Is this you? Claim this profile to correct it, add a bio and choose the work people see first.
Claim this profileAcademic lineage
Possible advisorsa guess from early papers, not confirmed
- Luigi Di StefanoSuggested from co-authorship
Is this you? Claim this profile to confirm or dismiss it.
Works414 from public data
- Deeper Depth Prediction with Fully Convolutional Residual Networks2,056
A fully convolutional architecture, encompassing residual learning, to model the ambiguous mapping between monocular images and depth maps is proposed and a novel way to efficiently learn feature map up-sampling within the network is presented.
- Unique Signatures of Histograms for Local Surface Description1,653
A novel comprehensive proposal for surface representation is formulated, which encompasses a new unique and repeatable local reference frame as well as a new 3D descriptor.
- SSD-6D: Making RGB-Based 3D Detection and 6D Pose Estimation Great Again1,042
A novel method for detecting 3D model instances and estimating their 6D poses from RGB data in a single shot that competes or surpasses current state-of-the-art methods that leverage RGBD data on multiple challenging datasets.
- SHOT: Unique signatures of histograms for surface and texture description1,012
A thorough experimental evaluation vouches that SHOT outperforms state-of-the-art local descriptors in experiments addressing descriptor matching for object recognition, 3D reconstruction and shape retrieval.
- Neural Fields in Visual Computing and Beyond866
A review of the literature on neural fields 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.
- CNN-SLAM: Real-Time Dense Monocular SLAM with Learned Depth Prediction788
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 monocularSLAM approaches tend to fail, e.g. along low-textured regions, and vice-versa.
- BOP: Benchmark for 6D Object Pose Estimation571
A benchmark for 6D pose estimation of a rigid object from a single RGB-D input image shows that methods based on point-pair features currently perform best, outperforming template matching methods, learning-based methods and methodsbased on 3D local features.
- Performance Evaluation of 3D Keypoint Detectors472
A categorization of existing methods in two classes, that allows for highlighting their common traits, is proposed, so as to abstract all algorithms to two general structures in terms of repeatability, distinctiveness and computational efficiency.
- 462
- 387
- Unique shape context for 3d data description383
This paper shows how to deploy a unique local Reference Frame to improve the accuracy and reduce the memory footprint of the well-known 3D Shape Context descriptor.
- 3D Point Capsule Networks371
3D capsule networks are proposed, an auto-encoder designed to process sparse 3D point clouds while preserving spatial arrangements of the input data and enables new applications such as part interpolation and replacement.
- Learning 3D Semantic Scene Graphs From 3D Indoor Reconstructions364
This work proposes a learned method that regresses a scene graph from the point cloud of a scene, based on PointNet and Graph Convolutional Networks, and introduces 3DSSG, a semiautomatically generated dataset, that contains semantically rich scene graphs of 3D scenes.
- 329
- 296
- Deep Learning of Local RGB-D Patches for 3D Object Detection and 6D Pose Estimation293
A 3D object detection method that uses regressed descriptors of locally-sampled RGB-D patches for 6D vote casting that generalizes well to previously unseen input data, and delivers robust detection results that compete with and surpass the state-of-the-art while being scalable in the number of objects is presented.
- SceneGraphFusion: Incremental 3D Scene Graph Prediction from RGB-D Sequences290
This work proposes a method to incrementally build up semantic scene graphs from a 3D environment given a sequence of RGB-D frames by means of a graph neural network and proposes a novel attention mechanism well suited for partial and missing graph data present in such an incremental reconstruction scenario.
- RIO: 3D Object Instance Re-Localization in Changing Indoor Environments283
A new data-driven approach that efficiently finds matching features using a fully-convolutional 3D correspondence network operating on multiple spatial scales is introduced, which outperforms state-of-the-art baselines on the authors' newly-established benchmark, achieving an accuracy of 30.58%.
- SPARF: Neural Radiance Fields from Sparse and Noisy Poses274
This work introduces Sparse Pose Adjusting Radiance Field (SPARF), to address the challenge of novel-view synthesis given only few wide-baseline input images with noisy camera poses, and sets a new state of the art in the sparse-view regime on multiple challenging datasets.
- SHIFT: A Synthetic Driving Dataset for Continuous Multi-Task Domain Adaptation273
This paper introduces the largest multi-task synthetic dataset for autonomous driving, SHIFT, which presents discrete and continuous shifts in cloudiness, rain and fog intensity, time of day, and vehicle and pedestrian density.
- Implicit Neural Representations for Image Compression249
This work proposes the first comprehensive compression pipeline based on INRs including quantization, quantization-aware retraining and entropy coding and finds that the approach to source compression with INRs vastly outperforms similar prior work, is competitive with common compression algorithms designed specifically for images and closes the gap to state-of-the-art learned approaches based on Rate-Distortion Autoencoders.
- ZNCC-based template matching using bounded partial correlation244
The algorithms shown in this paper generalize and extend the concept of Bounded Partial Correlation, previously devised for a template matching process based on the Normalized Cross-Correlation function (NCC), previously devised for a template matching process based on the Normalized Cross-Correlation function (NCC).
- 243
- Learning in an Uncertain World: Representing Ambiguity Through Multiple Hypotheses232
This work proposes a frame-work for reformulating existing single-prediction models as multiple hypothesis prediction (MHP) models and an associated meta loss and optimization procedure to train them, and finds that MHP models outperform their single-hypothesis counterparts in all cases and expose valuable insights into the variability of predictions.
- Segmentation-Based Adaptive Support for Accurate Stereo Correspondence221
A novel support aggregation strategy which includes information obtained from a segmentation process is proposed which is effective in improving the state of the art in dense stereo correspondence.
- Distortion-Aware Convolutional Filters for Dense Prediction in Panoramic Images220
This work proposes a learning approach for panoramic depth map estimation from a single image, thanks to a specifically developed distortion-aware deformable convolution filter, which can be trained by means of conventional perspective images, then used to regress depth forPanoramic images, thus bypassing the effort needed to create annotated pan oramic training dataset.
- ZebraPose: Coarse to Fine Surface Encoding for 6DoF Object Pose Estimation210
This work presents a discrete descriptor, which can represent the object surface densely by incorporating a hierarchical binary grouping, and proposes a coarse to fine training strategy, which enables fine-grained correspondence prediction of the 6DoF pose.
- 210
- Learning Graph Embeddings for Compositional Zero-shot Learning205
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 and significantly outperforms the state of the art on MIT-States and UT-Zappos in the challenging generalized compositional zero-shot setting.
- Fully-Convolutional Point Networks for Large-Scale Point Clouds203
This work proposes a general-purpose, fully-convolutional network architecture for efficiently processing large-scale 3D data and demonstrates its ability to effectively learn both low-level features as well as complex compositional relationships by evaluating it on benchmark datasets for semantic voxel segmentation, semantic part segmentation and 3D scene captioning.
- GPV-Pose: Category-level Object Pose Estimation via Geometry-guided Point-wise Voting192
GPV-Pose is proposed, a novel framework for robust category-level pose estimation, harnessing geometric insights to enhance the learning of category- level pose-sensitive features, which produces superior results to state-of-the-art competitors on common public benchmarks, whilst almost achieving real-time inference speed at 20 FPS.
- Long Short-Term Memory Kalman Filters: Recurrent Neural Estimators for Pose Regularization191
This work proposes to learn rich, dynamic representations of the motion and noise models from data using long shortterm memory, which allows representations that depend on all previous observations and all previous states.
- A Global Hypotheses Verification Method for 3D Object Recognition187
Peculiar to this approach is the inherent ability to detect significantly occluded objects without increasing the amount of false positives, so that the operating point of the object recognition algorithm can nicely move toward a higher recall without sacrificing precision.
- Batch normalization embeddings for deep domain generalization179
This work explicitly train domain-dependant representations by using ad-hoc batch normalization layers to collect independent domain's statistics and proposes to use these statistics to map domains in a shared latent space, where membership to a domain can be measured by means of a distance function.
- SparseFusion: Fusing Multi-Modal Sparse Representations for Multi-Sensor 3D Object Detection174
SparseFusion is proposed, a novel multi-sensor 3D detection method that exclusively uses sparse candidates and sparse representations that achieves state-of-the-art performance on the nuScenes benchmark while also running at the fastest speed, even outperforming methods with stronger backbones.
- SO-Pose: Exploiting Self-Occlusion for Direct 6D Pose Estimation174
This work addresses the challenge of directly regressing all 6 degrees-of-freedom for the object pose in a cluttered environment from a single RGB image 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.
- TextMesh: Generation of Realistic 3D Meshes From Text Prompts172
A novel method for generation of highly realistic-looking 3D meshes is proposed, extending NeRF to employ an SDF backbone, leading to improved 3D mesh extraction and a novel way to finetune the mesh texture, removing the effect of high saturation and improving the details of the output3D mesh.
- Deep Model-Based 6D Pose Refinement in RGB172
A new visual loss is proposed that drives the pose update by aligning object contours, thus avoiding the definition of any explicit appearance model and producing pose accuracies that come close to 3D ICP without the need for depth data.
- OUR-CVFH – Oriented, Unique and Repeatable Clustered Viewpoint Feature Histogram for Object Recognition and 6DOF Pose Estimation172
A novel method to estimate a unique and repeatable reference frame in the context of 3D object recognition from a single viewpoint based on global descriptors with substantial improvement is presented regarding accuracy in recognition and 6DOF pose estimation, as well as in terms of computational performance.
- 165
- Variational Transformer Networks for Layout Generation163
This work exploits the properties of self-attention layers to capture high level relationships between elements in a layout, and uses these as the building blocks of the well-known Variational Autoencoder (VAE) formulation.
- Self6D: Self-supervised Monocular 6D Object Pose Estimation161
This work proposes the idea of monocular 6D pose estimation by means of self-supervised learning, removing the need for real annotations, and demonstrates that the proposed self- supervision model is able to significantly enhance the model's original performance, outperforming all other methods relying on synthetic data or employing elaborate techniques from the domain adaptation realm.
- Concurrent Segmentation and Localization for Tracking of Surgical Instruments161
A novel method is proposed that takes advantage of the interdependency between localization and segmentation of the surgical tool and reformulate the 2D pose estimation as a heatmap regression and thereby enable a robust, concurrent regression of both tasks via deep learning.
- Sampling-Free Epistemic Uncertainty Estimation Using Approximated Variance Propagation160
This work presents a sampling-free approach for computing the epistemic uncertainty of a neural network and applies this approach to large-scale visual tasks to demonstrate the advantages of the method compared to sampling-based approaches in terms of quality of the uncertainty estimates as well as of computational overhead.
- RadSplat: Radiance Field-Informed Gaussian Splatting for Robust Real- Time Rendering with 900+ FPS152
RadSplat is presented, a lightweight method for robust real-time rendering of complex scenes that enables state-of-the-art synthesis of complex captures at 900+ FPS and proposes a novel test-time filtering approach that further accelerates rendering and allows to scale to larger, house-sized scenes.
- Query-Guided End-To-End Person Search151
A novel query-guided end-to-end person search network (QEEPS) to address both person detection and re-identification and outperform the previous state-of-the-art datasets by a large margin.
- Explaining the Ambiguity of Object Detection and 6D Pose From Visual Data141
This work proposes a new approach to 3D object detection and pose estimation which provides not only a better explanation for pose ambiguity, but also a higher accuracy in terms of pose estimation.
- Semantic Image Manipulation Using Scene Graphs140
This work introduces a spatio-semantic scene graph network that does not require direct supervision for constellation changes or image edits, and makes it possible to train the system from existing real-world datasets with no additional annotation effort.
- Structure-SLAM: Low-Drift Monocular SLAM in Indoor Environments135
The approach decouples rotation and translation estimation of the tracking process to reduce the long-term drift in indoor environments and outperforms the state of the art on common SLAM benchmarks such as ICL-NUIM and TUM RGB-D.
- 125
- 124
- Traffic sign detection via interest region extraction123
This work shows how a combination of solid image analysis and pattern recognition techniques can be used to tackle the problem of traffic sign detection in mobile mapping data, and presents in detail the design of a Traffic Sign Detection pipeline.
- Splat-SLAM: Globally Optimized RGB-Only SLAM with 3D Gaussians120
This work introduces the first RGB-only SLAM system with globally optimized tracking, dynamically adapting the Gaussian map to keyframe pose and depth updates, and incorporates so called Disparity, Scale and Pose Optimization (DSPO) for bundle adjustment, jointly optimizing pose, depth, and monocular depth scale.
- 120
- Dynamic Gaussian Marbles for Novel View Synthesis of Casual Monocular Videos117
Dynamic Gaussian Marbles is proposed, which 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.
- LayoutVLM: Differentiable Optimization of 3D Layout via Vision-Language Models115
This work introduces 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.
- Graph-to-3D: End-to-End Generation and Manipulation of 3D Scenes Using Scene Graphs112
This work proposes the first work that directly generates shapes from a scene graph in an end-to-end manner, and shows that the same model supports scene modification, using the respective scene graph as interface.
- RGB-D SLAM with Structural Regularities110
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.
- 106
- I2MVFormer: Large Language Model Generated Multi-View Document Supervision for Zero-Shot Image Classification103
This work provides a novel perspective on using an LLM to provide text supervision for a zero- shot image classification model and establishes a new state-of-the-art on three public benchmark datasets for zero-shot image classification with unsupervised semantic embeddings.
- 103
- 102
- 2D Image-To-3D Model: Knowledge-Based 3D Building Reconstruction (3DBR) Using Single Aerial Images and Convolutional Neural Networks (CNNs)101
A deep learning (DL)-based approach is proposed for the detection and reconstruction of buildings from a single aerial image and does not need any additional or auxiliary data and employs a single image to reconstruct the 3D models of buildings with the competitive precision.
- 100
- Quaternion Equivariant Capsule Networks for 3D Point Clouds96
A 3D capsule module for processing point clouds that is equivariant to 3D rotations and translations, as well as invariant to permutations of the input points, and builds a capsule network that disentangles geometry from pose, paving the way for more informative descriptors and a structured latent space.
- 94
- 90
- NeRFMeshing: Distilling Neural Radiance Fields into Geometrically-Accurate 3D Meshes89
This work proposes a novel compact and flexible architecture that enables easy 3D surface reconstruction from any NeRF-driven approach, and distill the volumetric 3D representation into a Signed Surface Approximation Network, allowing easy extraction of the 3D mesh and appearance.
- SecondPose: SE(3)-Consistent Dual-Stream Feature Fusion for Category-Level Pose Estimation87
Sec-ondPose is presented, a novel approach integrating object-specific ge-ometric features with semantic category priors from DI-NOv2 that achieves a 12.4% leap forward over the state-of-the-art on NOCS-REAL275 and still surpasses other competitors by a large margin.
- SoftPoolNet: Shape Descriptor for Point Cloud Completion and Classification87
This paper proves that the regional activation can be incorporated in many point cloud architectures like AtlasNet and PCN, leading to better performance for geometric completion.
- OpenNeRF: Open Set 3D Neural Scene Segmentation with Pixel-Wise Features and Rendered Novel Views86
This work proposes OpenNeRF which naturally operates on posed images and directly encodes the VLM features within the NeRF, and 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.
- ManhattanSLAM: Robust Planar Tracking and Mapping Leveraging Mixture of Manhattan Frames86
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, achieving superior performance compared to other state-of-the-art methods.
- 81
- Panoster: End-to-End Panoptic Segmentation of LiDAR Point Clouds79
This letter presents Panoster, a novel proposal-free panoptic segmentation method for LiDAR point clouds that reached state-of-the-art results among published approaches on the challenging SemanticKITTI benchmark, and further increased its lead by exploiting heuristic techniques.
- 79
- Introducing Language Guidance in Prompt-based Continual Learning78
Language Guidance for Prompt-based Continual Learning is proposed 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 and achieves performance improvements without needing any additional learnable parameters.
- CommonScenes: Generating Commonsense 3D Indoor Scenes with Scene Graph Diffusion76
CommonScenes is a fully generative model that converts scene graphs into corresponding controllable 3D scenes, which are semantically realistic and conform to commonsense and shows clear advantages over other methods regarding generation consistency, quality, and diversity.
- CONFORM: Contrast is All You Need For High-Fidelity Text-to-Image Diffusion Models73
This work introduces a novel perspective by tackling this challenge in a contrastive context of text-to-image diffusion models by intuitively promoting the segregation of objects in attention maps while also maintaining that pairs of related attributes are kept close to each other.
- Convolutional neural networks for real-time epileptic seizure detection73
It is conjecture that features learned via a convolutional neural network provide the ability to distinctively detect seizures from video, and even allow the system to generalize to different seizure types.
- 3D-VField: Adversarial Augmentation of Point Clouds for Domain Generalization in 3D Object Detection71
This work substantially improves the generalization of 3D object detectors to out-of-domain data by deforming point clouds during training with 3D-VField: a novel data augmentation method that plausibly deforms objects via vector fields learned in an adversarial fashion.
- Patient MoCap: Human Pose Estimation Under Blanket Occlusion for Hospital Monitoring Applications70
This work proposes a learning-based method that allows to automatically infer 3D patient pose from depth images using a combination of convolutional neural network and recurrent neural network, which is trained on a large database that covers a range of motions in the hospital bed.
- ForkNet: Multi-Branch Volumetric Semantic Completion From a Single Depth Image68
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 is proposed.
- Peeking behind objects: Layered depth prediction from a single image65
A novel approach based on Convolutional Neural Networks to jointly predict depth maps and foreground separation masks used to condition Generative Adversarial Networks for hallucinating plausible color and depths in the initially occluded areas is proposed.
- Learning to Prompt with Text Only Supervision for Vision-Language Models64
This work develops a language-only efficient training approach that enables prompts to distill rich contextual knowledge from LLM data and enables zero-shot transfer of prompts to new classes and datasets, potentially reducing the LLM prompt engineering cost.
- Incremental 3D Semantic Scene Graph Prediction from RGB Sequences64
A real-time framework that incrementally builds a consistent 3D semantic scene graph of a scene given an RGB image sequence using a novel incremental entity estimation pipeline and a scene graph prediction network.
- Learning Local Displacements for Point Cloud Completion63
A novel approach aimed at object and semantic scene completion from a partial scan represented as a 3D point cloud that relies on three novel layers that are used successively within an encoder-decoder structure and specifically developed for the task at hand.
- 60
- Multi‐modal imaging, model‐based tracking, and mixed reality visualisation for orthopaedic surgery60
A mixed reality support system, which incorporates multi-modal data fusion and model-based surgical tool tracking for creating a mixed reality environment supporting screw placement in orthopaedic surgery and provides an intuitive understanding of the surgical site.
- R4Dyn: Exploring Radar for Self-Supervised Monocular Depth Estimation of Dynamic Scenes59
R4Dyn is presented, a novel set of techniques to use cost-efficient radar data on top of a self-supervised depth estimation framework that substantially improves 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.
- Beyond Controlled Environments: 3D Camera Re-localization in Changing Indoor Scenes56
This paper adapts 3RScan - a recently introduced indoor RGB-D dataset designed for object instance re-localization - to create RIO10, a new long-term camera re- localization benchmark focused on indoor scenes and explores how state-of-the-art cameraRe-localizers perform according to these metrics.
- 56
- Stereo Vision Enabling Precise Border Localization Within a Scanline Optimization Framework56
A novel algorithm for obtaining accurate dense disparity measurements and precise border localization from stereo pairs is proposed, which embodies a very effective variable support approach based on segmentation within a Scanline Optimization framework.
- SSP-Pose: Symmetry-Aware Shape Prior Deformation for Direct Category-Level Object Pose Estimation55
An end-to-end trainable network SSP-Pose is proposed for category-level pose estimation, which integrates shape priors into a direct pose regression network and utilizes the symmetry information of each category to guide the shape prior deformation, and proposes a novel symmetry-aware loss to mitigate the matching ambiguity.
- Human Motion Analysis with Deep Metric Learning55
A novel metric learning objective based on a triplet architecture and Maximum Mean Discrepancy is proposed and a novel deep architecture based on attentive recurrent neural networks is proposed, which enforces a better separation within the learned embedding space of the different motion categories by means of the associated distribution moments.
- 55
- 53
- Omnia de EgoTempo: Benchmarking Temporal Understanding of Multi-Modal LLMs in Egocentric Videos51
This work introduces EgoTempo, a dataset specifically designed to evaluate temporal understanding in the egocentric domain, and shows that state-of-the-art Multi-Modal Large Language Models on these benchmarks achieve remarkably high performance using just text or a single frame as input.
- Attention-based Lane Change Prediction51
An attention-based recurrent model to tackle both understandability and prediction quality is proposed and metrics which reflect the discomfort felt by the driver are proposed.
- 51
- 50
- 50
- Real-time localization of articulated surgical instruments in retinal microsurgery50
A method that builds on random forests and addresses the task by modelling the instrument as an articulated object by relating the movement of the instrument to the induced changes on the image intensities and can outperform current state-of-the art in real-time tracking.
- Surgical Tool Tracking and Pose Estimation in Retinal Microsurgery50
This work proposes an algorithm for simultaneous tool tracking and pose estimation that is inspired by state-of-the-art computer vision techniques and introduces a method based on regression forests to track the tool tip and to recover the tool's articulated pose.
- Hashmod: A Hashing Method for Scalable 3D Object Detection50
It is shown empirically that the complexity of the method is sublinear with the number of objects and it enables detection and pose estimation of many 3D objects with high accuracy while outperforming the state-of-the-art in terms of runtime.
- MonoGraspNet: 6-DoF Grasping with a Single RGB Image49
This paper proposes 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.
- 49
- 49
- 49
- Deep Positional and Relational Feature Learning for Rotation-Invariant Point Cloud Analysis48
A novel deep network for point clouds analysis is proposed by incorporating positional information of points as inputs while yielding rotation-invariance, and experiments show state-of-the-art classification and segmentation performances on benchmark datasets.
- SG-Bot: Object Rearrangement via Coarse-to-Fine Robotic Imagination on Scene Graphs47
This paper presents SG-Bot, a novel rearrangement framework that utilizes a coarse-to-fine scheme with a scene graph as the scene representation, and demonstrates that SG-Bot outperforms competitors by a large margin.
- DNS-SLAM: Dense Neural Semantic-Informed SLAM47
DNS SLAM is proposed, the first semantic neural SLAM method that trains class-wise scene representations while providing stable camera tracking at the same time, and outputs class-wise decomposed reconstructions with better texture, capturing appearance and geometric details.
- Hough Voting for 3D Object Recognition under Occlusion and Clutter47
This work proposes a novel approach for the detection of free-form shapes in a3D space that matches 3D features through their descriptions to attain correspondences, then accumulates evidence of the presence of the object(s) being sought by verifying the consensus of correspondences within a 3D Hough space.
- Learning to Detect Good 3D Keypoints46
Through extensive experiments on standard datasets, this work shows how feature matching performance improves significantly by deploying 3D descriptors together with companion detectors learned by the methodology with respect to the adoption of established state-of-the-art 3D detectors based on hand-crafted saliency functions.
- Real-Time 3D Model Tracking in Color and Depth on a Single CPU Core45
A novel method to track 3D models in color and depth data is presented and it is shown 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.
- OPA-3D: Occlusion-Aware Pixel-Wise Aggregation for Monocular 3D Object Detection44
This work proposes to jointly estimate dense scene depth with depth-bounding box residuals and object bounding boxes, allowing a two-stream detection of 3D objects that harnesses both geometry and context information.
- 44
- CubeDiff: Repurposing Diffusion-Based Image Models for Panorama Generation43
A novel method for generating 360{\deg} panoramas from text prompts or images by employing multi-view diffusion models to jointly synthesize the six faces of a cubemap, simplifying the generation process and enabling the use of existing multi-view diffusion models.
- Opportunistic Interfaces for Augmented Reality: Transforming Everyday Objects into Tangible 6DoF Interfaces Using Ad hoc UI43
This paper presents the workflow and technical details of Ad hoc UI (AhUI), a prototyping toolkit to empower users to turn everyday objects into opportunistic interfaces on the fly to grant users complete freedom to summon virtual interfaces on everyday objects via voice commands or tapping gestures.
- A performance evaluation of point pair features43
A detailed analysis of PPFs is presented, which highlights under which conditions PPFs perform particularly well as well as its main weaknesses, and finds that PPFs compared to most local histogram features degrade faster under disturbances such as occlusion and clutter.
- 43
- 41
- Object-Driven Multi-Layer Scene Decomposition From a Single Image40
The 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, which enables various applications, such as 3D photography and diminished reality, all from asingle RGB image.
- Guide Me: Interacting with Deep Networks40
Methods to flexibly guide a trained convolutional neural network through user input to improve its performance during inference are explored by inserting a layer that acts as a spatio-semantic guide into the network.
- Search3D: Hierarchical Open-Vocabulary 3D Segmentation39
This work introduces Search3D, an approach to construct hierarchical open-vocabulary 3D scene representations, enabling 3D search at multiple levels of granularity: fine-grained object parts, entire objects, or regions described by attributes like materials.
- Shape, Pose, and Appearance from a Single Image via Bootstrapped Radiance Field Inversion39
This work introduces a principled end-to-end reconstruction framework for natural images, where accurate ground-truth poses are not available, and leverages an unconditional 3D-aware generator, to which a hybrid inversion scheme where a model produces a first guess of the solution which is then refined via optimization.
- Learning 3D Semantic Scene Graphs with Instance Embeddings39
This work proposes instead a novel neural network architecture for 3D data, where the aim is to learn to regress semantic graphs from a given 3D scene, and shows the effectiveness of graphs as an intermediate representation on a retrieval task.
- Interest Points via Maximal Self-Dissimilarities39
This work proposes a novel interest point detector stemming from the intuition that image patches which are highly dissimilar over a relatively large extent of their surroundings hold the property of being repeatable and distinctive, thereby achieving enhanced repeatability, distinctiveness and localization accuracy.
- 37
- Fast and Accurate Semantic Mapping through Geometric-based Incremental Segmentation37
An efficient and scalable method for incrementally building a dense, semantically annotated 3D map in real-time that 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.
- How Good is my Video-LMM? Complex Video Reasoning and Robustness Evaluation Suite for Video-LMMs36
A training-free Dual-Step Contextual Prompting technique is developed to effectively enhance the performance of existing Video-LMMs on the CVRR-ES benchmark, providing valuable in-sights for building the next generation of human-centric AI systems with advanced robustness and reasoning capabilities.
- 36
- Mixed Diffusion for 3D Indoor Scene Synthesis35
MiDiffusion is presented, a novel mixed discrete-continuous diffusion model designed to synthesize plausible 3D indoor scenes given a floor plan and pre-arranged objects that outperforms state-of-the-art autoregressive and diffusion models in floor-conditioned 3D scene synthesis.
- SuperGSeg: Open-Vocabulary 3D Segmentation with Structured Super-Gaussians35
A novel approach that fosters cohesive, context-aware hierarchical scene representation by disentangling segmentation and language field distillation and achieves remarkable performance on both open-vocabulary object selection and semantic segmentation tasks is introduced.
- 35
- Query-guided networks for few-shot fine-grained classification and person search35
A novel unified Query-Guided Network (QGN) applicable to few-shot fine-grained classification and person search is proposed, which improves on a few recent few- shot fine- grained datasets and performs competitively on the person search datasets.
- Efficient template matching for multi-channel images35
This paper proposes a fast, exhaustive technique based on the Zero-mean Normalized Cross-Correlation (ZNCC) inspired from previous work related to grayscale images to speed-up template matching on multi-channel images without deteriorating the outcome of the search.
- LIME: Localized Image Editing via Attention Regularization in Diffusion Models34
LIME does not require user-specified regions of interest (RoI) or additional text input, but rather employs features from pretrained methods and a straightforward clustering method to obtain precise editing mask, which consistently improves the performance of existing methods in various editing benchmarks.
- Feature matching evaluation for multimodal correspondence34
A study and evaluation of approaches aimed at image matching under different modalities, together with a survey of methodologies used for performance comparison in this specific context, and a novel algorithm for image matching.
- Domain-Specific Priors and Meta Learning for Few-Shot First-Person Action Recognition33
This work aims to develop an effective method for few-shot transfer learning for first-person action classification using 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.
- Unconditional Scene Graph Generation33
This work develops a deep auto-regressive model called SceneGraphGen which can directly learn the probability distribution over labelled and directed graphs using a hierarchical recurrent architecture and demonstrates the application of the generated graphs in image synthesis, anomaly detection and scene graph completion.
- Co-Planar Parametrization for Stereo-SLAM and Visual-Inertial Odometry33
A novel SLAM framework for stereo and visual inertial odometry estimation and an efficient and robust parametrization of co-planar points and lines which leverages specific geometric constraints to improve camera pose optimization in terms of both efficiency and accuracy.
- 33
- 32
- AnyUp: Universal Feature Upsampling31
This work proposes an inference-time feature-agnostic upsampling architecture to alleviate this limitation and improve upsampling quality, and sets a new state of the art for upsampled features, generalizes to different feature types, and preserves feature semantics while being efficient and easy to apply to a wide range of downstream tasks.
- 31
- Quantifying Aleatoric and Epistemic Uncertainty Using Density Estimation in Latent Space.31
This work investigates whether the distribution of latent representations indeed contains information about the uncertainty associated with the predictions of a neural network and concludes with the exciting finding that the hidden repesentations of a deterministic neural network indeed contain information about its uncertainty.
- 31
- 31
- On the Use of Implicit Shape Models for Recognition of Object Categories in 3D Data31
Modifications to three design points within the structure of a standard ISM are proposed to enhance its effectiveness for the categorization of databases entries, either 3D or 2D: namely, codebook size and composition, codeword activation strategy and vote weight strategy.
- 30
- 30
- 29
- 29
- Holistic OR domain modeling: a semantic scene graph approach29
A neural network-based SSG generation pipeline for semantic reasoning in the OR domain is introduced and the capabilities of the scene graphs are further highlighted by their successful application to clinical role prediction and surgical phase recognition tasks.
- SoftPool++: An Encoder–Decoder Network for Point Cloud Completion29
This work proposes a novel convolutional operator that extracts permutation-invariant features from the point cloud via a soft-pooling of feature activations, which are able to preserve fine-grained geometric details in the input shape structure.
- TSDF++: A Multi-Object Formulation for Dynamic Object Tracking and Reconstruction29
This work introduces a novel multi-object TSDF formulation that can encode multiple object surfaces at any given location in the map that allows maintaining accurate reconstruction of surfaces even while they become temporarily occluded by other objects moving in their proximity.
- CertainNet: Sampling-Free Uncertainty Estimation for Object Detection29
This work proposes a novel sampling-free uncertainty estimation method for object detection, and it is the first to provide separate uncertainties for each output signal: objectness, class, location and size.
- Real-Time Accurate 3D Head Tracking and Pose Estimation with Consumer RGB-D Cameras29
This work proposes another solution that utilizes a multi-camera system such that the data simultaneously acquired from multiple RGB-D sensors helps the tracker to handle challenging conditions that affect a subset of the cameras.
- One2Any: One-Reference 6D Pose Estimation for Any Object28
A novel method One2Any is proposed 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.
- NEWTON: Neural View-Centric Mapping for On-the-Fly Large-Scale SLAM28
Newton is proposed, a view-centric mapping method that dynamically constructs neural fields based on run-time observation that enables camera pose updates using loop closures and scene boundary updates by representing the scene with multiple neural fields, where each is defined in a local coordinate system of a selected keyframe.
- LoRACLR: Contrastive Adaptation for Customization of Diffusion Models27
LoRACLR is presented, 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.
- IPCC-TP: Utilizing Incremental Pearson Correlation Coefficient for Joint Multi-Agent Trajectory Prediction27
IPCC-TP is proposed, a novel relevance-aware module based on Incremental Pearson Correlation Coefficient to improve multi-agent interaction modeling and can be conveniently embedded into existing multi- agent prediction methods to extend original motion distribution decoders.
- Bending Graphs: Hierarchical Shape Matching using Gated Optimal Transport27
A hierarchical learning design is investigated, to which local patch-level information and global shape-level structures are incorporated, which enables correspondence prediction and provides rich features for the matching stage.
- From 2D to 3D: Re-thinking Benchmarking of Monocular Depth Prediction26
This work proposes a set of metrics well suited to evaluate the 3D geometry of MDP approaches and a novel indoor benchmark, RIO-D3D, crucial for the proposed evaluation methodology, based on a real-world dataset featuring high-quality rendered depth maps obtained from RGB-D reconstructions.
- Adversarial Domain Feature Adaptation for Bronchoscopic Depth Estimation26
This work proposes an alternative domain-adaptive approach to depth estimation from monocular images that improves the network's performance on real images by a considerable margin and can be employed in 3D reconstruction pipelines.
- 26
- Go with the Flows: Mixtures of Normalizing Flows for Point Cloud Generation and Reconstruction25
This work generalizes prior work by introducing additional discrete latent variable in mixture model, and demonstrates that by adding data augmentation, individual mixture components can learn to specialize in a semantically meaningful manner.
- 25
- Active Data Curation Effectively Distills Large-Scale Multimodal Models24
This work explores an alternative, yet simple approach—active data curation as effective distillation for contrastive multimodal pretraining, and finds such an active curation strategy to in fact be complementary to standard KD.
- Epipolar Geometry Improves Video Generation Models24
This work aligns diffusion models using pairwise epipolar geometry constraints using pairwise epipolar geometry constraints via preference-based optimization, directly addressing unstable trajectories and geometric artifacts through mathematically principled geometric enforcement.
- 24
- 24
- 24
- Binary DAD-Net: Binarized Driveable Area Detection Network for Autonomous Driving22
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, which outperforms state-of-the-art semantic segmentation networks on public datasets.
- 21
- Graphite: Graph-Induced Feature Extraction for Point Cloud Registration21
This work 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.
- 21
- 21
- UNOPose: Unseen Object Pose Estimation with an Unposed RGB-D Reference Image20
Evaluated on the 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.
- 20
- 3D Compositional Zero-Shot Learning with DeCompositional Consensus20
This work presents 3D Compositional Zero-shot Learning as a problem of part generalization from seen to unseen object classes for semantic segmentation, and proposes DeCompositional Consensus, which combines a part segmentation network with a part scoring network.
- 20
- 20
- Denoising Diffusion via Image-Based Rendering19
This work introduces the first diffusion model able to perform fast, detailed reconstruction and generation of real-world 3D scenes, and introduces 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.
- 19
- 19
- Lidar Upsampling With Sliced Wasserstein Distance19
A method is proposed for lidar point cloud upsampling which can reconstruct fine-grained lidar scan patterns and utilize edge-aware dense convolutions for both feature extraction and feature expansion.
- Joint detection and tracking in videos with identification features19
This work proposes the first joint optimization of detection, tracking and re-identification features for videos at low frame-rate, and maintains the detector performance, a typical multi-task challenge.
- Dealing with Ambiguity in Robotic Grasping via Multiple Predictions19
This work augments a fully convolutional neural network with a multiple hypothesis prediction model that predicts a set of grasp hypotheses in under 60ms, which is critical for real-time robotic applications.
- Hierarchical Multi-Organ Segmentation Without Registration in 3D Abdominal CT Images19
A novel framework for the segmentation of multiple organs in 3D abdominal CT images, which does not require registration with an atlas is presented, which uses discriminative classifiers that have been trained on an array of 3D volumetric features and implicitly model the appearance of the organs of interest.
- 19
- An Algorithm for Efficient and Exhaustive Template Matching19
Experimental results show that the proposed algorithm for efficient and exhaustive template matching based on the Zero mean Normalized Cross Correlation function is effective in speeding up the standard procedure and that the behavior follows that obtained by the BPC technique in the NCC case.
- 4D Gaussian Splatting SLAM18
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 is proposed and achieves robust tracking and high-quality view synthesis performance in real-world environments.
- RelationField: Relate Anything in Radiance Fields18
RelationField is proposed, the first method to extract inter-object relationships directly from neural radiance fields, and represents relationships between objects as pairs of rays within a neural radiance field, effectively extending its formulation to include implicit relationship queries.
- 18
- 18
- Real-Time Online Adaption for Robust Instrument Tracking and Pose Estimation18
A novel method for instrument tracking in Retinal Microsurgery which is apt to withstand the challenges of RM visual sequences in terms of varying illumination conditions and blur and is general enough to deal with different background and tool appearances.
- Graffiti Detection Using a Time-Of-Flight Camera18
Experimental results show promising capabilities of the proposed approach to video-based graffiti detection, with improvements expected as the technology gets more mature.
- 18
- MOHO: Learning Single-View Hand-Held Object Reconstruction with Multi-View Occlusion-Aware Supervision17
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.
- Unsupervised Traffic Scene Generation with Synthetic 3D Scene Graphs17
This work proposes a method based on domain-invariant scene representation to directly synthesize traffic scene imagery without rendering and introduces an unsupervised neural network architecture for realistic traffic scene synthesis.
- SRH-Net: Stacked Recurrent Hourglass Network for Stereo Matching17
This paper decouple the 4D cubic cost volume used by 3D convolutional filters into sequential cost maps along the direction of disparity instead of dealing with it at once by exploiting a recurrent cost aggregation strategy.
- 17
- Robust Segmentation of Various Anatomies in 3D Ultrasound Using Hough Forests and Learned Data Representations17
This work proposes a highly adaptive learning-based method for fully automatic segmentation of ultrasound volumes that can robustly and automatically adapt to different anatomies yet enforcing appearance and shape constraints.
- Markerless Augmented Reality Using Image Mosaics17
This work proposes a practical approach that is suitable for environment where big planar structures are present and all the objects laying on the structure are composed into a large reference object using image mosaicing techniques, so that the problem is reduced to that of finding the pose from a single plane.
- InseRF: Text-Driven Generative Object Insertion in Neural 3D Scenes16
This study introduces InseRF, a novel method for generative object insertion in the NeRF reconstructions of 3D scenes based on a user-provided textual description and a 2D bounding box in a reference viewpoint, which is capable of controllable and 3D-consistent object insertion without requiring explicit 3D information as input.
- TokenFormer: Rethinking Transformer Scaling with Tokenized Model Parameters16
TokenFormer is introduced, 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.
- 16
- Multi-modal Video Surveillance Aided by Pyroelectric Infrared Sensors16
A multi-modal video sensor node designed for low-power and low-cost video surveillance able to detect changes in the environment and the on-board microprocessor implements an NCC-based change detection algorithm is presented.
- I2DFormer+: Learning Image to Document Summary Attention for Zero-Shot Image Classification15
This work proposes I2DFormer+, a novel transformer-based ZSL framework that jointly learn to encode images and documents by aligning both modalities in a shared embedding space and significantly outperforms previous unsupervised semantic embeddings under both zero-shot and generalized zero-shot learning settings on three public datasets.
- Know Your Neighbors: Improving Single-View Reconstruction via Spatial Vision-Language Reasoning15
Kyn, a novel method for single-view scene reconstruction that reasons about semantic and spatial context to predict each point's density, is proposed and it is shown that KYN improves 3D shape recovery compared to predicting density for each 3D point in isolation.
- Dynamic Hyperbolic Attention Network for Fine Hand-object Reconstruction15
This work proposes 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.
- SupeRGB-D: Zero-Shot Instance Segmentation in Cluttered Indoor Environments15
This work explores zero-shot instance segmentation (ZSIS) from RGB-D data to identify unseen objects in a semantic category-agnostic manner and presents a method that uses annotated objects to learn the “objectness” of pixels and generalize to unseen object categories in cluttered indoor environments.
- 15
- HyperSDFusion: Bridging Hierarchical Structures in Language and Geometry for Enhanced 3D Text2Shape Generation14
This work proposes HyperSD-Fusion, a dual-branch diffusion model that generates 3D shapes from a given text using a hyperbolic hierarchical representation, and is the first to explore the hyperbolic hierarchical representation for text-to-shape generation.
- Zero123-6D: Zero-shot Novel View Synthesis for RGB Category-level 6D Pose Estimation14
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.
- Physics-Encoded Graph Neural Networks for Deformation Prediction under Contact14
This work focuses on modeling the dynamics between a rigid mesh contacting a deformable mesh under external forces under external forces, and jointly learning geometry and physics reconstructs consistent and detailed deformations.
- CCD-3DR: Consistent Conditioning in Diffusion for Single-Image 3D Reconstruction14
CCD-3DR is presented, which exploits a novel centered diffusion probabilistic model for consistent local feature conditioning and outperforms all competitors by a large margin on synthetic benchmark ShapeNet-R2N2.
- 14
- RobotFusion: Grasping with a Robotic Manipulator via Multi-view Reconstruction14
An efficient approach to segment the sought objects out of a planar workbench as well as a novel algorithm to automatically estimate grasping points are proposed for 3D object reconstruction and grasping based on an articulated robotic manipulator.
- 14
- 14
- Gaussians-to-Life: Text-Driven Animation of 3D Gaussian Splatting Scenes13
Gaussians2Life, a method for animating parts of high-quality 3D scenes in a Gaussian Splatting representation, enables the creation of consistent, immersive 3D experiences for arbitrary scenes and enables the animation of a large variety of object classes.
- 13
- 13
- Gatekeeper: Improving Model Cascades Through Confidence Tuning12
This work introduces a novel loss function called Gatekeeper for calibrating smaller models in cascade setups that fine-tunes the smaller model to confidently handle tasks it can perform correctly while deferring complex tasks to the larger model.
- Semantic Library Adaptation: LoRA Retrieval and Fusion for Open-Vocabulary Semantic Segmentation12
This work introduces Semantic Library Adaptation (SemLA), a novel framework for training-free, test-time domain adaptation that 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.
- 12
- Segmenting Known Objects and Unseen Unknowns without Prior Knowledge12
For the first time in panoptic segmentation with unknown objects, the U3HS is trained without unknown categories, reducing assumptions and leaving the settings as unconstrained as in real-life scenarios.
- 12
- 12
- 12
- 12
- 12
- 12
- Unsupervised Novel View Synthesis from a Single Image11
This work pre-train a purely generative decoder model using a GAN formulation while at the same time training an encoder network to invert the mapping from latent code to images and shows that the framework achieves results comparable to the state of the art on ShapeNet.
- Computer Vision and Machine Learning for Surgical Instrument Tracking11
This chapter outlines how computer vision can support the surgeon during an intervention via the example of surgical instrument tracking in retinal microsurgery, and shows how to derive algorithms for simultaneous tool tracking and pose estimation based on random forests.
- M2SVid: End-to-End Inpainting and Refinement for Monocular-to-Stereo Video Conversion10
The Stable Video Diffusion model is extended 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 and modify the attention layers in SVD to compute full attention for discoccluded pixels.
- Contrastive Test-Time Composition of Multiple LoRA Models for Image Generation10
CLoRA is introduced, a training-free approach that addresses limitations in multi-concept image generation using LoRA models by updating the attention maps of multiple LoRA models at test-time, and leveraging the attention maps to create semantic masks for fusing latent representations.
- Text-to-3D by Stitching a Multi-view Reconstruction Network to a Video Generator10
This work introduces VIST3A, a general framework that enables high-quality text-to-3D generation and adapts direct reward finetuning, a popular technique for human preference alignment, to adapt direct reward finetuning, a popular technique for human preference alignment.
- KP-RED: Exploiting Semantic Keypoints for Joint 3D Shape Retrieval and Deformation10
KP-RED is presented, 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.
- Time-to-Label: Temporal Consistency for Self-Supervised Monocular 3D Object Detection10
This work argues that the temporal consistency on the level of object poses, provides an important supervision signal given the strong prior on physical motion and proposes a self-supervised loss which uses this consistency, in addition to render-and-compare losses defined on lidar point clouds, to refine noisy pose predictions and derive high-quality pseudo labels.
- Graffiti Detection Using Two Views10
A novel video surveillance approach designed to detect vandal acts occurring on the background of the monitored scene, such as graffiti painting on walls and surfaces, public and private property defacing or etching, unauthorized post sticking is presented.
- Prior2former - Evidential Modeling of Mask Transformers for Assumption-Free Open-World Panoptic Segmentation9
This work proposes Prior2Former (P2F), the first approach for segmentation vision transformers rooted in evidential learning that 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.
- 9
- LatentSwap3D: Semantic Edits on 3D Image GANs9
This work proposes LatentSwap3D, a semantic edit approach based on latent space discovery that can be used with any off-the-shelf 3D or 2D GAN model and on any dataset and outperforms others both qualitatively and quantitatively.
- 3D Adversarial Augmentations for Robust Out-of-Domain Predictions9
This work improves the robustness and generalization of both 3D object detection and 3D semantic segmentation methods to out-of-domain data by augmenting the training set with adversarial examples.
- 9
- 9
- 3D Scene Graphs: Open Challenges and Future Directions8
This survey provides a unified and critical review of 3DSGs, with particular emphasis on open challenges and future directions, and formalize 3DSGs under a common definition and analyze the principal modeling choices that characterize existing formulations, including node and edge attributes, hierarchical structure, dynamic scene representations, and affordance-aware extensions.
- GALA: Guided Attention with Language Alignment for Open Vocabulary Gaussian Splatting8
This paper presents GALA, a novel framework for open-vocabulary 3D scene understanding with 3D Gaussian Splatting (3DGS), and introduces a core contribution of GALA, a cross-attention module with two learnable codebooks that encode view-independent semantic embeddings.
- PointDiT: Pixel-Space Diffusion for Monocular Geometry Estimation8
This work introduces a minimalist pixel-space Diffusion Transformer, built on a plain ViT, that operates directly on raw 3D point map patches and is conditioned on image tokens from a pre-trained DINOv3.
- Test-Time Visual In-Context Tuning8
This work proposes test-time Visual In-Context Tuning (VICT), a method that can adapt VICL models on the fly with a single test sample and flips the role between the task prompts and the test sample and uses a cycle consistency loss to reconstruct the original task prompt output.
- Unified Semantic Transformer for 3D Scene Understanding8
This work introduces UNITE, a Unified Semantic Transformer for 3D scene understanding, a novel feed-forward neural network that unifies a diverse set of 3D dense semantic indoor tasks within a single model.
- Towards Real-Time Open-Vocabulary Video Instance Segmentation8
This paper analyzes the computational bottlenecks of state-of-the-art foundation models that performs OV- VIS, and proposes a new method, TROY-VIS, that significantly improves processing speed while maintaining high accuracy, and achieves the best tradeoff between accuracy and speed.
- 8
- LaRI: Layered Ray Intersections for Single-view 3D Geometric Reasoning8
An annotation pipeline using rendering engines, construct annotations for five public datasets, including synthetic and real-world data covering 3D objects and scenes, and proposes to predict the ray stopping index, which identifies valid intersecting pixels and layers from LaRI's output.
- Extracting Training Data from Document-Based VQA Models8
This work shows Vision-Language Models can memorize responses for training samples and regurgitate them even when the relevant visual information has been removed, indicating these models could divulge memorised sensitive information and therefore pose a privacy risk.
- 8
- CloudAttention: Efficient Multi-Scale Attention Scheme For 3D Point Cloud Learning8
This work redesigns set transformers into a hierarchical framework for shape classification and part and scene segmentation and proposes a local attention unit, which captures features in a spatial neighborhood, in order to mitigate the non-heterogeneity of point clouds.
- 3DPointCaps++: Learning 3D Representations with Capsule Networks8
3DPointCaps++ is presented for learning robust, flexible and generalizable 3D object representations without requiring heavy annotation efforts or supervision and introduces a cluster loss ensuring that the points reconstructed by a single capsule remain local and do not spread across the object uncontrollably.
- Content Disentanglement for Semantically Consistent Synthetic-to-Real Domain Adaptation8
This work proposes a novel, unsupervised, end-to-end domain adaptation network architecture that enables semantically consistent sim2real image transfer and performs content disentanglement by employing shared content encoder and fixed style code.
- Joint motion boundary detection and CNN-based feature visualization for video object segmentation8
A video object segmentation method which jointly uses motion boundary and convolutional neural network (CNN)-based class-level maps to carry out the co-segmentation of the frames to create initial object and background regions.
- Large scale and long standing simultaneous reconstruction and segmentation8
This work proposes a method to segment a 3D point cloud of a scene while simultaneously reconstructing it via Simultaneous Localization And Mapping (SLAM) and incrementally merges segments obtained from each input depth image in an unified global model leveraging the camera pose estimated via SLAM.
- Semantic parametric body shape estimation from noisy depth sequences8
An original approach for fitting a pre-defined parametric shape model to depth data by exploiting the 3D body pose tracked through a sequence of range maps acquired from off-the-shelf depth cameras.
- 8
- 8
- 7
- 7
- 7
- Adversarial Appearance Learning in Augmented Cityscapes for Pedestrian Recognition in Autonomous Driving7
This paper provides a pipeline for augmentation of the Cityscapes dataset with virtual pedestrians and reveals a novel generative network architecture for adversarial learning of the data-set lighting conditions.
- An Octree-Based Approach towards Efficient Variational Range Data Fusion7
This work presents an efficient way to fuse range data via a variational Octree-based minimization approach by taking the actual range data geometry into account, and explains how to dynamically adjust the optimizer's geometric structure via joining/splitting of Octree nodes.
- 7
- 7
- 7
- OracleGS: Grounding Generative Priors for Sparse-View Gaussian Splatting6
This work proposes OracleGS, a novel framework that reconciles generative completeness with regressive fidelity for sparse view Gaussian Splatting, and conditions the powerful generative prior on multi-view geometric evidence on multi-view geometric evidence.
- RaNeuS: Ray-adaptive Neural Surface Reconstruction6
The proposed RaNeuS1 proposes 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.
- D-SCo: Dual-Stream Conditional Diffusion for Monocular Hand-Held Object Reconstruction6
This work introduces centroid-fixed dual-stream conditional diffusion for monocular hand-held object reconstruction (D-SCo), tackling two predominant challenges and introduces a dual-stream denoiser to semantically and geometrically model hand-object interactions with a novel unified hand-object semantic embedding.
- SST: Real-time End-to-end Monocular 3D Reconstruction via Sparse Spatial-Temporal Guidance6
An end-to-end 3D reconstruction network SST is proposed, 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.
- ManiFlow: Implicitly Representing Manifolds with Normalizing Flows6
This work establishes that NFs trained on perturbed data implicitly represent the manifold in regions of maximum likelihood, and proposes an optimization objective that recovers the most likely point on the manifold given a sample from the perturbed distribution.
- 6
- 6
- A Divide et Impera Approach for 3D Shape Reconstruction from Multiple Views6
This paper proposes to rely on viewpoint variant reconstructions by merging the visible information from the given views by estimating the relative pose between all the pairs of the object and builds a clean 3D model which preserves the details from each viewpoint.
- Learning without prejudice: Avoiding bias in webly-supervised action recognition6
This work trains two independent CNNs, a RGB network on web images and video frames and a second network using temporal information from optical flow and demonstrates that training the networks independently is vastly superior to selecting the frames for the flow classifier by using the authors' RGB network.
- 6
- A PRACTICAL STEREO SYSTEM BASED ON REGULARIZATION AND TEXTURE PROJECTION6
It is shown that in scenes with moving objects spatial regularization methods are more accurate than spacetime stereo, while remaining computationally simpler, and an extension of regularization-based algorithms to the temporal domain is proposed, to further improve the performance ofregularization methods within dynamic scenes.
- 6
- OVI-MAP:Open-Vocabulary Instance-Semantic Mapping5
This work proposes to build a class-agnostic 3D instance map that is incrementally constructed from RGB-D input, while semantic features are extracted only from a small set of automatically selected views using vision-language models.
- SmileSplat: Generalizable Gaussian Splats for Unconstrained Sparse Images5
A novel generalizable Gaussian Splatting method is proposed to reconstruct pixel-aligned Gaussian surfels for diverse scenarios only requiring unconstrained sparse multi-view images, and achieves state-of-the-art performance in various 3D vision tasks.
- Toward a Diffusion-Based Generalist for Dense Vision Tasks5
This paper proposes to perform diffusion in pixel space and provide a recipe for finetuning pre-trained text-to-image diffusion models for dense vision tasks and shows competitive performance to the other vision generalists.
- Open-Structure: Structural Benchmark Dataset for SLAM Algorithms5
A baseline is proposed using the Open-Structure dataset to evaluate widely used modules, including camera pose tracking, parametrization, and factor graph optimization, within SLAM systems, to discern each module's strengths and weaknesses in the context of camera tracking and optimization processes.
- Human Pose Descriptions and Subject-Focused Attention for Improved Zero-Shot Transfer in Human-Centric Classification Tasks5
The proposed FocusCLIP, integrating subject-level guidance– a specialized mechanism for target-specific supervision–into the CLIP framework for improved zero-shot transfer on human-centric tasks, shows a 2.47% improvement over CLIP in zero-shot bird classification using the CUB dataset.
- OpenSUN3D: 1st Workshop Challenge on Open-Vocabulary 3D Scene Understanding5
An overview of the challenge hosted at the OpenSUN3D Workshop on Open-Vocabulary 3D Scene Understanding is provided, the challenge dataset, the evaluation methodology, and brief descriptions of the winning methods are presented.
- 5
- SC-Explorer: Incremental 3D Scene Completion for Safe and Efficient Exploration Mapping and Planning5
This work proposes a novel way to integrate deep learning into exploration by leveraging 3D scene completion by leveraging 3D scene completion for informed, safe, and interpretable exploration mapping and planning.
- 5
- Semantic Image Alignment for Vehicle Localization5
This paper presents 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, formulate the localization task as a direct image alignment problem on semantic images, which allows the approach to robustly track the vehicle pose in semantically labeled maps.
- BUILDING OUTLINE EXTRACTION FROM AERIAL IMAGES USING CONVOLUTIONAL NEURAL NETWORKS5
A Deep Learning (DL)-based approach is proposed to localize buildings, estimate the relative height information, and extract the buildings’ boundaries using a single aerial image and could be employed for real time applications.
- 5
- A Template Analysis Methodology to Improve the Efficiency of Fast Matching Algorithms5
This paper outlines a methodology to order the succession of bounding functions deployed by this technique based on the analysis of the template only, and proves that the proposed approach is able to achieve improved efficiency.
- 4
- ESCAPE: Equivariant Shape Completion via Anchor Point Encoding4
ESAPE (Equivariant Shape Completion via Anchor Point Encoding), a novel framework designed to achieve rotation-equivariant shape completion, is introduced, which employs a distinctive encoding strategy by selecting anchor points from a shape and representing all points as a distance to all anchor points.
- CausNVS: Autoregressive Multi-view Diffusion for Flexible 3D Novel View Synthesis4
The proposed CausNVS is a multi-view diffusion model in an autoregressive setting, which supports arbitrary input-output view configurations and generates views sequentially, and achieves consistently strong visual quality across diverse settings.
- 4
- Unsupervised Template Warp Consistency for Implicit Surface Correspondences4
This work investigates the inherent assumption that the implicit neural field optimization naturally leads to consistently warped shapes, thus providing both good shape reconstruction and correspondences and introduces explicit constraints using unsupervised sparse point predictions, directly encouraging consistency of the warped shapes.
- DisPositioNet: Disentangled Pose and Identity in Semantic Image Manipulation4
This work proposes DisPositioNet, a model that learns a disentangled representation for each object for the task of image manipulation using scene graphs in a self-supervised manner and shows that disentangling the feature representations in the latent manifold of the model outperforms the previous works qualitatively and quantitatively.
- 3D Local Descriptors—from Handcrafted to Learned4
This chapter provides a comprehensive review of the main proposals in the field of 3D feature description, based on different, handcrafted ways to encode and exploit the geometric properties of a given surface.
- Guest Editors’ Introduction to the Special Issue on RGB-D Vision: Methods and Applications4
This special issue focuses 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, to stimulate researchers from different fields to present their state-of-the-art work.
- 4
- Co-segmentation via visualization4
Compared experiments conducted on two benchmark datasets, iCoseg and MSRC, illustrate the superior performance of the proposed approach over state-of-the-art algorithms.
- 4
- SING3R-SLAM: Submap-Based Indoor Monocular Gaussian SLAM With 3D Reconstruction Priors3
This work proposes SING3R-SLAM, a globally consistent Gaussian-based monocular indoor SLAM framework that represents the scene with a Global Gaussian Map that serves as a persistent, differentiable memory, incorporates local geometric reconstruction via submap-level global alignment, and leverages consistency of the global map to further refine local geometry.
- RefAM: Attention Magnets for Zero-Shot Referral Segmentation3
This work introduces a new method that directly exploits features from diffusion transformers for downstream tasks, requiring neither architectural modifications nor additional training, and develops RefAM, a simple training-free grounding framework that combines cross-attention maps, GAS handling, and redistribution.
- 3
- Elastic3D: Controllable Stereo Video Conversion with Guided Latent Decoding3
Elastic3D, a controllable, direct end-to-end method for upgrading a conventional video to a binocular one, based on (conditional) latent diffusion, avoids artifacts due to explicit depth estimation and warping and sets a new standard for reliable, controllable stereo video conversion.
- Neural Semantic Map-Learning for Autonomous Vehicles3
This work presents 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.
- Object-Aware Monocular Depth Prediction With Instance Convolutions3
This work proposes a novel convolutional operator which is explicitly tailored to avoid feature aggregation of different object parts, based on estimating per-part depth values by means of super-pixels, and demonstrates the advantages of Instance Convolutions over the classical convolution at estimating depth around occlusion boundaries.
- 3
- Universal Hough dictionaries for object tracking3
This paper aims to demonstrate the efforts towards in-situ applicability of EMMARM, as to provide real-time information about the physical properties of the central nervous system and its complications.
- Detecting motion by means of 2D and 3D information3
Two change detection strategies based on two views which exploits 2D and 3D information in order to deal with typical change detection issues are proposed and compared.
- Stitched Value Model for Diffusion Alignment2
This work proposes StitchVM, a model stitching framework that efficiently transfers reward models pretrained for clean images to the noisy latent regime, and shows that this approach yields improvements across a broad range of downstream steering and post-training methods.
- Future Dynamic 3D Reconstruction: Toward 3D World Modeling with Disentangled Ego-Motion2
FR3D is proposed, a world model that predicts a persistent 3D latent representation for future dynamic 3D reconstruction and introduces a teacher-student distillation strategy that leverages the spatial ”com-mon sense” of off-the-shelf foundation models, leading to robust zero-shot generalization.
- SSL-R1: Self-Supervised Visual Reinforcement Post-Training for Multimodal Large Language Models2
This work revisits self-supervised learning in visual domains and reformulates widely-used SSL tasks into a set of verifiable visual puzzles for RL post-training, requiring neither human nor external model supervision.
- DataComp-VLM: Improved Open Datasets for Vision-Language Models2
It is found that data mixing, not filtering, is key to a high-quality training dataset: instruction-heavy mixtures scale better than caption-heavy ones, with gains widening at larger scales.
- MOBIUS: Big-to-Mobile Universal Instance Segmentation via Multi-modal Bottleneck Fusion and Calibrated Decoder Pruning2
MOBIUS establishes a new benchmark for efficient segmentation on both highperformance computing platforms and mobile devices, designed for Pareto-optimal downscaling to support deployment across devices ranging from high-end accelerators to mobile hardware.
- 2
- 2
- 2
- 2
- Explicit Domain Adaptation With Loosely Coupled Samples2
This work proposes a transfer learning framework, core of which is learning an explicit mapping between domains, beneficial for safety-critical applications, like autonomous driving.
- 2
- Predicting Multiple Actions for Stochastic Continuous Control2
This work introduces a new approach to estimate continuous actions using actor-critic algorithms for reinforcement learning problems and predicts M actions with the policy network and then uniformly sample one action during training as well as testing at each state.
- Automatic Initialization and Failure Detection for Surgical Tool Tracking in Retinal Microsurgery2
An algorithm for robustly detecting the bounding box around the tool tip together with a failure detection of the tracking algorithm is proposed, transforming the user input dependent algorithm into a completely automatic framework without the need of an assistant.
- 2
- 2
- 2
- A Radial Search Method for Fast Nearest Neighbor Search on Range Images2
When tested against open source implementations of state-of-the-art NNS algorithms, radial search obtains better performance than the other algorithms in terms of speedup, while yielding the same level of accuracy.
- Bounded Non-Local Means for Fast and Effective Image Denoising2
This paper proposes a method aimed at improving the computational efficiency of NLM by quick pre-selection of dissimilar patches thanks to a rapidly computable upper bound of the weighting function.
- The Maximal Self-dissimilarity Interest Point Detector2
This work proposes a novel interest point detector stemming from the intuition that image patches which are highly dissimilar over a relatively large extent of their surroundings hold the property of being repeatable and distinctive, thereby achieving enhanced repeatability, distinctiveness and localization accuracy.
- Featurising Pixels from Dynamic 3D Scenes with Linear In-Context Learners1
This work presents a framework that learns pixel-accurate feature descriptors from videos, LILA, which leverages spatio-temporal cue maps and trains effectively on uncurated video datasets, embedding semantic and geometric properties in a temporally consistent manner.
- FLAT: Feedforward Latent Triangle Splatting for Geometrically Accurate Scene Generation1
FLAT is introduced and, for the first time, it is shown that triangle splats can be decoded directly from video diffusion latents and converted into a fully opaque, game-engine-ready representation that supports real-time rendering.
- Good Token Hunting: A Hitchhiker's Guide to Token Selection for Visual Geometry Transformers1
This work introduces a two-stage framework for effective token selection, and highlights the advantage of a diversity-based strategy for inter-frame selection, which ensures broad coverage of the scene and offers a superior speed-accuracy trade-off compared to existing solutions.
- 1
- OpenGaFF: Open-Vocabulary Gaussian Feature Field with Codebook Attention1
OpenGaFF is presented, a novel framework for open-vocabulary 3D scene understanding built upon 3D Gaussian Splatting that models semantics as a continuous function of Gaussian geometry and appearance and introduces a structured codebook that serves as a set of shared semantic primitives.
- 3D Segmentation Using Viewpoint-Dependent Spatial Relationships1
A viewpoint representation is introduced that encodes camera poses and conditions the model on the observation viewpoint, improving segmentation accuracy on viewpoint-dependent relations and increasing mIoU from 0.30 to 0.47 compared to a model without viewpoint conditioning.
- PARCEL: Pool-Anchored Resampling with Conditioned Elastic Queries for Efficient Vision-Language Understanding1
PARCEL improves the performance-efficiency Pareto frontier, consistently outperforming existing matryoshka baselines across visual-token budgets while preserving the"train once, deploy anywhere"paradigm.
- Lost in Translation? Vocabulary Alignment for Source-Free Adaptation in Open-Vocabulary Semantic Segmentation1
This work introduces VocAlign, a novel source-free domain adaptation framework specifically designed for VLMs in open-vocabulary semantic segmentation that adopts a student-teacher paradigm enhanced with a vocabulary alignment strategy, which improves pseudo-label generation by incorporating additional class concepts.
- UIP2P: Unsupervised Instruction-Based Image Editing via Edit Reversibility Constraint1
An unsupervised instruction-based image editing approach that removes the need for ground-truth edited images during training and introduces a novel editing mechanism called Edit Reversibility Constraint (ERC), which applies forward and reverse edits in one training step and enforces alignment in image, text, and attention spaces.
- 1
- 1
- 1
- View-to-Label: Multi-View Consistency for Self-Supervised 3D Object Detection1
This work proposes a novel approach to self-supervise 3D object detection purely from RGB sequences alone, leveraging multi-view constraints and weak labels.
- 3D Compression Using Neural Fields1
This work leverages previous advances and proposes a novel NF-based compression algorithm for 3D data that excels at geometry compression on 3D point clouds as well as meshes and shows that, due to the NF formulation, it is straightforward to extend the compression algorithm to compress both geometry and attribute of3D data.
- Re-Nerfing: Improving Novel View Synthesis through Novel View Synthesis1
Re-Nerfing is proposed, a simple and general multi-stage approach that leverages NeRF’s own view synthesis to address limitations in shape-radiance ambiguities and brings improvements to the state-of-the-art Zip-NeRF, even when trained with all views.
- ParGAN: Learning Real Parametrizable Transformations1
This work proposes ParGAN, a generalization of the cycle-consistent GAN framework to learn image transformations with simple and intuitive controls, which takes as input both an image and a parametrization of the transformation.
- 1
- 1
- 1
- 1
- Abstract: Real-Time Online Adaption for Robust Instrument Tracking and Pose Estimation1
A robust closed-loop framework to track and localize the instrument parts in real-time, based on a dual-random forest (RF), which predicts the locations of the tool joints based on HOG features.
- 1
- Adaptive Low Resolution Pruning for fast Full Search-equivalent pattern matching1
This paper proposes a computational analysis of several incremental techniques specifically designed to enhance the efficiency of Low Resolution Pruning and proposes a novel LRP algorithm aimed at minimizing the theoretical number of operations by adaptively exploiting different incremental approaches.
- Second-Order Polynomial Models for Background Subtraction1
This paper proposes here a thorough theoretical analysis aimed at identifying a family of suitable models and deriving the closed-form solutions of the associated regression problems and presents a detailed quantitative experimental evaluation aimed at comparing the different background subtraction algorithms resulting from theoretical analysis.
- IMPROVING GEOMETRIC HASHING BY MEANS OF FEATURE DESCRIPTORS1
A novel method aimed at improving the performance of Geometric Hashing in terms of robustness toward occlusion and clutter is proposed, which employs feature descriptors to notably decrease the amount of false positives that generally arise under these conditions.
- Methodologies for visual correspondence1
The dissertation presents original contributions to the theory of visual correspondence, as well as applications dealing with 3D reconstruction and multi-view video surveillance, including fast pattern matching, stereo correspondence and robust image matching.
- Enhanced Low-Resolution Pruning for Fast Full-Search Template Matching1
This paper proposes an original contribution with respect to Gharavi-Alkhansari's work that is based on the exploitation of an initial estimation of the global minimum aimed at increasing the efficiency of the pruning process.
- 1
- Robust Visual Correspondence: Theory and Applications1
The visual correspondence task can be extremely challenging in presence of disturbance factors which typically affect images, and the acquisition of images from different spatial points or different time instants can also induce occlusions.
- –
- –
- R-CoV: Region-Aware Chain-of-Verification for Alleviating Object Hallucinations in LVLMs–
R-CoV is proposed, a visual chain-of-verification method to alleviate object hallucinations in LVLMs in a post-hoc manner that can be seamlessly integrated into various LVLMs in a training-free manner and without relying on external detection models.
- –
- –
- –
- –
- –
- –
- –
- –
- SA4Depth: Consistent Pose-Depth Scale Alignment for Self-Supervised Monocular Depth Estimation–
This work uses the depth estimated during training to reproject learnable visual features across consecutive frames and refine the pose estimates by reducing feature alignment residuals and introduces SA4Depth, an approach to improve this alignment and boost the depth predictions while keeping the inference time unchanged.
- DynaTok: Token-Based 4D Reconstruction from Partial Point Clouds–
D DynaTok is proposed, a point-based framework for correspondence-free 4D reconstruction from partial point cloud sequences without images that encodes frames into compact latent tokens, aggregates incomplete observations over time with a Transformer-based spatiotemporal encoder, and decouples geometry and motion through residual tokens in a unified model.
- –
- –
- –
- SG2Loc: Sequential Visual Localization on 3D Scene Graphs–
This paper introduces a novel, lightweight approach to sequential visual localization using 3D scene graphs, leveraging a compact scene graph and efficient semantic matching, which significantly reduces storage while maintaining performance on real-world datasets.
- –
- The Art of Interrogation: Consistency Amplifies Factuality in Spatial Reasoning–
This work proposes a self-supervised reinforcement learning (RL) framework that targets the internal reasoning process without requiring ground-truth annotations and shows that this label-free consistency training approaches the accuracy of models trained with ground-truth supervision and achieves similar generalization across diverse tasks and data domains.
- –
- Prompting Diffusion Models for Zero-Shot Instance Segmentation–
The results suggest that the rich priors encoded in generative pretraining, combined with principled spatial conditioning, offer a compelling path toward broadly generalizing interactive segmentation without large-scale mask supervision.
- Language-Guided Open-World Anomaly Segmentation–
Clipomaly is proposed, the first CLIP-based open-world and anomaly segmentation method for autonomous driving that dynamically extends its vocabulary at inference time without retraining, enabling robust detection and naming of anomalies beyond common class definitions such as those in Cityscapes.
- –
- –
- –
- PRISM: Probabilistic Representation for Integrated Shape Modeling and Generation–
This work introduces PRISM, a novel compositional approach for 3D shape generation that integrates categorical diffusion models with Statistical Shape Models (SSM) and Gaussian Mixture Models (GMM) and uses GMM to represent part semantics in a continuous space.
- –
- –
- –
- –
- –
- –
- –
- –
- –
- –
- –
- Abstract: Leveraging Web Data for Skin Lesion Classification–
The success of deep learning is mainly based on the assumption that for the given application, there is access to a large amount of annotated data, and the representativeness of the training set is still limited by the number of available samples.
- –
- –
- –
- –
- –
- –
- –
- –
Publication data from OpenAlex, with missing venues and authors filled in from Crossref; citation counts are the higher of OpenAlex and Semantic Scholar, last synced 2026-10-15. One-sentence summaries under some papers are written by Semantic Scholar’s model. Citation counts may be lower than on Google Scholar, which indexes more sources.
Report an error
Wrong papers, two people merged into one, or a profile that should not be here? Tell us and we will fix or hide it.