better than alternative representations. Yet, these methods assume dense multi-view videos as supervision. In this work, we are interested in extending the capability of Gaussian scene representations to casually captured monocular videos. We show that existing 4D Gaussian methods dramatically fail in this setup because the monocular setting is underconstrained. Building off this finding, we propose a method we call Dynamic Gaussian Marbles, which consist of three core modifications that target the difficulties of the monocular setting. First, we use isotropic Gaussian "marbles'', reducing the degrees of freedom of each Gaussian. Second, we employ a hierarchical divide and-conquer learning strategy to efficiently guide the optimization towards solutions with globally coherent motion. Finally, we add image-level and geometry-level priors into the optimization, including a tracking loss that takes advantage of recent progress in point tracking. By constraining the optimization, Dynamic Gaussian Marbles learns Gaussian trajectories that enable novel-view rendering and accurately capture the 3D motion of the scene elements. We evaluate on the Nvidia Dynamic Scenes dataset and the DyCheck iPhone dataset, and show that Gaussian Marbles significantly outperforms other Gaussian baselines in quality, and is on-par with non-Gaussian representations, all while maintaining the efficiency, compositionality, editability, and tracking benefits of Gaussians. Our project page can be found here https://geometry.stanford.edu/projects/dynamic-gaussian-marbles.github.io/.

Abstract Since real-world training datasets cannot properly sample the long tail of the underlying data distribution, corner cases and rare out-of-domain samples can severely hinder the performance of state-of-the-art models. This problem becomes even more severe for dense tasks, such as 3D semantic segmentation, where points of non-standard objects can be confidently associated to the wrong class. In this work, we focus on improving the generalization to out-of-domain data. We achieve this by augmenting the training set with adversarial examples. First, we learn a set of vectors that deform the objects in an adversarial fashion. To prevent the adversarial examples from being too far from the existing data distribution, we preserve their plausibility through a series of constraints, ensuring sensor-awareness and shapes smoothness. Then, we perform adversarial augmentation by applying the learned sample-independent vectors to the available objects when training a model. We conduct extensive experiments across a variety of scenarios on data from KITTI, Waymo, and CrashD for 3D object detection, and on data from SemanticKITTI, Waymo, and nuScenes for 3D semantic segmentation. Despite training on a standard single dataset, our approach substantially improves the robustness and generalization of both 3D object detection and 3D semantic segmentation methods to out-of-domain data.

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.

Deep learning approaches process data in a layer-by-layer way with intermediate (or latent) features. We aim at designing a general solution to optimize the latent manifolds to improve the performance on classification, segmentation, completion and/or reconstruction through probabilistic models. This paper proposes a variational inference model which leads to a clustered embedding. We introduce additional variables in the latent space, called nebula anchors, that guide the latent variables to form clusters during training. To prevent the anchors from clustering among themselves, we employ the variational constraint that enforces the latent features within an anchor to form a Gaussian distribution, resulting in a generative model we refer as Nebula Variational Coding (NVC). Since each latent feature can be labeled with the closest anchor, we also propose to apply metric learning in a self-supervised way to make the separation between clusters more explicit. As a consequence, the latent variables of our variational coder form clusters which adapt to the generated semantic of the training data, e.g., the categorical labels of each sample. We demonstrate experimentally that it can be used within different architectures designed to solve different problems including text sequence, images, 3D point clouds and volumetric data, validating the advantage of our proposed method.

Recent works have shown that unstructured text (doc-uments) from online sources can serve as useful auxiliary information for zero-shot image classification. However, these methods require access to a high-quality source like Wikipedia and are limited to a single source of information. Large Language Models (LLM) trained on web-scale text show impressive abilities to repurpose their learned knowledge for a multitude of tasks. In this work, we provide a novel perspective on using an LLM to provide text supervision for a zero-shot image classification model. The LLM is provided with a few text descriptions from different annota-tors as examples. The LLM is conditioned on these exam-ples to generate multiple text descriptions for each class (re-ferred to as views). Our proposed model, I2MVFormer, learns multi-view semantic embeddings for zero-shot image classification with these class views. We show that each text view of a class provides complementary information allowing a model to learn a highly discriminative class embed-ding. Moreover, we show that I2MVFormer is better at consuming the multi-view text supervision from LLM compared to baseline models. I2MVFormer establishes a new state-of-the-art on three public benchmark datasets for zero-shot image classification with unsupervised semantic embeddings. Code available at https://github.com/ferjad/I2DFormer

Neural Radiance Fields (NeRF) coupled with CANs represent a promising direction in the area of 3D reconstruction from a single view, owing to their ability to efficiently model arbitrary topologies. Recent work in this area, however, has mostly focused on synthetic datasets where exact ground-truth poses are known, and has overlooked pose estimation, which is important for certain down-stream applications such as augmented reality (AR) and robotics. We introduce a principled end-to-end reconstructionframeworkfor natural images, where accurate ground-truth poses are not available. Our approach recovers an SDF-parameterized 3D shape, pose, and appearance from a single image of an object, without exploiting multiple views during training. More specifically, we leverage an unconditional 3D-aware generator, to which we apply a hybrid inversion scheme where a model produces a first guess of the solution which is then refined via optimization. Our frame-work can de-render an image in as few as 10 steps, enabling its use in practical scenarios. We demonstrate state-of-the-art results on a variety of real and synthetic benchmarks.

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 multiview 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.

3D semantic scene graphs are a powerful holistic representation as they describe the individual objects and depict the relation between them. They are compact high-level graphs that enable many tasks requiring scene reasoning. In real-world settings, existing 3D estimation methods produce robust predictions that mostly rely on dense inputs. In this work, we propose a real-time framework that incrementally builds a consistent 3D semantic scene graph of a scene given an RGB image sequence. Our method consists of a novel incremental entity estimation pipeline and a scene graph prediction network. The proposed pipeline simultaneously reconstructs a sparse point map and fuses entity estimation from the input images. The proposed network estimates 3D semantic scene graphs with iterative message passing using multi-view and geometric features extracted from the scene entities. Extensive experiments on the 3RScan dataset show the effectiveness of the proposed method in this challenging task, outperforming state-of-the-art approaches. Our implementation is available at https://shunchengwu.github.io/MonoSSG.

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.

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.

Panoptic segmentation methods assign a known class to each pixel given in input. Even for state-of-the-art approaches, this inevitably enforces decisions that systematically lead to wrong predictions for objects outside the training categories. However, robustness against out-of-distribution samples and corner cases is crucial in safety-critical settings to avoid dangerous consequences. Since real-world datasets cannot contain enough data points to adequately sample the long tail of the underlying distribution, models must be able to deal with unseen and unknown scenarios as well. Previous methods targeted this by re-identifying already-seen unlabeled objects. In this work, we propose the necessary step to extend segmentation with a new setting which we term holistic segmentation. Holistic segmentation aims to identify and separate objects of unseen, unknown categories into instances without any prior knowledge about them while performing panoptic segmentation of known classes. We tackle this new problem with U3HS, which finds unknowns as highly uncertain regions and clusters their corresponding instance-aware embeddings into individual objects. By doing so, for the first time in panoptic segmentation with unknown objects, our U3HS is trained without unknown categories, reducing assumptions and leaving the settings as unconstrained as in real-life scenarios. Extensive experiments on public data from MS COCO, Cityscapes, and Lost&Found demonstrate the effectiveness of U3HS for this new, challenging, and assumptions-free setting called holistic segmentation. Project page: https://holisticseg.github.io.

While state-of-the-art monocular depth estimation approaches achieve impressive results in ideal settings, they are highly unreliable under challenging illumination and weather conditions, such as at nighttime or in the presence of rain. In this paper, we uncover these safety-critical issues and tackle them with md4all: a simple and effective solution that works reliably under both adverse and ideal conditions, as well as for different types of learning supervision. We achieve this by exploiting the efficacy of existing methods under perfect settings. Therefore, we provide valid training signals independently of what is in the input. First, we generate a set of complex samples corresponding to the normal training ones. Then, we train the model by guiding its self- or full-supervision by feeding the generated samples and computing the standard losses on the corresponding original images. Doing so enables a single model to recover information across diverse conditions without modifications at inference time. Extensive experiments on two challenging public datasets, namely nuScenes and Oxford RobotCar, demonstrate the effectiveness of our techniques, outperforming prior works by a large margin in both standard and challenging conditions. Source code and data are available at: https://md4all.github.io.

Continual Learning aims to learn a single model on a sequence of tasks without having access to data from previous tasks. The biggest challenge in the domain still remains catastrophic forgetting: a loss in performance on seen classes of earlier tasks. Some existing methods rely on an expensive replay buffer to store a chunk of data from previous tasks. This, while promising, becomes expensive when the number of tasks becomes large or data can not be stored for privacy reasons. As an alternative, prompt-based methods have been proposed that store the task information in a learnable prompt pool. This prompt pool instructs a frozen image encoder on how to solve each task. While the model faces a disjoint set of classes in each task in this setting, we argue that these classes can be encoded to the same embedding space of a pre-trained language encoder. In this work, we propose Language Guidance for Prompt-based Continual Learning (LGCL) as a plug-in for prompt-based methods. LGCL is model agnostic and introduces language guidance at the task level in the prompt pool and at the class level on the output feature of the vision encoder. We show with extensive experimentation that LGCL consistently improves the performance of prompt-based continual learning methods to set a new state-of-the art. LGCL achieves these performance improvements without needing any additional learnable parameters.

Reconstructing both objects and hands in 3D from a single RGB image is complex. Existing methods rely on manually defined hand-object constraints in Euclidean space, leading to suboptimal feature learning. Compared with Euclidean space, hyperbolic space better preserves the geometric properties of meshes thanks to its exponentially-growing space distance, which amplifies the differences between the features based on similarity. In this work, we propose the first precise hand-object reconstruction method in hyperbolic space, namely Dynamic Hyperbolic Attention Network (DHANet), which leverages intrinsic properties of hyperbolic space to learn representative features. Our method that projects mesh and image features into a unified hyperbolic space includes two modules, i.e. dynamic hyperbolic graph convolution and image-attention hyperbolic graph convolution. With these two modules, our method learns mesh features with rich geometry-image multi-modal information and models better hand-object interaction. Our method provides a promising alternative for fine hand-object reconstruction in hyperbolic space. Extensive experiments on three public datasets demonstrate that our method outperforms most state-of-the-art methods.

By identifying four important components of existing LiDAR-camera 3D object detection methods (LiDAR and camera candidates, transformation, and fusion outputs), we observe that all existing methods either find dense candidates or yield dense representations of scenes. However, given that objects occupy only a small part of a scene, finding dense candidates and generating dense representations is noisy and inefficient. We propose SparseFusion, a novel multi-sensor 3D detection method that exclusively uses sparse candidates and sparse representations. Specifically, SparseFusion utilizes the outputs of parallel detectors in the LiDAR and camera modalities as sparse candidates for fusion. We transform the camera candidates into the LiDAR coordinate space by disentangling the object representations. Then, we can fuse the multi-modality candidates in a unified 3D space by a lightweight self-attention module. To mitigate negative transfer between modalities, we propose novel semantic and geometric cross-modality transfer modules that are applied prior to the modality-specific detectors. SparseFusion achieves state-of-the-art performance on the nuScenes benchmark while also running at the fastest speed, even outperforming methods with stronger backbones. We perform extensive experiments to demonstrate the effectiveness and efficiency of our modules and overall method pipeline. Our code will be made publicly available at https://github.com/yichen928/SparseFusion.

Real-time monocular 3D reconstruction is a challenging problem that remains unsolved. Although recent end-to-end methods demonstrate promising results, tiny structures and geometric boundaries are hardly captured due to their insufficient supervision neglecting spatial details and oversimplified feature fusion ignoring temporal cues. To address the problems, we propose an end-to-end 3D reconstruction network SST, which utilizes Sparse estimated points from visual SLAM system as additional Spatial guidance and fuses Temporal features via a cross-modal attention mechanism, achieving more detailed reconstruction results. We propose a Local Spatial-Temporal Fusion module to exploit more informative spatial-temporal cues from multi-view color information and sparse priors, as well a Global Spatial-Temporal Fusion module to refine the local TSDF volumes with the world-frame model from coarse to fine. Extensive experiments on ScanNet and 7-Scenes demonstrate that SST outperforms all state-of-the-art competitors, whilst keeping a high inference speed at 59 FPS, enabling real-world applications with real-time requirements.

6-DoF robotic grasping is a long-lasting but un-solved problem. Recent methods utilize strong 3D networks to extract geometric grasping representations from depth sensors, demonstrating superior accuracy on common objects but performing 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 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 a keypoint heatmap and a 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 annotate and open-source a multi-view grasping dataset in the real world containing 44 sequence collections of mixed photometric complexity with nearly 20M accurate grasping labels.

We introduce the task of open-vocabulary 3D instance segmentation. Current approaches for 3D instance segmentation can typically only recognize object categories from a pre-defined closed set of classes that are annotated in the training datasets. This results in important limitations for real-world applications where one might need to perform tasks guided by novel, open-vocabulary queries related to a wide variety of objects. Recently, open-vocabulary 3D scene understanding methods have emerged to address this problem by learning queryable features for each point in the scene. While such a representation can be directly employed to perform semantic segmentation, existing methods cannot separate multiple object instances. In this work, we address this limitation, and propose OpenMask3D, which is a zero-shot approach for open-vocabulary 3D instance segmentation. Guided by predicted class-agnostic 3D instance masks, our model aggregates per-mask features via multi-view fusion of CLIP-based image embeddings. Experiments and ablation studies on ScanNet200 and Replica show that OpenMask3D outperforms other open-vocabulary methods, especially on the long-tail distribution. Qualitative experiments further showcase OpenMask3D’s ability to segment object properties based on free-form queries describing geometry, affordances, and materials.

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.

Abstract Unsupervised template discovery via implicit representation in a category of shapes has recently shown strong performance. At the core, such methods deform input shapes to a common template space which allows establishing correspondences as well as implicit representation of the shapes. In this work we investigate the inherent assumption that the implicit neural field optimization naturally leads to consistently warped shapes, thus providing both good shape reconstruction and correspondences. Contrary to this convenient assumption, in practice we observe that such is not the case, consequently resulting in sub‐optimal point correspondences. In order to solve the problem, we re‐visit the warp design and more importantly introduce explicit constraints using unsupervised sparse point predictions, directly encouraging consistency of the warped shapes. We use the unsupervised sparse keypoints in order to further condition the deformation warp and enforce the consistency of the deformation warp. Experiments in dynamic non‐rigid DFaust and ShapeNet categories show that our problem identification and solution provide the new state‐of‐the‐art in unsupervised dense correspondences.

The lack of large-scale real datasets with annotations makes transfer learning a necessity for video activity understanding. We aim to develop an effective method for few-shot transfer learning for first-person action classification. We leverage independently trained local visual cues to learn representations that can be transferred from a source domain, which provides primitive action labels, to a different target domain using only a handful of examples. Visual cues we employ include object-object interactions, hand grasps and motion within regions that are a function of hand locations. We employ a framework based on meta-learning to extract the distinctive and domain invariant components of the deployed visual cues. This enables transfer of action classification models across public datasets captured with diverse scene and action configurations. We present comparative results of our transfer learning methodology and report superior results over state-of-the-art action classification approaches for both inter-class and inter-dataset transfer.

Normalizing Flows (NFs) are flexible explicit generative models that have been shown to accurately model complex real-world data distributions. However, their invertibility constraint imposes limitations on data distributions that reside on lower dimensional manifolds embedded in higher dimensional space. Practically, this shortcoming is often bypassed by adding noise to the data which impacts the quality of the generated samples. In contrast to prior work, we approach this problem by generating samples from the original data distribution given full knowledge about the perturbed distribution and the noise model. To this end, we establish that NFs trained on perturbed data implicitly represent the manifold in regions of maximum likelihood. Then, we propose an optimization objective that recovers the most likely point on the manifold given a sample from the perturbed distribution. Finally, we focus on 3D point clouds for which we utilize the explicit nature of NFs, i.e. surface normals extracted from the gradient of the log-likelihood and the log-likelihood itself, to apply Poisson surface re-construction to refine generated point sets.

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.

As 3D object detection on point clouds relies on the geometrical relationships between the points, non-standard object shapes can hinder a method's detection capability. However, in safety-critical settings, robustness to out-of-domain and long-tail samples is fundamental to circumvent dangerous issues, such as the misdetection of damaged or rare cars. In this work, we substantially improve the generalization of 3D object detectors to out-of-domain data by deforming point clouds during training. We achieve this with 3D-VField: a novel data augmentation method that plausibly deforms objects via vector fields learned in an adversarial fashion. Our approach constrains 3D points to slide along their sensor view rays while neither adding nor removing any of them. The obtained vectors are transferable, sample-independent and preserve shape and occlusions. Despite training only on a standard dataset, such as KITTI, augmenting with our vector fields significantly improves the generalization to differently shaped objects and scenes. Towards this end, we propose and share CrashD: a synthetic dataset of realistic damaged and rare cars, with a variety of crash scenarios. Extensive experiments on KITTI, Waymo, our CrashD and SUN RGB-D show the generalizability of our techniques to out-of-domain data, different models and sensors, namely LiDAR and ToF cameras, for both indoor and outdoor scenes. Our CrashD dataset is available at https://crashd-cars.github.io.

Shape matching has been a long-studied problem for the computer graphics and vision community. The objective is to predict a dense correspondence between meshes that have a certain degree of deformation. Existing methods either consider the local description of sampled points or discover correspondences based on global shape information. In this work, we investigate a hierarchical learning design, to which we incorporate local patch-level information and global shape-level structures. This flexible representation enables correspondence prediction and provides rich features for the matching stage. Finally, we propose a novel optimal transport solver by recurrently updating features on non-confident nodes to learn globally consistent correspondences between the shapes. Our results on publicly available datasets suggest robust performance in presence of severe deformations without the need of extensive training or refinement.

Establishing correspondences from image to 3D has been a key task of 6DoF object pose estimation for a long time. To predict pose more accurately, deeply learned dense maps replaced sparse templates. Dense methods also improved pose estimation in the presence of occlusion. More recently researchers have shown improvements by learning object fragments as segmentation. In this work, we present a discrete descriptor, which can represent the object surface densely. By incorporating a hierarchical binary grouping, we can encode the object surface very efficiently. Moreover, we propose a coarse to fine training strategy, which enables fine-grained correspondence prediction. Finally, by matching predicted codes with object surface and using a PnP solver, we estimate the 6DoF pose. Results on the public LM-O and YCB-V datasets show major improvement over the state of the art w.r.t. ADD(-S) metric, even surpassing RGB-D based methods in some cases.

Adapting to a continuously evolving environment is a safety-critical challenge inevitably faced by all autonomous-driving systems. Existing image- and video-based driving datasets, however, fall short of capturing the mutable nature of the real world. In this paper, we introduce the largest multi-task synthetic dataset for autonomous driving, SHIFT. It presents discrete and continuous shifts in cloudiness, rain and fog intensity, time of day, and vehicle and pedestrian density. Featuring a comprehensive sensor suite and annotations for several mainstream perception tasks, SHIFT allows to investigate how a perception systems' performance degrades at increasing levels of domain shift, fostering the development of continuous adaptation strategies to mitigate this problem and assessing the robustness and generality of a model. Our dataset and benchmark toolkit are publicly available at www.vis.xyz/shift.

We propose a novel approach aimed at object and semantic scene completion from a partial scan represented as a 3D point cloud. Our architecture relies on three novel layers that are used successively within an encoder-decoder structure and specifically developed for the task at hand. The first one carries out feature extraction by matching the point features to a set of pre-trained local descriptors. Then, to avoid losing individual descriptors as part of standard operations such as max-pooling, we propose an alternative neighbor-pooling operation that relies on adopting the feature vectors with the highest activations. Finally, upsampling in the decoder modifies our feature extraction in order to increase the output dimension. While this model is already able to achieve competitive results with the state of the art, we further propose a way to increase the versatility of our approach to process point clouds. To this aim, we introduce a second model that assembles our layers within a transformer architecture. We evaluate both architectures on object and indoor scene completion tasks, achieving state-of-the-art performance.

A set of novel approaches for estimating epistemic uncertainty in deep neural networks with a single forward pass has recently emerged as a valid alternative to Bayesian Neural Networks. On the premise of informative representations, these deterministic uncertainty methods (DUMs) achieve strong performance on detecting out-of-distribution (OOD) data while adding negligible computational costs at inference time. However, it remains unclear whether DUMs are well calibrated and can seamlessly scale to real-world applications - both prerequisites for their practical deployment. To this end, we first provide a taxonomy of DUMs, and evaluate their calibration under continuous distributional shifts. Then, we extend them to semantic segmentation. We find that, while DUMs scale to realistic vision tasks and perform well on OOD detection, the practicality of current methods is undermined by poor calibration under distributional shifts.

Processing 3D data efficiently has always been a challenge. Spatial operations on large-scale point clouds, stored as sparse data, require extra cost. Attracted by the success of transformers, researchers are using multi-head attention for vision tasks. However, attention calculations in transformers come with quadratic complexity in the number of inputs and miss spatial intuition on sets like point clouds. We redesign set transformers in this work and incorporate them into a hierarchical framework for shape classification and part and scene segmentation. We propose our local attention unit, which captures features in a spatial neighborhood. We also compute efficient and dynamic global cross attentions by leveraging sampling and grouping at each iteration. Finally, to mitigate the non-heterogeneity of point clouds, we propose an efficient Multi-Scale Tokenization (MST), which extracts scale-invariant tokens for attention operations. The proposed hierarchical model achieves state-of-the-art shape classification in mean accuracy and yields results on par with the previous segmentation methods while requiring significantly fewer computations. Our proposed architecture predicts segmentation labels with around half the latency and parameter count of the previous most effi-cient method with comparable performance. The code is available at https://github.com/YigeWang-WHU/CloudAttention.

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 demon-strate that SSP-Pose produces superior performance compared with competitors with a real-time inference speed at about 25Hz. The codes will be released soon.

Despite the tremendous progress in zero-shot learning (ZSL), the majority of existing methods still rely on human-annotated attributes, which are difficult to annotate and scale. An unsupervised alternative is to represent each class using the word embedding associated with its semantic class name. However, word embeddings extracted from pre-trained language models do not necessarily capture visual similarities, resulting in poor zero-shot performance. In this work, we argue that online textual documents e.g., Wikipedia, contain rich visual descriptions about object classes, therefore can be used as powerful unsupervised side information for ZSL. To this end, we propose I2DFormer, a novel transformer-based ZSL framework that jointly learns to encode images and documents by aligning both modalities in a shared embedding space. In order to distill discriminative visual words from noisy documents, we introduce a new cross-modal attention module that learns fine-grained interactions between image patches and document words. Consequently, our I2DFormer not only learns highly discriminative document embeddings that capture visual similarities but also gains the ability to localize visually relevant words in image regions. Quantitatively, we demonstrate that our I2DFormer significantly outperforms previous unsupervised semantic embeddings under both zero-shot and generalized zero-shot learning settings on three public datasets. Qualitatively, we show that our method leads to highly interpretable results where document words can be grounded in the image regions.

Abstract Recent advances in machine learning have led to increased interest in solving visual computing problems using methods that employ coordinate‐based neural networks. These methods, which we call neural fields , parameterize physical properties of scenes or objects across space and time. They have seen widespread success in problems such as 3D shape and image synthesis, animation of human bodies, 3D reconstruction, and pose estimation. Rapid progress has led to numerous papers, but a consolidation of the discovered knowledge has not yet emerged. We provide context, mathematical grounding, and a review of over 250 papers in the literature on neural fields. In Part I , we focus on neural field techniques by identifying common components of neural field methods, including different conditioning, representation, forward map, architecture, and manipulation methods. In Part II , we focus on applications of neural fields to different problems in visual computing, and beyond (e.g., robotics, audio). Our review shows the breadth of topics already covered in visual computing, both historically and in current incarnations, and highlights the improved quality, flexibility, and capability brought by neural field methods. Finally, we present a companion website that acts as a living database that can be continually updated by the community.

Abstract A 3D scene is more than the geometry and classes of the objects it comprises. An essential aspect beyond object-level perception is the scene context, described as a dense semantic network of interconnected nodes. Scene graphs have become a common representation to encode the semantic richness of images, where nodes in the graph are object entities connected by edges, so-called relationships. Such graphs have been shown to be useful in achieving state-of-the-art performance in image captioning, visual question answering and image generation or editing. While scene graph prediction methods so far focused on images, we propose instead a novel neural network architecture for 3D data, where the aim is to learn to regress semantic graphs from a given 3D scene. With this work, we go beyond object-level perception, by exploring relations between object entities. Our method learns instance embeddings alongside a scene segmentation and is able to predict semantics for object nodes and edges. We leverage 3DSSG, a large scale dataset based on 3RScan that features scene graphs of changing 3D scenes. Finally, we show the effectiveness of graphs as an intermediate representation on a retrieval task.

Abstract We propose a novel convolutional operator for the task of point cloud completion. One striking characteristic of our approach is that, conversely to related work it does not require any max-pooling or voxelization operation. Instead, the proposed operator used to learn the point cloud embedding in the encoder extracts permutation-invariant features from the point cloud via a soft-pooling of feature activations, which are able to preserve fine-grained geometric details. These features are then passed on to a decoder architecture. Due to the compression in the encoder, a typical limitation of this type of architectures is that they tend to lose parts of the input shape structure. We propose to overcome this limitation by using skip connections specifically devised for point clouds, where links between corresponding layers in the encoder and the decoder are established. As part of these connections, we introduce a transformation matrix that projects the features from the encoder to the decoder and vice-versa. The quantitative and qualitative results on the task of object completion from partial scans on the ShapeNet dataset show that incorporating our approach achieves state-of-the-art performance in shape completion both at low and high resolutions.

for learning robust, flexible and generalizable 3D object representations without requiring heavy annotation efforts or supervision. Unlike conventional 3D generative models, our algorithm aims for building a structured latent space where certain factors of shape variations, such as object parts, can be disentangled into independent sub-spaces. Our novel decoder then acts on these individual latent sub-spaces (i.e. capsules) using deconvolution operators to reconstruct 3D points in a self-supervised manner. We further introduce a cluster loss ensuring that the points reconstructed by a single capsule remain local and do not spread across the object uncontrollably. These contributions allow our network to tackle the challenging tasks of part segmentation, part interpolation/replacement as well as correspondence estimation across rigid / non-rigid shape, and across / within category. Our extensive evaluations on ShapeNet objects and human scans demonstrate that our network can learn generic representations that are robust and useful in many applications.

Predicting future events accurately is a task of great importance for autonomous vehicles. In this work we focus on lane change events. For this, we propose a novel attention mechanism on top of recurrent neural networks for the prediction task, which improves performance and yields more interpretable models. As critical corner cases are often not considered and reflected in traditional prediction metrics, we additionally introduce a new scenario-based evaluation scheme, which we posit be considered for further maneuver prediction works. Prediction and planning tasks often are correlated, usually sharing input representations and differing in expected outputs and their subsequent consideration. Here, we detail a supporting layer for planning tasks, which analyzes situations w.r.t. their suitability for lane changes and can serve as decision-making support for any planning algorithm. Exploitation of similarities between this task and the aforementioned prediction problem further improves performance of the prediction task, as well as labelling quality of the assessment task. Additionally, we extend our evaluation to urban scenarios, showcasing the generalizability of our proposed prediction models.

While self-supervised monocular depth estimation in driving scenarios has achieved comparable performance to supervised approaches, violations of the static world assumption can still lead to erroneous depth predictions of traffic participants, posing a potential safety issue. In this paper, we present R4Dyn, a novel set of techniques to use cost-efficient radar data on top of a self-supervised depth estimation framework. In particular, we show how radar can be used during training as weak supervision signal, as well as an extra input to enhance the estimation robustness at inference time. Since automotive radars are readily available, this allows to collect training data from a variety of existing vehicles. Moreover, by filtering and expanding the signal to make it compatible with learning-based approaches, we address radar inherent issues, such as noise and sparsity. With R4Dyn we are able to overcome a major limitation of self-supervised depth estimation, i.e. the prediction of traffic participants. We substantially improve the estimation on dynamic objects, such as cars by 37% on the challenging nuScenes dataset, hence demonstrating that radar is a valuable additional sensor for monocular depth estimation in autonomous vehicles.

Recently Normalizing Flows (NFs) have demonstrated state-of-the-art performance on modeling 3D point clouds while allowing sampling with arbitrary resolution at inference time. However, these flow-based models still have fundamental limitations on complicated geometries. This work generalizes prior work by introducing additional discrete latent variable, i.e. mixture model. This circumvents limitations of prior approaches, leads to more parameter efficient models and reduces the inference runtime. Moreover, in this more general framework each component learns to specialize in a particular subregion of an object in a completely unsupervised fashion yielding promising clustering properties. We further demonstrate that by adding data augmentation, individual mixture components can learn to specialize in a semantically meaningful manner. We evaluate mixtures of NFs on generation, autoencoding and single-view reconstruction based on the ShapeNet dataset.

6D pose estimation from a single RGB image is a fundamental task in computer vision. The current top-performing deep learning-based methods rely on an indirect strategy, i.e., first establishing 2D-3D correspondences between the coordinates in the image plane and object coordinate system, and then applying a variant of the PnP/RANSAC algorithm. However, this two-stage pipeline is not end-to-end trainable, thus is hard to be employed for many tasks requiring differentiable poses. On the other hand, methods based on direct regression are currently inferior to geometry-based methods. In this work, we perform an in-depth investigation on both direct and indirect methods, and propose a simple yet effective Geometry-guided Direct Regression Network (GDR-Net) to learn the 6D pose in an end-to-end manner from dense correspondence-based intermediate geometric representations. Extensive experiments show that our approach remarkably outperforms state-of-the-art methods on LM, LM-O and YCB-V datasets. Code is available at https://git.io/GDR-Net.

Generative models able to synthesize layouts of different kinds (e.g. documents, user interfaces or furniture arrangements) are a useful tool to aid design processes and as a first step in the generation of synthetic data, among other tasks. We exploit the properties of self-attention layers to capture high level relationships between elements in a layout, and use these as the building blocks of the well-known Variational Autoencoder (VAE) formulation. Our proposed Variational Transformer Network (VTN) is capable of learning margins, alignments and other global design rules without explicit supervision. Layouts sampled from our model have a high degree of resemblance to the training data, while demonstrating appealing diversity. In an extensive evaluation on publicly available benchmarks for different layout types VTNs achieve state-of-the-art diversity and perceptual quality. Additionally, we show the capabilities of this method as part of a document layout detection pipeline.

In compositional zero-shot learning, the goal is to recognize unseen compositions (e.g. old dog) of observed visual primitives states (e.g. old, cute) and objects (e.g. car, dog) in the training set. This is challenging because the same state can for example alter the visual appearance of a dog drastically differently from a car. As a solution, we propose a novel graph formulation called Compositional Graph Embedding (CGE) that learns image features, compositional classifiers and latent representations of visual primitives in an end-to-end manner. The key to our approach is exploiting the dependency between states, objects and their compositions within a graph structure to enforce the relevant knowledge transfer from seen to unseen compositions. By learning a joint compatibility that encodes semantics between concepts, our model allows for generalization to unseen compositions without relying on an external knowledge base like WordNet. We show that in the challenging generalized compositional zero-shot setting our CGE significantly outperforms the state of the art on MIT-States and UT-Zappos. We also propose a new benchmark for this task based on the recent GQA dataset. Code is available at: https://github.com/ExplainableML/czsl

Scene graphs are a compact and explicit representation successfully used in a variety of 2D scene understanding tasks. This work proposes a method to incrementally build up semantic scene graphs from a 3D environment given a sequence of RGB-D frames. To this end, we aggregate PointNet features from primitive scene components by means of a graph neural network. We also propose a novel attention mechanism well suited for partial and missing graph data present in such an incremental reconstruction scenario. Although our proposed method is designed to run on submaps of the scene, we show it also transfers to entire 3D scenes. Experiments show that our approach outperforms 3D scene graph prediction methods by a large margin and its accuracy is on par with other 3D semantic and panoptic segmentation methods while running at 35Hz.

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.

Directly regressing all 6 degrees-of-freedom (6DoF) for the object pose (i.e. 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 PnP/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.

Despite recent advancements in single-domain or single-object image generation, it is still challenging to generate complex scenes containing diverse, multiple objects and their interactions. Scene graphs, composed of nodes as objects and directed-edges as relationships among objects, offer an alternative representation of a scene that is more semantically grounded than images. We hypothesize that a generative model for scene graphs might be able to learn the underlying semantic structure of real-world scenes more effectively than images, and hence, generate realistic novel scenes in the form of scene graphs. In this work, we explore a new task for the unconditional generation of semantic scene graphs. We develop a deep auto-regressive model called SceneGraphGen which can directly learn the probability distribution over labelled and directed graphs using a hierarchical recurrent architecture. The model takes a seed object as input and generates a scene graph in a sequence of steps, each step generating an object node, followed by a sequence of relationship edges connecting to the previous nodes. We show that the scene graphs generated by SceneGraphGen are diverse and follow the semantic patterns of real-world scenes. Additionally, we demonstrate the application of the generated graphs in image synthesis, anomaly detection and scene graph completion.

The ability to simultaneously track and reconstruct multiple objects moving in the scene is of the utmost importance for robotic tasks such as autonomous navigation and interaction. Virtually all of the previous attempts to map multiple dynamic objects have evolved to store individual objects in separate reconstruction volumes and track the relative pose between them. While simple and intuitive, such formulation does not scale well with respect to the number of objects in the scene and introduces the need for an explicit occlusion handling strategy. In contrast, we propose a map representation that allows maintaining a single volume for the entire scene and all the objects therein. To this end, we introduce a novel multi-object TSDF formulation that can encode multiple object surfaces at any given location in the map. In a multiple dynamic object tracking and reconstruction scenario, our representation allows maintaining accurate reconstruction of surfaces even while they become temporarily occluded by other objects moving in their proximity. We evaluate the proposed TSDF++ formulation on a public synthetic dataset and demonstrate its ability to preserve reconstructions of occluded surfaces when compared to the standard TSDF map representation. Code is available at https://github.com/ethz-asl/tsdf-plusplus.

Lightweight and semantically meaningful environment maps are crucial for many applications in robotics and autonomous driving to facilitate higher-level tasks such as navigation and planning. In this paper we present a novel approach to incrementally build a meaningful and lightweight semantic map directly as a 3D mesh from a monocular or stereo sequence. Our system leverages existing feature-based visual odometry paired with learned depth prediction and semantic image segmentation to identify and reconstruct semantically relevant environment structure. We introduce a probabilistic fusion scheme to incrementally refine and extend a 3D mesh with semantic labels for each face without intermediate voxel-based fusion. To demonstrate its effectiveness, we evaluate our system in outdoor driving scenarios with monocular depth prediction and stereo and present quantitative and qualitative reconstruction results with comparison to ground truth. Our results show that the proposed approach achieves reconstruction quality comparable to current state-of-the-art voxel-based methods while being much more lightweight both in storage and computation.

This work proposes a RGB-D SLAM system specifically designed for structured environments and aimed at improved tracking and mapping accuracy by relying on geometric features that are extracted from the surrounding. Structured environments offer, in addition to points, also an abundance of geometrical features such as lines and planes, which we exploit to design both the tracking and mapping components of our SLAM system. For the tracking part, we explore geometric relationships between these features based on the assumption of a Manhattan World (MW). We propose a decoupling-refinement method based on points, lines, and planes, as well as the use of Manhattan relationships in an additional pose refinement module. For the mapping part, different levels of maps from sparse to dense are reconstructed at a low computational cost. We propose an instance-wise meshing strategy to build a dense map by meshing plane instances independently. The overall performance in terms of pose estimation and reconstruction is evaluated on public benchmarks and shows improved performance compared to state-of-the-art methods. The code is released at https://github.com/yanyan-li/PlanarSLAM.

In this paper, a robust RGB-D SLAM system is proposed to utilize the structural information in indoor scenes, allowing for accurate tracking and efficient dense mapping on a CPU. Prior works have used the Manhattan World (MW) assumption to estimate low-drift camera pose, in turn limiting the applications of such systems. This paper, in contrast, proposes a novel approach delivering robust tracking in MW and non-MW environments. We check orthogonal relations between planes to directly detect Manhattan Frames, modeling the scene as a Mixture of Manhattan Frames. For MW scenes, we decouple pose estimation and provide a novel drift-free rotation estimation based on Manhattan Frame observations. For translation estimation in MW scenes and full camera pose estimation in non-MW scenes, we make use of point, line and plane features for robust tracking in challenging scenes. Additionally, by exploiting plane features detected in each frame, we also propose an efficient surfel-based dense mapping strategy, which divides each image into planar and non-planar regions. Planar surfels are initialized directly from sparse planes in our map while non-planar surfels are built by extracting superpixels. We evaluate our method on public benchmarks for pose estimation, drift and reconstruction accuracy, achieving superior performance compared to other state-of-the-art methods. We will open-source our code in the future.

Accurate and reliable localization is a fundamental requirement for autonomous vehicles to use map information in higher-level tasks such as navigation or planning. In this paper, we present a novel approach to vehicle localization in dense semantic maps, including vectorized high-definition maps or 3D meshes, using semantic segmentation from a monocular camera. We formulate the localization task as a direct image alignment problem on semantic images, which allows our approach to robustly track the vehicle pose in semantically labeled maps by aligning virtual camera views rendered from the map to sequences of semantically segmented camera images. In contrast to existing visual localization approaches, the system does not require additional keypoint features, handcrafted localization landmark extractors or expensive LiDAR sensors. We demonstrate the wide applicability of our method on a diverse set of semantic mesh maps generated from stereo or LiDAR as well as manually annotated HD maps and show that it achieves reliable and accurate localization in real-time.

Synthetic data generation is an appealing approach to generate novel traffic scenarios in autonomous driving. However, deep learning perception algorithms trained solely on synthetic data encounter serious performance drops when they are tested on real data. Such performance drops are commonly attributed to the domain gap between real and synthetic data. Domain adaptation methods that have been applied to mitigate the aforementioned domain gap achieve visually appealing results, but usually introduce semantic inconsistencies into the translated samples. In this work, we propose a novel, unsupervised, end-to-end domain adaptation network architecture that enables semantically consistent sim2real image transfer. Our method performs content disentanglement by employing shared content encoder and fixed style code.

Image synthesis driven by computer graphics achieved recently a remarkable realism, yet synthetic image data generated this way reveals a significant domain gap with respect to real-world data. This is especially true in autonomous driving scenarios, which represent a critical aspect for over-coming utilizing synthetic data for training neural networks. We propose a method based on domain-invariant scene representation to directly synthesize traffic scene imagery without rendering. Specifically, we rely on synthetic scene graphs as our internal representation and introduce an unsupervised neural network architecture for realistic traffic scene synthesis. We enhance synthetic scene graphs with spatial information about the scene and demonstrate the effectiveness of our approach through scene manipulation.

Driven by deep learning, object recognition has recently made a tremendous leap forward. Nonetheless, its accuracy often still suffers from several sources of variation that can be found in real-world images. Some of the most challenging variations are induced by changing lighting conditions. This paper presents a novel approach for tackling brightness variation in the domain of 2D object detection and 6D object pose estimation. Existing works aiming at improving robustness towards different lighting conditions are often grounded on classical computer vision contrast normalisation techniques or the acquisition of large amounts of annotated data in order to achieve invariance during training. While the former cannot generalise well to a wide range of illumination conditions, the latter is neither practical nor scalable. Hence, We propose the usage of Generative Adversarial Networks in order to learn how to normalise the illumination of an input image. Thereby, the generator is explicitly designed to normalise illumination in images so to enhance the object recognition performance. Extensive evaluations demonstrate that leveraging the generated data can significantly enhance the detection performance, outperforming all other state-of-the-art methods. We further constitute a natural extension focusing on white balance variations and introduce a new dataset for evaluation.

3D Point clouds are a rich source of information that enjoy growing popularity in the vision community. However, due to the sparsity of their representation, learning models based on large point clouds is still a challenge. In this work, we introduce Graphite, a GRAPH-Induced feaTure Extraction pipeline, a simple yet powerful feature transform and keypoint detector. Graphite enables intensive down-sampling of point clouds with keypoint detection accompanied by a descriptor. We construct a generic graph-based learning scheme to describe point cloud regions and extract salient points. To this end, we take advantage of 6D pose information and metric learning to learn robust descriptions and keypoints across different scans. We Reformulate the 3D keypoint pipeline with graph neural networks which allow efficient processing of the point set while boosting its descriptive power which ultimately results in more accurate 3D registrations. We demonstrate our lightweight descriptor on common 3D descriptor matching and point cloud registration benchmarks [76], [71] and achieve comparable results with the state of the art. Describing 100 patches of a point cloud and detecting their keypoints takes only 0.018 seconds with our proposed network.

Estimating the 3D shape of an object from a single or multiple images has gained popularity thanks to the recent breakthroughs powered by deep learning. Most approaches regress the full object shape in a canonical pose, possibly extrapolating the occluded parts based on the learned priors. However, their viewpoint invariant technique often discards the unique structures visible from the input images. In contrast, this paper proposes to rely on viewpoint variant reconstructions by merging the visible information from the given views. Our approach is divided into three steps. Starting from the sparse views of the object, we first align them into a common coordinate system by estimating the relative pose between all the pairs. Then, inspired by the traditional voxel carving, we generate an occupancy grid of the object taken from the silhouette on the images and their relative poses. Finally, we refine the initial reconstruction to build a clean 3D model which preserves the details from each viewpoint. To validate the proposed method, we perform a comprehensive evaluation on the ShapeNet reference benchmark in terms of relative pose estimation and 3D shape reconstruction.

Real-time scene reconstruction from depth data inevitably suffers from occlusion, thus leading to incomplete 3D models. Partial reconstructions, in turn, limit the performance of algorithms that leverage them for applications in the context of, e.g., augmented reality, robotic navigation, and 3D mapping. Most methods address this issue by predicting the missing geometry as an offline optimization, thus being incompatible with real-time applications. We propose a framework that ameliorates this issue by performing scene reconstruction and semantic scene completion jointly in an incremental and real-time manner, based on an input sequence of depth maps. Our framework relies on a novel neural architecture designed to process occupancy maps and leverages voxel states to accurately and efficiently fuse semantic completion with the 3D global model. We evaluate the proposed approach quantitatively and qualitatively, demonstrating that our method can obtain accurate 3D semantic scene completion in real-time.

Image manipulation can be considered a special case of image generation where the image to be produced is a modification of an existing image. Image generation and manipulation have been, for the most part, tasks that operate on raw pixels. However, the remarkable progress in learning rich image and object representations has opened the way for tasks such as text-to-image or layout-to-image generation that are mainly driven by semantics. In our work, we address the novel problem of image manipulation from scene graphs, in which a user can edit images by merely applying changes in the nodes or edges of a semantic graph that is generated from the image. Our goal is to encode image information in a given constellation and from there on generate new constellations, such as replacing objects or even changing relationships between objects, while respecting the semantics and style from the original image. We introduce a spatio-semantic scene graph network that does not require direct supervision for constellation changes or image edits. This makes it possible to train the system from existing real-world datasets with no additional annotation effort.

Scene understanding has been of high interest in computer vision. It encompasses not only identifying objects in a scene, but also their relationships within the given context. With this goal, a recent line of works tackles 3D semantic segmentation and scene layout prediction. In our work we focus on scene graphs, a data structure that organizes the entities of a scene in a graph, where objects are nodes and their relationships modeled as edges. We leverage inference on scene graphs as a way to carry out 3D scene understanding, mapping objects and their relationships. In particular, we propose a learned method that regresses a scene graph from the point cloud of a scene. Our novel architecture is based on PointNet and Graph Convolutional Networks (GCN). In addition, we introduce 3DSSG, a semiautomatically generated dataset, that contains semantically rich scene graphs of 3D scenes. We show the application of our method in a domain-agnostic retrieval task, where graphs serve as an intermediate representation for 3D-3D and 2D-3D matching.

Driveable area detection is a key component for various applications in the field of autonomous driving (AD), such as ground-plane detection, obstacle detection and maneuver planning. Additionally, bulky and over-parameterized networks can be easily forgone and replaced with smaller networks for faster inference on embedded systems. The driveable area detection, posed as a two class segmentation task, can be efficiently modeled with slim binary networks. This paper proposes a novel binarized driveable area detection network (binary DAD-Net), which uses only binary weights and activations in the encoder, the bottleneck, and the decoder part. The latent space of the bottleneck is efficiently increased (×32→×16 downsampling) through binary dilated convolutions, learning more complex features. Along with automatically generated training data, the binary DAD-Net outperforms state-of-the-art semantic segmentation networks on public datasets. In comparison to a full-precision model, our approach has a ×14.3 reduced compute complexity on an FPGA and it requires only 0.9MB memory resources. Therefore, commodity SIMD-based AD-hardware is capable of accelerating the binary DAD-Net.

In the autonomous driving area synthetic data is crucial for cover specific traffic scenarios which autonomous vehicle must handle. This data commonly introduces domain gap between synthetic and real domains. In this paper we deploy data augmentation to generate custom traffic scenarios with VRUs in order to improve pedestrian recognition. We provide a pipeline for augmentation of the Cityscapes dataset with virtual pedestrians. In order to improve augmentation realism of the pipeline we reveal a novel generative network architecture for adversarial learning of the data-set lighting conditions. We also evaluate our approach on the tasks of semantic and instance segmentation.

Synthetic data has been applied in many deep learning based computer vision tasks. Limited performance of algorithms trained solely on synthetic data has been approached with domain adaptation techniques such as the ones based on generative adversarial framework. We demonstrate how adversarial training alone can introduce semantic inconsistencies in translated images. To tackle this issue we propose density prematching strategy using KLIEP-based density ratio estimation procedure. Finally, we show that aforementioned strategy improves quality of translated images of underlying method and their usability for the semantic segmentation task in the context of autonomous driving.

The twenty-six papers in this special issue focus on Red Blue Green (RBG)-D vision, an emerging research topic in computer vision, with a number of applications in robotics, entertainment, biometrics and multimedia. Compared to 2D images and 3D data (including depth images, point clouds and meshes), RGB-D images represent both the photometric and geometric information of a scene. Moreover, low-cost consumer depth cameras (e.g., Microsoft Kinect v2, Intel Realsense, Orbbec Astra) can enable realtime applications due to their high acquisition frame-rate. In the last few years, a large number of RGB-D datasets have also been publicly released to tackle various vision tasks. Although remarkable progress has been achieved, several critical problems still remain open. The aim of this special issue is to stimulate researchers from different fields to present their state-of-the-art work, and to provide a cross-fertilization ground for discussions on the next steps in this important research area.

We introduce GFrames, a novel local reference frame (LRF) construction for 3D meshes and point clouds. GFrames are based on the computation of the intrinsic gradient of a scalar field defined on top of the input shape. The resulting tangent vector field defines a repeatable tangent direction of the local frame at each point; importantly, it directly inherits the properties and invariance classes of the underlying scalar function, making it remarkably robust under strong sampling artifacts, vertex noise, as well as non-rigid deformations. Existing local descriptors can directly benefit from our repeatable frames, as we showcase in a selection of 3D vision and shape analysis applications where we demonstrate state-of-the-art performance in a variety of challenging settings.

Person search has recently gained attention as the novel task of finding a person, provided as a cropped sample, from a gallery of non-cropped images, whereby several other people are also visible. We believe that i. person detection and re-identification should be pursued in a joint optimization framework and that ii. the person search should leverage the query image extensively (e.g. emphasizing unique query patterns). However, so far, no prior art realizes this. We introduce a novel query-guided end-to-end person search network (QEEPS) to address both aspects. We leverage a most recent joint detector and re-identification work, OIM [37]. We extend this with i. a query-guided Siamese squeeze-and-excitation network (QSSE-Net) that uses global context from both the query and gallery images, ii. a query-guided region proposal network (QRPN) to produce query-relevant proposals, and iii. a query-guided similarity subnetwork (QSimNet), to learn a query-guided re-identification score. QEEPS is the first end-to-end query-guided detection and re-id network. On both the most recent CUHK-SYSU [37] and PRW [46] datasets, we outperform the previous state-of-the-art by a large margin.

In this paper, we propose 3D point-capsule networks, an auto-encoder designed to process sparse 3D point clouds while preserving spatial arrangements of the input data. 3D capsule networks arise as a direct consequence of our unified formulation of the common 3D auto-encoders. The dynamic routing scheme and the peculiar 2D latent space deployed by our capsule networks bring in improvements for several common point cloud-related tasks, such as object classification, object reconstruction and part segmentation as substantiated by our extensive evaluations. Moreover, it enables new applications such as part interpolation and replacement.

We present a method that tackles the challenge of predicting color and depth behind the visible content of an image. Our approach aims at building up a Layered Depth Image (LDI) from a single RGB input, which is an efficient representation that arranges the scene in layers, including originally occluded regions. Unlike previous work, we enable an adaptive scheme for the number of layers and incorporate semantic encoding for better hallucination of partly occluded objects. Additionally, our approach is object-driven, which especially boosts the accuracy for the occluded intermediate objects. The framework consists of two steps. First, we individually complete each object in terms of color and depth, while estimating the scene layout. Second, we rebuild the scene based on the regressed layers and enforce the recomposed image to resemble the structure of the original input. The learned representation enables various applications, such as 3D photography and diminished reality, all from a single RGB image.

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 these ambiguities. For each object instance we predict multiple 6D pose 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.

We present a sampling-free approach for computing the epistemic uncertainty of a neural network. Epistemic uncertainty is an important quantity for the deployment of deep neural networks in safety-critical applications, since it represents how much one can trust predictions on new data. Recently promising works were proposed using noise injection combined with Monte-Carlo sampling at inference time to estimate this quantity (e.g. Monte-Carlo dropout). Our main contribution is an approximation of the epistemic uncertainty estimated by these methods that does not require sampling, thus notably reducing the computational overhead. We apply our approach to large-scale visual tasks (\ie, semantic segmentation and depth regression) to demonstrate the advantages of our method compared to sampling-based approaches in terms of quality of the uncertainty estimates as well as of computational overhead.

In this work, we introduce the task of 3D object instance re-localization (RIO): given one or multiple objects in an RGB-D scan, we want to estimate their corresponding 6DoF poses in another 3D scan of the same environment taken at a later point in time. We consider RIO a particularly important task in 3D vision since it enables a wide range of practical applications, including AI-assistants or robots that are asked to find a specific object in a 3D scene. To address this problem, we first introduce 3RScan, a novel dataset and benchmark, which features 1482 RGB-D scans of 478 environments across multiple time steps. Each scene includes several objects whose positions change over time, together with ground truth annotations of object instances and their respective 6DoF mappings among re-scans. Automatically finding 6DoF object poses leads to a particular challenging feature matching task due to varying partial observations and changes in the surrounding context. To this end, we introduce a new data-driven approach that efficiently finds matching features using a fully-convolutional 3D correspondence network operating on multiple spatial scales. Combined with a 6DoF pose optimization, our method outperforms state-of-the-art baselines on our newly-established benchmark, achieving an accuracy of 30.58%.

We propose a novel model for 3D semantic completion from a single depth image, based on a single encoder and three separate generators used to reconstruct different geometric and semantic representations of the original and completed scene, all sharing the same latent space. To transfer information between the geometric and semantic branches of the network, we introduce paths between them concatenating features at corresponding network layers. Motivated by the limited amount of training samples from real scenes, an interesting attribute of our architecture is the capacity to supplement the existing dataset by generating a new training dataset with high quality, realistic scenes that even includes occlusion and real noise. We build the new dataset by sampling the features directly from latent space which generates a pair of partial volumetric surface and completed volumetric semantic surface. Moreover, we utilize multiple discriminators to increase the accuracy and realism of the reconstructions. We demonstrate the benefits of our approach on standard benchmarks for the two most common completion tasks: semantic 3D scene completion and 3D object completion.

Highly accurate maps of the road infrastructure are a crucial cornerstone for self-driving cars to enable navigation in complex traffic scenarios. Traditional methods for creating detailed maps of road environments involve expensive survey vehicles that cannot keep up with the frequent changes in the road network. In this paper, we propose a novel method to derive detailed high-definition maps by crowd sourcing data using commodity sensors. Our system uses multi-session feature-based visual SLAM to align submaps recorded by individual vehicles on a central backend server. We reconstruct 3D boundaries of road infrastructure elements such as road markings and road boundaries from semantic object contours detected in keyframes by a neural network. The result is a concise map of semantically meaningful objects suitable both for localization and higher-level planning tasks of automated vehicles. We evaluate our method on real-world data against a globally referenced ground-truth map demonstrating a high level of detail and metric accuracy.

When driving at night, a good illumination of the road ahead is crucial. With autonomous driving at close temporal proximity, this not only concerns human drivers but also autonomous systems capable of controlling the car. To achieve fully autonomous driving, a variety of sensors are integrated into the vehicles. Cameras act as one of the major sensors. However, due to their passivity, cameras cannot see well in the dark. To mitigate this shortcoming, modern cars are equipped with powerful headlights that provide proper illumination of the road ahead while avoiding the dazzling of other traffic participants. To use the headlights' full potential and to also provide advanced light functionality like glare-free high beam, they need to be properly adjusted. After the initial calibration during production, this setting is prone to undesirable degradation, primarily due to mechanical reasons. We present a completely new application of computer vision and machine learning to automatically detect wrongly adjusted headlights by estimating their pitch angle from the images of a vehicle-attached camera for advanced driving assistance systems (ADAS). We show that we can achieve high performance in terms of accuracy and robustness by training a deep neural network in an end-to-end fashion. To demonstrate the benefits of our proposed approach, an additional handcrafted baseline method is implemented.

Synthetic data find application in computer vision tasks for a long time. Limited performance of algorithms trained solely on synthetic data has been approached with domain adaptation techniques such as the ones based on generative adversarial framework [1]. In this work we demonstrate how using adversarial training alone can introduce semantic inconsistencies in refined images. We suggest leveraging available semantic labels from target domain using naive re-sampling approach alongside with adversarial loss. We also show that aforementioned strategy improves quality of translated images of underlying method and their usability for the semantic segmentation task in the context of autonomous driving. This method will be also put in comparison with existing state-of-the-art synthetic to real domain adaptation methods.

3-D object retrieval has been a hot research topic in recent years. Within such a field, view-based approaches are attracting increasing attention because of the flexibility of data representation as well as the reported state-of-the-art performance. One of the most important issues related to view-based 3-D object retrieval is how to learn embedding features that are discriminative across classes while being compactly distributed within each class. In this paper, we analyze the difference between the two tasks of classification and retrieval, and propose a novel way to learn a view-pooling feature via a triplet network. In addition, we propose a new loss, named cube loss, which is able to sample a number of triplets equal to the cube of the samples in a batch. With the new loss, both hard-negative and hard-positive pairs can be effectively investigated. The experimental results on the ModelNet benchmark demonstrate that the proposed method achieves superior performance compared to state-of-the-art approaches.

We propose a method to reconstruct, complete and semantically label a 3D scene from a single input depth image. We improve the accuracy of the regressed semantic 3D maps by a novel architecture based on adversarial learning. In particular, we suggest using multiple adversarial loss terms that not only enforce realistic outputs with respect to the ground truth, but also an effective embedding of the internal features. This is done by correlating the latent features of the encoder working on partial 2.5D data with the latent features extracted from a variational 3D auto-encoder trained to reconstruct the complete semantic scene. In addition, differently from other approaches that operate entirely through 3D convolutions, at test time we retain the original 2.5D structure of the input during downsampling to improve the effectiveness of the internal representation of our model. We test our approach on the main benchmark datasets for semantic scene completion to qualitatively and quantitatively assess the effectiveness of our proposal.

Interaction and collaboration between humans and intelligent machines has become increasingly important as machine learning methods move into real-world applications that involve end users. While much prior work lies at the intersection of natural language and vision, such as image captioning or image generation from text descriptions, less focus has been placed on the use of language to guide or improve the performance of a learned visual processing algorithm. In this paper, we explore methods to flexibly guide a trained convolutional neural network through user input to improve its performance during inference. We do so by inserting a layer that acts as a spatio-semantic guide into the network. This guide is trained to modify the network's activations, either directly via an energy minimization scheme or indirectly through a recurrent model that translates human language queries to interaction weights. Learning the verbal interaction is fully automatic and does not require manual text annotations. We evaluate the method on two datasets, showing that guiding a pre-trained network can improve performance, and provide extensive insights into the interaction between the guide and the CNN.

We present a novel regularization technique for learning local feature descriptors based on statistical information extracted from batches of training samples. With the proposed regularization term, we learn a descriptor distribution in Euclidean space that aims at minimizing the overlap between the distributions of positive pairs and that of negative pairs. The proposed method is able to improve the performance of pairwise and triplet losses with various deep convolution network architectures. This improvement is demonstrated through two different types of architectures, able to obtain state-of-the-art results on the reference benchmark for local feature matching.

We introduce an extension of the Dubins Traveling Salesman Problem with Neighborhoods into the 3D space in which a fixed-wing aerial vehicle is requested to visit a set of target regions while the vehicle motion constraints are satisfied, i.e., the minimum turning radius and maximum climb and dive angles. The primary challenge is to address both the combinatorial optimization part of finding the sequence of target visits and the continuous optimization part of the final trajectory determination. Due to its high complexity, we propose to address both parts of the problem separately by a decoupled approach in which the sequence is determined by a new distance function designed explicitly for the utilized 3D Dubins Airplane model. The final trajectory is then found by a local optimization which improves the solution quality. The proposed approach provides significantly better solutions than using Euclidean distance in the sequencing part of the problem. Moreover, the found solutions are of the competitive quality to the sampling-based algorithm while its computational requirements are about two orders of magnitude lower.

One of the greatest challenges towards fully autonomous cars is the understanding of complex and dynamic scenes. Such understanding is needed for planning of maneuvers, especially those that are particularly frequent such as lane changes. While in recent years advanced driver-assistance systems have made driving safer and more comfortable, these have mostly focused on car following scenarios, and less on maneuvers involving lane changes. In this work we propose a situation assessment algorithm for classifying driving situations with respect to their suitability for lane changing. For this, we propose a deep learning architecture based on a Bidirectional Recurrent Neural Network, which uses Long Short-Term Memory units, and integrates a prediction component in the form of the Intelligent Driver Model. We prove the feasibility of our algorithm on the publicly available NGSIM datasets, where we outperform existing methods.

Recent advances in monocular SLAM have enabled real-time capable systems which run robustly under the assumption of a static environment, but fail in presence of dynamic scene changes and motion, since they lack an explicit dynamic outlier handling. We propose a semantic monocular SLAM framework designed to deal with highly dynamic environments, combining feature-based and direct approaches to achieve robustness under challenging conditions. The proposed approach exploits semantic information extracted from the scene within an explicit probabilistic model, which maximizes the probability for both tracking and mapping to rely on those scene parts that do not present a relative motion with respect to the camera. We show more stable pose estimation in dynamic environments and comparable performance to the state of the art on static sequences on the Virtual KITTI and Synthia datasets.

We propose a new approach for generating training data for the task of drivable area segmentation with deep neural networks (DNN). The impressive progress of deep learning in recent years demonstrated a superior performance of DNNs over traditional machine learning and deterministic algorithms for various tasks. Nevertheless, the acquisition of large-scale datasets with associated ground truth labels still poses an expensive and labor-intensive problem. We contribute to the solution of this problem for the task of road segmentation by proposing an automatic labeling pipeline which leverages a deterministic stereo-based approach for ground plane detection to create large datasets suitable for training neural networks. Based on the popular Cityscapes [1] and KITTI dataset [2] and two off-the-shelf DNNs for semantic segmentation, we show that we can achieve good segmentation results on monocular images, which substantially exceed the performance of the algorithm employed for automatic labeling without the need of any manual annotation.

We propose an efficient and scalable method for incrementally building a dense, semantically annotated 3D map in real-time. The proposed method assigns class probabilities to each region, not each element (e.g., surfel and voxel), of the 3D map which is built up through a robust SLAM framework and incrementally segmented with a geometric-based segmentation method. Differently from all other approaches, our method has a capability of running at over 30Hz while performing all processing components, including SLAM, segmentation, 2D recognition, and updating class probabilities of each segmentation label at every incoming frame, thanks to the high efficiency that characterizes the computationally intensive stages of our framework. By utilizing a specifically designed CNN to improve the frame-wise segmentation result, we can also achieve high accuracy. We validate our method on the NYUv2 dataset by comparing with the state of the art in terms of accuracy and computational efficiency, and by means of an analysis in terms of time and space complexity.

3D Human shape tracking consists in fitting a template model to temporal sequences of visual observations. It usually comprises an association step, that finds correspondences between the model and the input data, and a deformation step, that fits the model to the observations given correspondences. Most current approaches follow the Iterative-Closest-Point (ICP) paradigm, where the association step is carried out by searching for the nearest neighbors. It fails when large deformations occur and errors in the association tend to propagate over time. In this paper, we propose a discriminative alternative for the association, that leverages random forests to infer correspondences in one shot. Regardless the choice of shape parameterizations, being surface or volumetric meshes, we convert 3D shapes to volumetric distance fields and thereby design features to train the forest. We investigate two ways to draw volumetric samples: voxels of regular grids and cells from Centroidal Voronoi Tessellation (CVT). While the former consumes considerable memory and in turn limits us to learn only subject-specific correspondences, the latter yields much less memory footprint by compactly tessellating the interior space of a shape with optimal discretization. This facilitates the use of larger cross-subject training databases, generalizes to different human subjects and hence results in less overfitting and better detection. The discriminative correspondences are successfully integrated to both surface and volumetric deformation frameworks that recover human shape poses, which we refer to as 'tracking-by-detection of 3D human shapes.' It allows for large deformations and prevents tracking errors from being accumulated. When combined with ICP for refinement, it proves to yield better accuracy in registration and more stability when tracking over time. Evaluations on existing datasets demonstrate the benefits with respect to the state-of-the-art.

We present a novel method to track 3D models in color and depth data. To this end, we introduce approximations that accelerate the state-of-the-art in region-based tracking by an order of magnitude while retaining similar accuracy. Furthermore, we show how the method can be made more robust in the presence of depth data and consequently formulate a new joint contour and ICP tracking energy. We present better results than the state-of-the-art while being much faster then most other methods and achieving all of the above on a single CPU core.

Given the recent advances in depth prediction from Convolutional Neural Networks (CNNs), this paper investigates how predicted depth maps from a deep neural network can be deployed for the goal of accurate and dense monocular reconstruction. We propose a method where CNN-predicted dense depth maps are naturally fused together with depth measurements obtained from direct monocular SLAM, based on a scheme that privileges depth prediction in image locations where monocular SLAM approaches tend to fail, e.g. along low-textured regions, and vice-versa. We demonstrate the use of depth prediction to estimate the absolute scale of the reconstruction, hence overcoming one of the major limitations of monocular SLAM. Finally, we propose a framework to efficiently fuse semantic labels, obtained from a single frame, with dense SLAM, so to yield semantically coherent scene reconstruction from a single view. Evaluation results on two benchmark datasets show the robustness and accuracy of our approach.

One-shot pose estimation for tasks such as body joint localization, camera pose estimation, and object tracking are generally noisy, and temporal filters have been extensively used for regularization. One of the most widely-used methods is the Kalman filter, which is both extremely simple and general. However, Kalman filters require a motion model and measurement model to be specified a priori, which burdens the modeler and simultaneously demands that we use explicit models that are often only crude approximations of reality. For example, in the pose-estimation tasks mentioned above, it is common to use motion models that assume constant velocity or constant acceleration, and we believe that these simplified representations are severely inhibitive. In this work, we propose to instead learn rich, dynamic representations of the motion and noise models. In particular, we propose learning these models from data using long shortterm memory, which allows representations that depend on all previous observations and all previous states. We evaluate our method using three of the most popular pose estimation tasks in computer vision, and in all cases we obtain state-of-the-art performance.

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 RGBD 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.

3D object temporal trackers estimate the 3D rotation and 3D translation of a rigid object by propagating the transformation from one frame to the next. To confront this task, algorithms either learn the transformation between two consecutive frames or optimize an energy function to align the object to the scene. The motivation behind our approach stems from a consideration on the nature of learners and optimizers. Throughout the evaluation of different types of objects and working conditions, we observe their complementary nature - on one hand, learners are more robust when undergoing challenging scenarios, while optimizers are prone to tracking failures due to the entrapment at local minima; on the other, optimizers can converge to a better accuracy and minimize jitter. Therefore, we propose to bridge the gap between learners and optimizers to attain a robust and accurate RGB-D temporal tracker that runs at approximately 2 ms per frame using one CPU core. Our work is highly suitable for Augmented Reality (AR), Mixed Reality (MR) and Virtual Reality (VR) applications due to its robustness, accuracy, efficiency and low latency. Aiming at stepping beyond the simple scenarios used by current systems, often constrained by having a single object in the absence of clutter, averting to touch the object to prevent close-range partial occlusion or selecting brightly colored objects to easily segment them individually, we demonstrate the capacity to handle challenging cases under clutter, partial occlusion and varying lighting conditions.

This paper addresses the problem of estimating the depth map of a scene given a single RGB image. We propose a fully convolutional architecture, encompassing residual learning, to model the ambiguous mapping between monocular images and depth maps. In order to improve the output resolution, we present a novel way to efficiently learn feature map up-sampling within the network. For optimization, we introduce the reverse Huber loss that is particularly suited for the task at hand and driven by the value distributions commonly present in depth maps. Our model is composed of a single architecture that is trained end-to-end and does not rely on post-processing techniques, such as CRFs or other additional refinement steps. As a result, it runs in real-time on images or videos. In the evaluation, we show that the proposed model contains fewer parameters and requires fewer training data than the current state of the art, while outperforming all approaches on depth estimation. Code and models are publicly available.

While the main trend of 3D object recognition has been to infer object detection from single views of the scene - i.e., 2.5D data - this work explores the direction on performing object recognition on 3D data that is reconstructed from multiple viewpoints, under the conjecture that such data can improve the robustness of an object recognition system. To achieve this goal, we propose a framework whichreal-time segmentation is able (i) to carry out incremental real-time segmentation of a 3D scene while being reconstructed via Simultaneous Localization And Mapping (SLAM), and (ii) to simultaneously and incrementally carry out 3D object recognition and pose estimation on the reconstructed and segmented 3D representations. Experimental results demonstrate the advantages of our approach with respect to traditional single view-based object recognition and pose estimation approaches, as well as its usefulness in robotic perception and augmented reality applications.

We present an architecture for online, incremental scene modeling which combines a SLAM-based scene understanding framework with semantic segmentation and object pose estimation. The core of this approach comprises a probabilistic inference scheme that predicts semantic labels for object hypotheses at each new frame. From these hypotheses, recognized scene structures are incrementally constructed and tracked. Semantic labels are inferred using a multi-domain convolutional architecture which operates on the image time series and which enables efficient propagation of features as well as robust model registration. To evaluate this architecture, we introduce a large-scale RGB-D dataset JHUSEQ-25 as a new benchmark for the sequence-based scene understanding in complex and densely cluttered scenes. This dataset contains 25 RGB-D video sequences with 100,000 labeled frames in total. We validate our method on this dataset and demonstrate improved performance of semantic segmentation and 6-DoF object pose estimation compared with methods based on the single view.

There are many applications where domain-specific sensing, such as accelerometers, kinematics, or force sensing, provide unique and important information for control or for analysis of motion. However, it is not always the case that these sensors can be deployed or accessed beyond laboratory environments. For example, it is possible to instrument humans or robots to measure motion in the laboratory in ways that it is not possible to replicate in the wild. An alternative, which we explore in this paper, is to address situations where accurate sensing is available while training an algorithm, but for which only video is available for deployment. We present two examples of this sensory substitution methodology. The first variation trains a convolutional neural network to regress real-valued signals, including robot end-effector pose, from video. The second example regresses binary signals derived from accelerometer data which signifies when specific objects are in motion. We evaluate these on the JIGSAWS dataset for robotic surgery training assessment and the 50 Salads dataset for modeling complex structured cooking tasks. We evaluate the trained models for video-based action recognition and show that the trained models provide information that is comparable to the sensory signals they replace.

Pipelines to recognize 3D objects despite clutter and occlusions usually end up with a final verification stage whereby recognition hypotheses are validated or dismissed based on how well they explain sensor measurements. Unlike previous work, we propose a Global Hypothesis Verification (GHV) approach which regards all hypotheses jointly so as to account for mutual interactions. GHV provides a principled framework to tackle the complexity of our visual world by leveraging on a plurality of recognition paradigms and cues. Accordingly, we present a 3D object recognition pipeline deploying both global and local 3D features as well as shape and color. Thereby, and facilitated by the robustness of the verification process, diverse object hypotheses can be gathered and weak hypotheses need not be suppressed too early to trade sensitivity for specificity. Experiments demonstrate the effectiveness of our proposal, which significantly improves over the state-of-art and attains ideal performance (no false negatives, no false positives) on three out of the six most relevant and challenging benchmark datasets.

Local coordinate frame (LCF) is a key component deployed in most 3D descriptors for invariant representations of 3D surfaces. This paper addresses the problem of attaching a LCF to non-rigidly deforming objects, in particular humanoid surfaces, with the application of recovering correspondences between the template model and input data for 3D human motion tracking. We facilitate this by extending two current LCF paradigms for rigid surface matching to the non-rigid case. Such an adaptation is motivated by the assumption that interpolating locally rigid movements often amounts to smooth globally non-rigid deformations. Both approaches leverage spatial distributions, based on signed distance and principal component analysis, respectively. Furthermore, we advocate a new strategy that incorporates multiple LCF candidates. This way we relax the requirement of perfectly repeatable LCFs, and yet still achieve improved data-model associations. Ground truth for non-rigid LCFs are synthetically generated by interpolating locally-rigidly transformed LCFs. Therefore, the proposed methods can be evaluated extensively in terms of repeatability of LCFs, robustness on estimating correspondences, and accuracy of final tracking results. All the experiments demonstrate the benefits of the proposed methods with respect to the state-of-the-art.

We propose a real-time method for 3D head pose estimation from RGB-D sequences. Our algorithm relies on a Random Forest framework that is able to regress the head pose at every frame in a temporal tracking manner. Such framework is learned once from a generic dataset of 3D head models and refined online to adapt the forest to the specific characteristics of each subject. Through the qualitative experiments under different conditions, it demonstrates remarkable properties in terms of robustness to occlusions, computational efficiency and capacity of handling a variety of challenging head poses. In addition, it also outperforms the state of the art on the reference benchmark dataset with regards to the accuracy of the estimated head poses.

We present a scalable method for detecting objects and estimating their 3D poses in RGB-D data. To this end, we rely on an efficient representation of object views and employ hashing techniques to match these views against the input frame in a scalable way. While a similar approach already exists for 2D detection, we show how to extend it to estimate the 3D pose of the detected objects. In particular, we explore different hashing strategies and identify the one which is more suitable to our problem. We show empirically that the complexity of our method is sublinear with the number of objects and we enable detection and pose estimation of many 3D objects with high accuracy while outperforming the state-of-the-art in terms of runtime.

We propose a novel approach to online visual tracking that combines the robustness of sparse coding with the flexibility of voting-based methods. Our algorithm relies on a dictionary that is learned once and for all from a large set of training patches extracted from images unrelated to the test sequences. In this way we obtain basis functions, also known as atoms, that can be sparsely combined to reconstruct local image content. In order to adapt the generic knowledge encoded in the dictionary to the specific object being tracked, we associate a set of votes and local object appearances to each atom: this is the only information being updated during online tracking. In each frame of the sequence the object's bounding box position is retrieved through a voting strategy. Our method exhibits robustness towards occlusions, sudden local and global illumination changes as well as shape changes. We test our method on 50 standard sequences obtaining results comparable or superior to the state of the art.

Keypoint detection represents the first stage in the majority of modern computer vision pipelines based on automatically established correspondences between local descriptors. However, no standard solution has emerged yet in the case of 3D data such as point clouds or meshes, which exhibit high variability in level of detail and noise. More importantly, existing proposals for 3D keypoint detection rely on geometric saliency functions that attempt to maximize repeatability rather than distinctiveness of the selected regions, which may lead to sub-optimal performance of the overall pipeline. To overcome these shortcomings, we cast 3D keypoint detection as a binary classification between points whose support can be correctly matched by a predefined 3D descriptor or not, thereby learning a descriptor-specific detector that adapts seamlessly to different scenarios. Through experiments on several public datasets, we show that this novel approach to the design of a keypoint detector represents a flexible solution that, nonetheless, can provide state-of-the-art descriptor matching performance.

This paper proposes a temporal tracking algorithm based on Random Forest that uses depth images to estimate and track the 3D pose of a rigid object in real-time. Compared to the state of the art aimed at the same goal, our algorithm holds important attributes such as high robustness against holes and occlusion, low computational cost of both learning and tracking stages, and low memory consumption. These are obtained (a) by a novel formulation of the learning strategy, based on a dense sampling of the camera viewpoints and learning independent trees from a single image for each camera view, as well as, (b) by an insightful occlusion handling strategy that enforces the forest to recognize the object's local and global structures. Due to these attributes, we report state-of-the-art tracking accuracy on benchmark datasets, and accomplish remarkable scalability with the number of targets, being able to simultaneously track the pose of over a hundred objects at 30~fps with an off-the-shelf CPU. In addition, the fast learning time enables us to extend our algorithm as a robust online tracker for model-free 3D objects under different viewpoints and appearance changes as demonstrated by the experiments.

This work proposes a real-time segmentation method for 3D point clouds obtained via Simultaneous Localization And Mapping (SLAM). The proposed method incrementally merges segments obtained from each input depth image in a unified global model using a SLAM framework. Differently from all other approaches, our method is able to yield segmentation of scenes reconstructed from multiple views in real-time, with a complexity that does not depend on the size of the global model. At the same time, it is also general, as it can be deployed with any frame-wise segmentation approach as well as any SLAM algorithm. We validate our proposal by a comparison with the state of the art in terms of computational efficiency and accuracy on a benchmark dataset, as well as by showing how our method can enable real-time segmentation from reconstructions of diverse real indoor environments.

Fusing intraoperative X-ray data with real-time video in a common reference frame is not trivial since both modalities have to be acquired from the same viewpoint. The goal of this work is to design a flexible system comprising two RGBD sensors that can be attached to any mobile C-arm, with the objective of synthesizing projective color images from the X-ray source viewpoint. To achieve this, we calibrate the RGBD sensors to the X-ray source with a 3D calibration object. Then, we synthesize the projective color image from the X-ray viewpoint by applying a volumetric-based rendering method. Finally, the X-ray image is overlaid on the projective image without any further registration, offering a multimodal visualization of X-ray and color images. In this paper we present the different steps of development (i.e. hardware setup, calibration and rendering algorithm) and discuss clinical applications for the new video augmented C-arm. By placing X-ray markers on a hand patient and a spine model, we show that the overlay accuracy between the X-ray image and the synthetized image is in average 1.7 mm.

In Spatial Augmented Reality (SAR) applications, real world objects are augmented with virtual content by means of a calibrated camera-projector system. A computer generated model (CAD) of the real object is used to plan the positions where the virtual content is to be projected. It is often the case that the real object deviates from its CAD model, this resulting in misregistered augmentations. We propose a new method to dynamically correct the planned augmentation by accommodating for the unknown deviations in the object geometry. We use a closed loop approach where the projected features are detected in the camera image and deployed as feedback. As a result, the registration misalignment is identified and the augmentations are corrected in the areas affected by the deviation. Our work is especially focused on SAR applications related to the industrial domain, where this problem is omnipresent. We show that our method is effective and beneficial for multiple industrial applications.

This work aims at automatic detection of man-made pole-like structures in scans of urban environments acquired by a 3D sensor mounted on top a moving vehicle. Pole-like structures, such as e.g. road signs and streetlights, are widespread in these environments, and their reliable detection is relevant to applications dealing with autonomous navigation, facility damage detection, city planning and maintenance. Yet, due to the characteristic thin shape, detection of man-made pole-like structures is significantly prone to both noise as well as occlusions and clutter, the latter being pervasive nuisances when scanning urban environments. Our approach is based on a “local” stage, whereby local features are classified and clustered together, followed by a “global” stage aimed at further classification of candidate entities. The proposed pipeline turns out effective in experiments on a standard publicly available dataset as well as on a challenging dataset acquired during the project for validation purposes.

Object detection in images withstanding significant clutter and occlusion is still a challenging task whenever the object surface is characterized by poor informative content. We propose to tackle this problem by a compact and distinctive representation of groups of neighboring line segments aggregated over limited spatial supports and invariant to rotation, translation and scale changes. Peculiarly, our proposal allows for leveraging on the inherent strengths of descriptor-based approaches, i.e. robustness to occlusion and clutter and scalability with respect to the size of the model library, also when dealing with scarcely textured objects.

This paper proposes an effective algorithm for recognizing objects and accurately estimating their 6DOF pose in scenes acquired by a RGB-D sensor. The proposed method is based on a combination of different recognition pipelines, each exploiting the data in a diverse manner and generating object hypotheses that are ultimately fused together in an Hypothesis Verification stage that globally enforces geometrical consistency between model hypotheses and the scene. Such a scheme boosts the overall recognition performance as it enhances the strength of the different recognition pipelines while diminishing the impact of their specific weaknesses. The proposed method outperforms the state-of-the-art on two challenging benchmark datasets for object recognition comprising 35 object models and, respectively, 176 and 353 scenes.

In this paper we present a pipeline for automatic detection of traffic signs in images. The proposed system can deal with high appearance variations, which typically occur in traffic sign recognition applications, especially with strong illumination changes and dramatic scale changes. Unlike most existing systems, our pipeline is based on interest regions extraction rather than a sliding window detection scheme. The proposed approach has been specialized and tested in three variants, each aimed at detecting one of the three categories of Mandatory, Prohibitory and Danger traffic signs. Our proposal has been evaluated experimentally within the German Traffic Sign Detection Benchmark competition.

The ability to perceive possible interactions with the environment is a key capability of task-guided robotic agents. An important subset of possible interactions depends solely on the objects of interest and their position and orientation in the scene. We call these object-based interactions 0-order affordances and divide them among non-hidden and hidden whether the current configuration of an object in the scene renders its affordance directly usable or not. Conversely to other works, we propose that detecting affordances that are not directly perceivable increase the usefulness of robotic agents with manipulation capabilities, so that by appropriate manipulation they can modify the object configuration until the seeked affordance becomes available. In this paper we show how 0-order affordances depending on the geometry of the objects and their pose can be learned using a supervised learning strategy on 3D mesh representations of the objects allowing the use of the whole object geometry. Moreover, we show how the learned affordances can be detected in real scenes obtained with a low-cost depth sensor like the Microsoft Kinect through object recognition and 6D0F pose estimation and present results for both learning on meshes and detection on real scenes to demonstrate the practical application of the presented approach.

Pattern matching is widely used in signal processing, computer vision, and image and video processing. Full search equivalent algorithms accelerate the pattern matching process and, in the meantime, yield exactly the same result as the full search. This paper proposes an analysis and comparison of state-of-the-art algorithms for full search equivalent pattern matching. Our intention is that the data sets and tests used in our evaluation will be a benchmark for testing future pattern matching algorithms, and that the analysis concerning state-of-the-art algorithms could inspire new fast algorithms. We also propose extensions of the evaluated algorithms and show that they outperform the original formulations.

Motivated by the increasing availability of 3D sensors capable of delivering both shape and texture information, this paper presents a novel descriptor for feature matching in 3D data enriched with texture. The proposed approach stems from the theory of a recently proposed descriptor for 3D data which relies on shape only, and represents its generalization to the case of multiple cues associated with a 3D mesh. The proposed descriptor, dubbed CSHOT, is demonstrated to notably improve the accuracy of feature matching in challenging object recognition scenarios characterized by the presence of clutter and occlusions.

We propose a method to perform automatic segmentation of 3D scenes based on a standard classifier, whose learning model is continuously improved by means of new samples, and a grouping stage, that enforces local consistency among classified labels. The new samples are automatically delivered to the system by a feedback loop based on a feature selection approach that exploits the outcome of the grouping stage. By experimental results on several datasets we demonstrate that the proposed online learning paradigm is effective in increasing the accuracy of the whole 3D segmentation thanks to the improvement of the learning model of the classifier by means of newly acquired, unsupervised data.

Template matching is a computationally intensive problem aimed at locating a template within a image. When dealing with images having more than one channel, the computational burden becomes even more dramatic. For this reason, in this paper we investigate on a methodology to speed-up template matching on multi-channel images without deteriorating the outcome of the search. In particular, we propose a fast, exhaustive technique based on the Zero-mean Normalized Cross-Correlation (ZNCC) inspired from previous work related to grayscale images. Experimental testing performed over thousands of template matching instances demonstrates the efficiency of our proposal.

We present a background subtraction approach aimed at efficiency and accuracy also in presence of common sources of disturbance such as illumination changes, camera gain and exposure variations, noise. The novelty of the proposal relies on a-priori modeling the local effect of disturbs on small neighborhoods of pixel intensities as a monotonic, homogeneous, second-degree polynomial transformation plus additive Gaussian noise. This allows for classifying pixels as changed or unchanged by an efficient inequality-constrained least-squares fitting procedure. Experiments prove that the approach is state-of-the-art in terms of efficiency-accuracy tradeoff on challenging sequences characterized by disturbs yielding sudden and strong variations of the background appearance.

Despite the significant number of stereo vision algorithms proposed in literature in the last decade, most proposals are notably computationally demanding and/or memory hungry so that it is unfeasible to employ them in application scenarios requiring real-time or near real-time processing on platforms with limited resources such as embedded devices. In this paper, we have selected the subset of proposals that appears more suited to the above requirements and, since literature lacks a proper comparison between these methods, we propose a quantitative experimental evaluation aimed at highlighting the best performing approach under the two criteria of accuracy and efficiency. The evaluation is performed on a standard benchmark dataset as well as on a novel dataset, acquired by means of an active technique, characterized by realistic working conditions.

Spacetime stereo is a promising technique for accurate 3D reconstruction based on randomly varying illumination and temporal integration of the stereo matching cost. In this paper we show that the standard spacetime stereo approach can be improved in terms of accuracy of disparity estimation and convergence speed by adoption of suitable matching algorithms based on adaptive support windows. We also present a practical and cost-effective 3D reconstruction system that deploys the proposed improved spacetime method together with cheap commercial off-the-shelf hardware (a PC, a stereo camera and a projector). Experimental results show that the proposed system can yield rapidly accurate 3D reconstruction of various types of objects and faces.

We present a background subtraction algorithm aimed at efficiency and robustness to common sources of disturbance such as illumination changes, camera gain and exposure variations, noise. The approach relies on modeling the local effect of disturbance factors on a neighborhood of pixel intensities as a second-degree polynomial transformation plus additive Gaussian noise. This allows for classifying pixels as changed or unchanged by a simple least-squares polynomial fitting procedure. Experimental results prove that the approach is state-of-the-art in challenging sequences characterized by sources of disturbance yielding sudden and strong background appearance changes.

Low-cost and low-power video surveillance systems based on networks of wireless video sensors will enter soon the marketplace with the promise of flexibility, quick deployment and providing accurate and real-time visual data. Energy autonomy and efficiency of the implemented algorithms are undoubtedly the primary design challenges to be addressed on systems subject to low computational capabilities and memory constraints. In this paper we present a low-power video sensor node designed for low-cost video surveillance which is able to detect abandoned and removed objects. The system exploits multi-modal sensor integration which saves on-board power consumption. In particular a pyroelectric infrared (PIR) sensor is exploited to optimize the use of the camera, grabbing images only when required in order to obtain the maximum efficiency from event recognition. Our fixed-point ARM-based approach is characterized in terms of runtime execution and power consumption, while efficiency is demonstrated by experimental results and compared with floating point implementations.

This paper proposes a novel method for fast pattern matching based on dissimilarity functions derived from the Lp norm, such as the Sum of Squared Differences (SSD) and the Sum of Absolute Differences (SAD). The proposed method is full-search equivalent, i.e. it yields the same results as the Full Search (FS) algorithm. In order to pursue computational savings the method deploys a succession of increasingly tighter lower bounds of the adopted Lp norm-based dissimilarity function. Such bounding functions allow for establishing a hierarchy of pruning conditions aimed at skipping rapidly those candidates that cannot satisfy the matching criterion. The paper includes an experimental comparison between the proposed method and other full-search equivalent approaches known in literature, which proves the remarkable computational efficiency of our proposal.

We present a method aimed at monitoring access to interlocks and secured entrance areas, which deploys two views in order to robustly perform intrusion detection and singularization. The main contributions are represented by an original approach to perform background subtraction, which is particularly robust against sudden illumination changes, shadows and photometric distortions, and by the use of a feature extraction and classification approach which allows to reliably determine an estimation of the number of people currently occupying the monitored area. Our system is designed to operate in very small interlocks and can work in a substantially unstructured environment.

In the last decades several cost aggregation methods aimed at improving the robustness of stereo correspondence within local and global algorithms have been proposed. Given the recent developments and the lack of an appropriate comparison, in this paper we survey, classify and compare experimentally on a standard data set the main cost aggregation approaches proposed in literature. The experimental evaluation addresses both accuracy and computational requirements, so as to outline the best performing methods under these two criteria.

This paper presents a method that reduces the computational cost of template matching based on the zero-mean normalized cross-correlation (ZNCC) without compromising the accuracy of the results. A very effective condition is determined at a small and fixed cost that allow to rapidly detect a large number of mismatching candidates with no need to compute the ZNCC score. Then, thanks to the use of an additional set of conditions, the computation of the whole ZNCC function is typically required only for a very small number of candidates. Experimental results demonstrate the effectiveness of our approach.

Recent research activity on stereo matching has proved the efficacy of local approaches based on advanced cost aggregation strategies in accurately retrieving 3D information. However, accuracy is typically achieved at expense of computational efficiency, with best methods being far from meeting real-time requirements. On the other side, basic real-time local algorithms relying on a rectangular correlation window suffer from significant ambiguity along depth borders and untextured areas. This work proposes a novel local approach aimed at maximizing the speed-accuracy trade-off by means of an efficient segmentation-based cost aggregation strategy.

We propose a novel algorithm, referred to as enhanced bounded correlation (EBC), that significantly reduces the number of computations required to carry out template matching based on normalized cross correlation (NCC) and yields exactly the same result as the full search algorithm. The algorithm relies on the concept of bounding the matching function: finding an efficiently computable upper bound of the NCC rapidly prunes those candidates that cannot provide a better NCC score with respect to the current best match. In this framework, we apply a succession of increasingly tighter upper bounding functions based on Cauchy-Schwarz inequality. Moreover, by including an online parameter prediction step into EBC, we obtain a parameter free algorithm that, in most cases, affords computational advantages very similar to those attainable by optimal offline parameter tuning. Experimental results show that the proposed algorithm can significantly accelerate a full-search equivalent template matching process and outperforms state-of-the-art methods.

This paper proposes a novel algorithm aimed at speeding-up template matching based on the L_p norm. The algorithm is exhaustive, i.e. it yields the same results as a Full Search (FS) template matching process, and is based on the deployment of tight lower bounds that can be derived by using together the triangular inequality and partial evaluations of the L_p norm. In order to deploy this, template and image subwindows are properly partitioned. The experimental results prove that the proposed algorithm allows speeding-up the FS process and also (when applied to the L_2 norm) the exhaustive approach based on the Fast Fourier Transform.