ECCV2026 3D Vision AI paper notes paper summaries 3D Gaussian Splatting Dynamic Scenes Human Pose Remote Sensing Compression Few-/Zero-Shot Learning

🧊 3D Vision¶

🎞️ ECCV2026 · 85 paper notes

📌 Same area in other venues: 📷 CVPR2026 (751) · 🔬 ICLR2026 (197) · 🧪 ICML2026 (30) · 🤖 AAAI2026 (79) · 🧠 NeurIPS2025 (116) · 📹 ICCV2025 (267)

🔥 Top topics: 3D Gaussian Splatting ×18 · Dynamic Scenes ×5 · Human Pose ×5 · Remote Sensing ×4 · Compression ×4

2D Features Are All You Need for 3D Shape Understanding: MeshFM fits corrected 2D foundation-model features into continuous 3D teacher fields, then trains a rotation-augmented feedforward predictor to reproduce them without manual 3D annotations, supporting segmentation, correspondence, and deformation with the same features and achieving 0.549/0.539 semantic mIoU on original/rotated PartObjaverse-Tiny shapes.
3D-LENS: A 3D Lifting-based Elevated Novel-view Synthesis method for Single-View Aerial-Ground Re-Identification: 3D-LENS lifts images from a single real viewpoint domain into fixed textured meshes, renders missing aerial or ground views, and combines appearance alignment with robust mixed-domain training, reaching 37.8% mAP on AG-ReID.v2 CCTV-to-aerial retrieval under aerial-only training, 21.2 percentage points above the strongest prior method in that column, PASS, without real target-view training samples.
3D-ReGen: A Unified 3D Geometry Regeneration Framework: MeshReGen represents coarse meshes, incomplete point clouds, and masked geometry as VecSet conditions, concatenates their tokens with noisy target tokens, and learns image-guided 3D regeneration from automatically constructed degradation pairs, reaching CD 0.0081 and F-score 0.4913 on four-view GSO reconstruction while also supporting enhancement and local editing.
3D Gaussian Splatting Compression with Object Scalability: RecastGS reorganizes pretrained 3DGS into a hierarchy with object-specific quality choices, and LayeredCGS compresses it for progressive decoding; on kitchen near 2.3 MB, ROI PSNR rises from 19.65 to 21.81 dB, at the cost of lower background and full-image quality.
3D Gaussian Texture for Real-time Mesoscale Appearance Synthesis and Rendering: The paper organizes 3D Gaussians into regular texture layers at similar heights above a surface, enabling real-time exemplar-based synthesis, transfer to new meshes, and basic relighting of structures such as grass, durian spikes, and wicker gaps; it improves SIFID over NeRF-Texture in all four comparisons, with reported times of 3.28–4.98 ms in Figure 4.
3D Scene-Adaptive Trajectory-Controllable Human Image Animation with Camera Movement: 3STC-HIA adjusts human altitude and orientation along a user-defined path in a unified 3D scene, then guides a pretrained video model with visible scene point clouds for training-free human and camera control, reducing Trajectory100 translation error from RealisMotion's 1.305 m to 0.478 m without a corresponding improvement in yaw accuracy.
3DGS3: Joint Super Sampling and Frame Interpolation for Real-Time Large-Scale 3DGS Rendering: Instead of rasterizing every output frame at high resolution, 3DGS3 jointly recovers spatial and temporal resolution through analytical-gradient-guided super sampling and lightweight frame interpolation, increasing 3DGS throughput from 17.35 to 96.47 FPS at 4096×2160 on Mip-NeRF 360 while achieving 49.22 dB PSNR against high-resolution reference renderings.
4D-VGGT: A SpatioTemporal Foundation Model for Dynamic Scene Geometry Estimation: 4D-VGGT organizes inputs with an adaptive view-time grid, learns spatial and motion cues through parallel cross-view global and cross-time local fusion, and routes them to five geometry heads, achieving a camera-pose ATE of 0.017 on Bonn while using ablations to demonstrate distinct spatial and temporal contributions.
4DGS360: 360° Gaussian Reconstruction of Dynamic Objects from a Single Video: 4DGS360 uses confident 2D tracks as spatiotemporal anchors for 3D tracking, initializes motion with geometry that includes occluded regions, and then optimizes dynamic Gaussians without diffusion in its base pipeline, reducing mean LPIPS on five iPhone scenes from HiMoR's 0.4639 to 0.3877.
A Benchmark for Heterogeneous Stereo Deblurring with Physically- and Epipolar-constrained Cross Attention: The paper introduces HSD for asymmetric wide/ultra-wide smartphone deblurring and PECA for cross-view retrieval within a physically plausible, directional epipolar window, improving NAFNet from 32.16 to 32.92 dB on HSD without the cost of global matching.
A second-order theory of texture for depth from focus: The paper explains why apparently textureless surfaces can provide subjective speckle for depth from focus, and enhances that signal with narrowband filtering under ambient light, at the cost of longer exposure and with material- and illumination-dependent benefits.
Abstract the Layout, Focus the Detail: A Dual-Granularity Representation Framework for Zero-Shot 3D Visual Grounding: The framework recasts a 3D scene as an abstract layout for spatial reasoning and object-centric patches for appearance verification, enabling an off-the-shelf VLM to select targets in a spatial-first, visual-second workflow and achieve 45.0% [email protected] on ScanRefer and 63.2% accuracy on Nr3D without GT object classes.
AC3S: Adaptive Conditioning for 3D-Aware Synthetic Data Generation: This paper proposes AC3S, which dynamically attenuates the conditioning injection strength of ControlNet using an adaptive visual prompt modulator, supplemented by a multi-agent VLM system to generate text prompts consistent with geometric structures. While maintaining precise 3D pose alignment, this approach significantly enhances the realism and diversity of synthetic images, reducing the FID by 15.95 points compared to the baseline 3D-DST and achieving notable improvements in downstream classification and pose estimation tasks.
ActiveStructure: Plane Scene Graph-Guided Active 3D Gaussian Splatting: ActiveStructure augments active 3DGS mapping with a persistent plane scene graph, uses relational anomalies and localized 3D gaps to select views, and turns established walls from exploration attractors into suppressors, achieving 97.47% coverage and 32.13 dB PSNR on Replica within a fixed 2000-frame budget.
Ada-VNNs: Adaptive Equivariance for Vector Neural Networks: Ada-VNNs attach a residual branch with relaxed coordinate-wise weight sharing to a strictly equivariant vector-neuron backbone, raising VNN's 10° registration recall from 8.87% to 70.31% on approximately continuously self-symmetric ShapeNet categories, while introducing task-dependent trade-offs rather than universal improvements.
AdaptiveSplat: Texture Aware Controllable 3D Gaussian Allocation for Feed-Forward Reconstruction: AdaptiveSplat uses local texture complexity to decide where to prune 3D Gaussians, then predicts updated attributes through a mask-conditioned head, enabling budget-controlled reconstruction without test-time scene optimization and reaching 20.740 dB PSNR on RE10K after pruning 80% of the primitives.
AerialMetric: Benchmarking and Adapting UAV Monocular Metric Depth Estimation in the Real World: This paper constructs a large-scale UAV aerial monocular metric depth benchmark, AerialMetric, consisting of four complementary subsets (52K real + 16K synthetic image-depth pairs). It systematically reveals the severe domain gap of existing SOTA models under aerial perspectives where δ1 is close to 0, and improves the aerial δ1 to over 84% through parameter-efficient fine-tuning of MoGe2 with LoRA, with almost no loss in ground generalization capability.
Agentic Collaborative Cognition for Zero-Shot 3D Understanding: This paper reformulates zero-shot 3D scene understanding as an iterative collaboration between a Planning Agent and a Perception Agent centered around an explicit holistic cognitive map. By actively planning key views, recording object attributes across viewpoints, and filtering candidates through feedback, the proposed method significantly outperforms existing zero-shot approaches on 6 benchmarks, including ScanRefer, Nr3D, SQA3D, and ScanQA.
Aggregating Cross-Domain Knowledge via Learnable Tokens for Multi-Teacher Distillation: ACTok places channel-concatenated learnable tokens and alternating cross-space/intra-space interactions after a student backbone to consolidate CLIP, MAE, and DINO knowledge using ImageNet-1K, reaching 94.4% on Caltech101 with ViT-B without universally outperforming the strongest teacher or distillation baseline.
AIMold: An Autonomous AI-based Pipeline for Complex Mold Design: AIMold combines orientation selection, coarse voxel assembly generation, latent geometry refinement, and CAD post-processing, raising auxiliary-component COV from the VecSet baseline's 21.24% to 57.29% on the MoldCAD test split while still requiring downstream manufacturing verification.
AirSplat: Alignment and Rating for Robust Feed-Forward 3D Gaussian Splatting: AirSplat fine-tunes only DA3's Gaussian prediction head, corrects misaligned supervision through training-time self-consistent pose alignment, and suppresses floaters with teacher-rated opacity constraints, raising PSNR from 20.78 to 23.08 dB on RE10K with 12 input views.
AirZoo: A Unified Large-Scale Dataset for Grounding Aerial Geometric 3D Vision: AirZoo constructs a million-scale synthetic aerial dataset. By utilizing an AirSim-Cesium-Unreal tri-simulation pipeline, it automatically renders UAV aerial sequences with pixel-level depth and 6-DoF georeferenced poses from Google 3D Tiles, covering 377 multi-weather trajectories across 22 countries. Fine-tuning existing SOTA models on three tasks—aerial image retrieval, cross-view matching, and multi-view 3D reconstruction—yields significant improvements, particularly enhancing zero-shot generalization capabilities in real-world scenes.
AnchorSplat: Fast and Structure Consistent Detail Synthesis for Gaussian Splatting: AnchorSplat treats a low-quality Gaussian asset as an attributed point cloud and directly synthesizes 3D detail through local point anchors and single-pass multi-Gaussian generation, reaching 36.57 dB PSNR on 3DGS-SR with a reported processing time of approximately 0.01 seconds per asset without accessing original multi-view images at inference.
Any to Full: Prompting Depth Anything for Depth Completion in One Stage: Any2Full encodes sparse depth as scale prompts that guide geometrically consistent relative depth prediction inside frozen Depth Anything v2, then recovers metric scale in a single final fit, achieving an average rank of 2.3 across domain and missing-pattern evaluations and approximately 1.4 times the inference speed of PriorDA with the same DA-L backbone.
AnyGround3D: Towards Grounding Any 3D Object in the Wild via 2D-to-3D Lifting: AnyGround3D automatically lifts monocular images from sources such as COCO and KITTI into 3D grounding supervision containing scene point clouds, complete-object 3D boxes, and natural-language descriptions, then distills DINOv2 semantics into its 3D encoder; it obtains 66.76/31.34 [email protected]/0.5 on the 3EED union split and 62.71/54.79 on the Mono3DRefer test set without target-specific fine-tuning.
AnyMatch: Supercharging Universal Multi-Modal Image Matching with Large-Scale Single-View Images: AnyMatch proposes a framework for synthesizing multimodal, multiview matching data from vast collections of single-view images at zero cost. Monocular depth estimation + 3D reprojection + diffusion inpainting enable geometrically consistent viewpoint transformations, while cross-modal image translation generates four modalities—infrared/depth/normals/events—and sample-level geometric consistency verification (SGCV) filters generation artifacts; after fine-tuning LoFTR/EDM/RoMa on the synthetic Any-syn dataset, cross-modal matching performance substantially surpasses existing synthesis methods such as MINIMA, with strong zero-shot generalization to unseen medical/remote-sensing modalities.
Audio-Visual Camera Pose Estimation with Passive Scene Sounds and In-the-Wild Video: This paper proposes the first method that leverages passive scene audio to assist vision-based relative camera pose estimation. By performing late fusion of a spatial audio encoder (SAE)—composed of a direction-of-arrival (DOA) spectrum and a binauralized embedding—with the SOTA visual pose estimation model Reloc3r, this method achieves significant improvements in both real-world scenes and simulated datasets.
BioMTBee: Biologically Constrained Multi-View Template-Based 3D Reconstruction of Bumblebee: BioMTBee fits a micro-CT-derived bumblebee template to multi-view video using filtered 3D poses, silhouettes, and biological constraints, reducing mesh-keypoint error from MAMMAL's 0.189 mm to 0.152 mm while increasing silhouette IoU from 0.696 to 0.758 on the DBS test split.
Boosting 6D Object Pose Estimation via Monocular Depth Cues: MDC-Net couples monocular-depth calibration, geometric inlier selection, and recurrent pose refinement, achieving 66.9% mean AR across seven BOP datasets with one pose hypothesis and no real depth at inference, although several auxiliary results and training descriptions are inconsistent.
Boosting Correspondence Learning with Structure-Aware Estimator: A structure-aware estimator replaces independent weighted least squares at the end of correspondence-learning networks, using learned inlier relationships to improve DeMatch's YFCC100M AUC@5° from 30.95 to 36.58, with important limitations on indoor–outdoor transfer and correspondence-pruning architectures.
BrepLLM: Enabling Large Language Models to Understand Boundary Representations: BrepLLM is the first framework that enables large language models to directly parse and reason over native B-rep (Boundary Representation) CAD models. By performing adaptive UV sampling to convert B-rep models into face-edge topology graphs, and using a hierarchical BrepEncoder to extract decoupled face, edge, and topology features, followed by CLIP contrastive pretraining and two-stage progressive LLM fine-tuning, it comprehensively outperforms point cloud- and image-based baseline models on 3D object description and classification tasks.
Capacity-Controlled Multi-View Stylization of 3D Gaussian Splatting: This paper proposes a multi-view stylization framework for 3D Gaussian Splatting based on capacity-controlled optimal transport. It replaces greedy nearest-neighbor matching with semi-balanced optimal transport, suppressing the many-to-one problem through column capacity constraints. Simultaneously, it introduces cross-view match guidance to ensure style consistency between adjacent views, assisted by geometric regularization to enhance reconstruction quality. The proposed method comprehensively outperforms existing approaches in style quality, content fidelity, and multi-view consistency.
DefenseSplat: Enhancing the Robustness of 3D Gaussian Splatting via Frequency-Aware Filtering: DefenseSplat significantly improves the robustness of 3DGS against adversarial attacks without clean ground-truth supervision. It decomposes input images using the wavelet transform and filters out high-frequency subbands to suppress adversarial perturbations, while introducing a ReLU-based scale regularization loss to constrain the generation of elongated Gaussians.
DeWorldSG: Depth-Aware 3D Semantic Scene Graph Generation via World-Model Priors: DeWorldSG proposes a framework for generating spatiotemporally robust 3D semantic scene graphs (SSGs) from RGB-D sequences: SAM-mask-guided dual-domain depth filtering estimates an instance-level probabilistic 3D Gaussian distribution for each object, while the V-JEPA 2 world model accumulates cross-frame relational evidence and refines predicate edges through uncertainty-aware fusion, improving relationship recall by 77.4% and predicate recall by 23.2% on 3DSSG.
Diffusion-Based Material Regularization for Physics-Based Inverse Rendering: This paper proposes regularizing physically disentangled inverse rendering with similarity kernels derived from diffusion-predicted G-buffers (albedo, roughness, metallic, and normals): rather than fitting these predictions as target values, it uses them to define material similarity between surface regions, applying joint bilateral filtering to encourage consistent optimized parameters within the same material region, thereby suppressing artifacts caused by lighting baked into materials and significantly outperforming SOTA methods such as Neural-PBIR and MaterialFusion on Stanford-ORB.
DLGStream: Dynamic Language-embedded Gaussian Splatting for Open-vocabulary Enabled Free-viewpoint Video Streaming: DLGStream introduces a dual-opacity language Gaussian representation and a deformation field based on temporal feature interpolation, embedding streamable spatiotemporal language features while reconstructing high-quality free-viewpoint video from multiview dynamic scenes, using an average frame size of 43 KB to enable real-time 4D open-vocabulary queries and frame interpolation.
EPO: Boosting 3D Foundation Models with Edge-based Pose Optimization: This paper proposes EPO, a geometric optimization framework that does not require explicit feature tracks. It utilizes Canny edge maps and distance transform fields to construct a differentiable edge reprojection loss, refining the camera parameters and depth maps output by 3D foundation models through a two-stage adaptive optimization. It achieves or exceeds the geometric accuracy of traditional bundle adjustment (BA) methods within seconds on consumer-grade GPUs.
Face Anything: 4D Face Reconstruction from Any Image Sequence: The paper proposes Face Anything, unifying 4D face reconstruction and tracking as canonical coordinate prediction—the network jointly estimates depth and canonical face coordinates from arbitrary image sequences, and dense correspondences are obtained directly through nearest-neighbor search in canonical space without explicitly modeling inter-frame motion. It outperforms existing methods across depth estimation (16% lower RMSE), correspondence accuracy (3× improvement), and inference efficiency (32× speedup).
G2P: Gaussian-to-Point Attribute Alignment for Boundary-Aware 3D Segmentation: The paper proposes G2P, which establishes precise Gaussian-to-point correspondences using Mahalanobis distance and transfers opacity and scale attributes from 3D Gaussian Splatting to point clouds. This introduces appearance cues while preserving the original point geometry, effectively reducing segmentation confusion between objects with similar geometry but different appearances.
Gaussian Belief Propagation Network for Depth Completion: GBPN reformulates depth completion as "performing Gaussian belief propagation on a Markov Random Field (MRF) dynamically constructed by a network." Instead of directly regressing depth, the network learns the MRF's potential functions, edge weights, and non-local edge structures, and then uses differentiable Gaussian BP to iteratively infer the dense depth distribution (treating the mean as depth and precision as confidence). It achieves SOTA performance on NYUv2 and KITTI, exhibiting robustness under extremely sparse inputs that far surpasses pure regression networks.
GaussianGPT: Towards Autoregressive 3D Gaussian Scene Generation: GaussianGPT compresses 3D Gaussian scenes into discrete token sequences and performs token-by-token autoregressive prediction using a causal Transformer with 3D rotary position embeddings. It realizes the pure autoregressive paradigm for 3D Gaussian scene generation for the first time, achieving or surpassing state-of-the-art (SOTA) diffusion models in shape synthesis and scene generation, and inherently supporting scene completion and outpainting.
GENA3D: Generative Amodal 3D Modeling by Bridging 2D Priors and 3D Coherence: GENA3D proposes a generative amodal 3D modeling framework. By parallelly fusing multi-view 2D completion features via View-Wise Cross Attention and injecting MVS point clouds as geometric gating into attention logits via Stereo-Conditioned Cross Attention, it simultaneously achieves generative diversity and geometric coherence under sparse occluded inputs. It reaches a FID of 30.73 on the GSO dataset (4-view), which is a 4.4 improvement over Amodal3R, while achieving an SSIM of 0.838 in visible regions.
GenSP: Consistent Spherical Parameterization via Learning Shape Generative Models: GenSP proposes a data-driven method to learn a neural generative model that continuously deforms a unit sphere into arbitrary genus-0 shapes, thereby obtaining cross-shape consistent and low-distortion spherical parameterization via inverse mapping. On ShapeNet, it improves isometric distortion by over 14× and consistency by over 3.6× compared to existing methods.
Geo-ID: Test-Time Geometric Consensus for Cross-View Consistent Intrinsics: Geo-ID proposes a training-free, test-time framework that couples the predictions of independently run pre-trained single-view intrinsic diffusion models using sparse 3D geometric correspondences. It aggregates these predictions into a confidence-weighted, voxel-level robust consensus, which is back-projected and injected as a guide into the subsequent diffusion sampling process via gradient updates on latent variables. This enforces consistency in albedo, roughness, and metallicity across views without degrading single-view decomposition quality, and enables direct use in downstream editable neural scene representations.
Geometry-Aware Style Transfer in 3D Gaussian Splatting: This paper proposes a geometry-aware 3DGS style transfer framework that alternately updates color and geometry parameters via decoupled optimization. It introduces a multimodal (RGB + depth + edge) contrastive feature matching loss (GCFM) to guide the transfer of style structures and geometric textures, significantly outperforming existing 3DGS style transfer methods in style fidelity and multi-view consistency.
GRAFT: Geometric Refinement and Fitting Transformer for Human Scene Reconstruction: GRAFT is a learned "human-scene interaction prior" that amortizes the optimization-based physical plausibility fitting process into a feed-forward Transformer. By encoding the relative human-scene relationship into 24 compact tokens via geometric probes, it predicts "interaction gradients" in a recurrent loop to correct floating and penetration. Ultimately, it achieves optimization-level interaction quality with a feed-forward speed of 0.2s and can serve as a plug-and-play prior to enhance any feed-forward method.
HAT-4D: Lifting Monocular Video for 4D Multi-Object Interactions via Human-Agent Collaboration: HAT-4D proposes the first agentic framework for monocular-video 4D multi-object interaction reconstruction. It uses an interaction knowledge graph (IKG) as a causal engine to drive 3D generation, spatial composition, and memory-augmented 4D propagation, while introducing multi-level human-in-the-loop (HITL) collaboration to address depth ambiguity and occlusion. On the MVOIK-4D benchmark, it comprehensively outperforms existing monocular 4D baselines, and a small amount of human feedback substantially improves interaction reconstruction quality.
Hierarchical 3D Scene Graph Construction and Belief-based Planning for Semantic Navigation: This paper proposes a zero-shot semantic navigation framework that online incrementally constructs an object-zone-region hierarchical 3D scene graph (HSG) as structured global memory. It fuses LLM semantic priors with exploration evidence to maintain hierarchical belief states, and uses POUCT within an HSG simulator for finite-step forward rollouts to explicitly evaluate the long-horizon expected returns of macro-actions, improving long-range navigation SR/SPL by 9.4%/5.0%, respectively.
Improving Sparse-View 3DGS Generalization via Flat Minima Optimization: This work addresses overfitting in sparse-view 3DGS from the perspective of flat minima optimization: through scale-adaptive position perturbation (SAP) + random application + magnitude scheduling + periodic reinitialization of non-position parameters, it drives 3DGS to converge to flatter regions of the loss landscape without changing the network architecture, achieving SOTA or near-SOTA rendering quality under sparse-view settings on LLFF and Mip-NeRF360.
Interact3D: Compositional 3D Generation of Interactive Objects: Interact3D proposes a training-free “generate-then-compose” framework. It leverages the spatial priors of 2D/3D generative models to formulate 3D compositional generation as a structured registration problem. Through a two-stage composition pipeline—global-to-local geometric alignment to anchor the main object, followed by SDF collision-aware optimization to place the remaining objects—and a VLM-driven agent closed-loop for iterative correction, it automatically generates physically plausible and collision-free interactive 3D compositional scenes from a single image and text prompt, while releasing a dataset of 8,000+ interactive 3D pairs.
JanusMesh: Fast and Zero-Shot 3D Visual Illusion Generation via Cross-Space Denoising: JanusMesh proposes a zero-shot, two-stage framework that utilizes dual-branch denoising and SDF fusion in the structured 3D latent space of TRELLIS to generate multi-semantic geometries, followed by viewpoint-constrained texture synthesis to assign corresponding appearances to each view, generating high-quality 3D visual illusion meshes within 3–5 minutes.
Large-Scale High-Quality 3D Gaussian Head Reconstruction from Multi-View Captures: HeadsUp proposes a scalable feed-forward method that compresses multi-view images from dynamic camera rigs into compact 2D latent variables via a cross-attention Transformer. These latents are then decoded into UV-parameterized 3D Gaussians anchored to a neutral head template. This decouples the output Gaussian count from the input image resolution and view count, achieving SOTA reconstruction quality on an internal dataset of 10,000+ subjects without requiring test-time optimization.
Learning Video Dynamics with Predictive Differentiable Rendering: This paper proposes the Predictive Differentiable Rendering (PDR) framework. By appending a lightweight PredGS adapter to existing pixel-space predictors, it maps coarse predictions to 2D Gaussian parameters. A CUDA-accelerated differentiable renderer then generates continuous-space predictions, which are fused with pixel predictions. This significantly alleviates the over-smoothing issue in deterministic video prediction with minimal computational overhead.
LUNA: Learning Universal 3D Human Animation Beyond Skinning: LUNA proposes the first end-to-end 3D human animation framework that completely abandons Linear Blend Skinning (LBS). It directly maps 2D driving signals (RGB images, keypoints, hand-drawn sketches, or even unseen character images) to 3D Gaussian deformations via a decoupled Transformer animator. During training, LBS distillation is utilized to provide structural priors to prevent geometric collapse. During inference, it completely bypasses LBS constraints, achieving zero-shot cross-identity and cross-modal 3D human animation.
MeGAS: Thermomechanical Dynamic Gaussian Splatting for Thermophysical Scene Editing: MeGAS integrates thermodynamic phase-change dynamics (heat conduction, convection, melting/solidification) into the 3D Gaussian Splatting framework, enabling the reconstructed real-world scenes to undergo controllable, physics-based thermomechanical deformation, while maintaining rasterization quality under extreme large deformations through a topology-adaptive rendering strategy.
Monocular Avatar Reconstruction via Cascaded Diffusion Priors and UV-Space Differentiable Shading: Using a unified pretrained diffusion backbone to pipeline three sub-tasks: "texture completion \(\rightarrow\) illumination homogenization \(\rightarrow\) material decomposition", with each sub-task attaching only a single LoRA, combined with a UV-space differentiable BRDF rendering loss, high-fidelity and relightable 3D facial PBR avatars at 4K resolution can be reconstructed from a single in-the-wild image using fewer than 100 real 3D scans.
Monte Carlo Energy Aggregation for Mobile 3D Gaussian Splatting: Flux-GS compresses 3rd-order Spherical Harmonics (SH) into a 1st-order representation via Monte Carlo specular energy aggregation. Complemented by an attribute-conditioned enhancement module with zero inference overhead and a multi-view alpha-weighted densification/pruning strategy, it achieves real-time, high-fidelity novel view synthesis at 130+ FPS on a mobile Snapdragon 8 Gen 3. It also compresses storage to 2-5 MB and reduces training time from Mobile-GS's 80+ minutes to 11 minutes.
Multi4D: High-Fidelity Dynamic Gaussian Splatting via Multi-Level Competitive Allocation: Multi4D explicitly decomposes a dynamic scene into three subsets of Gaussians: "static structure + persistent dynamic geometry + transient appearance", allowing them to competitively explain photometric residuals under shared rasterization. This concurrently preserves long-term motion consistency and high-frequency appearance details. Relying on 25\(\times\) fewer dynamic Gaussians than 4DGS, it achieves SOTA rendering quality and real-time frame rates, while casually enabling 4D segmentation with a 10\(\times\) speedup.
One Video, One World: Turning Monocular Video into Physical 4D Scenes: OVOW is the first training-free system that transforms monocular videos into instance-level, physically simulatable 4D mesh scenes. Through a four-stage pipeline consisting of VLM-guided scene decomposition, category-aware mesh reconstruction, iterative scale-pose recovery, and physically grounded assembly, it outputs watertight mesh instances without any fine-tuning, which can be directly imported into physical simulators like MuJoCo.
OP3DSG: Open-Vocabulary Part-Aware 3D Scene Graph Generation for Real-World Environments: OP3DSG proposes a unified 3D scene graph generation framework. By leveraging knowledge-guided part detection, part-aware multi-view fusion, and a geometrically initialized prior graph combined with verification-gated LLM reasoning, it jointly models objects, interactive parts, spatial relations, functional relations, and affordances. On the self-constructed UniGraph3D benchmark, its part-node recall outperforms prior methods by an absolute margin of over 31 points, and it can directly serve as a perception backend for robotic QA, navigation, and task planning.
OrthoTrack: Continuous 6-DoF UAV Trajectory Estimation Anchored in Public Orthophotos: OrthoTrack proposes a training-free, 6-DoF continuous UAV trajectory estimation system. By leveraging public orthophotos (DOP) and Digital Surface Models (DSM) using an alternating strategy of "keyframe dense matching anchoring + inter-frame optical flow propagation", it achieves sub-meter absolute metric pose estimation at real-time frame rates without requiring GPS, IMUs, or post-processing alignment. It also introduces MovingDrone, a large-scale multimodal benchmark.
Pano3D: Unified 3D Reconstruction and Panoptic Segmentation: Pano3D directly appends a Mask2Former-style set-based mask decoder to a Feedforward Reconstruction Model (FRM, e.g., MUSt3R / Pi3), and jointly fine-tunes the geometric decoder using both geometric and semantic losses. This represents the first approach to simultaneously output dense point clouds and 3D panoptic segmentation in a single feedforward pass without requiring external 2D models or post-processing clustering. It significantly outperforms state-of-the-art (SOTA) methods in semantic segmentation mIoU on ScanNet (+16.6), ScanNet200, and ScanNet++.
Physically Grounded 3D Generative Reconstruction under Hand Occlusion using Proprioception and Multi-Contact Touch: This work integrates monocular RGB vision, proprioception (hand pose), and multi-contact touch within a flow matching diffusion framework to generate the metric-scale SDF of an object under severe hand occlusion. It guides shape reconstruction using two physical losses—non-penetration and contact consistency—during both training and inference sampling stages, yielding physically feasible outputs and implicitly estimating the object pose.
PoseShield: Neural Collision Fields for Human Self-Collision Resolution: PoseShield learns a neural collision field with Eikonal regularization (approximating pose-space SDF) within the SMPL pose space, embedding it as a differentiable constraint function in gradient-based constrained optimizers like SLSQP. This achieves post-processing correction from self-penetrating poses to the nearest collision-free poses. This method scales to motion sequence collision resolution without retraining downstream motion generation models, achieving a 95.8% success rate on the self-built HwC dataset, significantly outperforming COAP (44.6%).
Progressive Pose-Guided 4D Animal Reconstruction from Monocular Video: This paper proposes a progressive test-time optimization framework based on 3D Gaussian Splatting. By decoupling skeletal motion from non-rigid deformation, it reconstructs high-fidelity 4D animal models from monocular videos, requiring only coarse-grained shape priors to generalize to diverse species.
RAGA: Real Time Ray Traced Gaussian Shadow Casting for 3DGS Avatar-Scene Interaction: RAGA proposes a ray tracing method that computes shadows entirely in the Gaussian space. For each shadow ray, it calculates a precise Gaussian line integral and normalizes it to obtain a thickness factor describing "how much" volume of the occluder is traversed, which drives the shadow transmittance. Simultaneously, it introduces an avatar proxy to eliminate temporal jitter during animation, achieving interactive shadow casting at approximately 50 FPS on NVIDIA OptiX.
RePer-360: Releasing Perspective Priors for 360\(^\circ\) Depth Estimation via Self-Modulation: RePer-360 proposes a self-conditioned framework based on normalization modulation. Without overwriting pre-trained perspective priors, it utilizes complementary projection (CP and ERP) features as guidance signals for geometric alignment to modulate a depth foundation model, achieving state-of-the-art (SOTA) omnidirectional depth estimation with an extremely small amount of panoramic data (approx. 1%).
SemCityLoc: Aerial 6DoF Localization Using Semantic 3D City Models: SemCityLoc redefines UAV aerial 6DoF pose estimation as a structured surface registration problem between foundation model visual priors (semantic segmentation + monocular depth) and standardized semantic 3D city models. Through a two-stage pipeline consisting of "4D semantic cost volume coarse localization \(\rightarrow\) particle filter-based joint semantic-depth fine alignment", without requiring dense radiometric reconstruction, it improves the localization recall in urban canyon scenarios from 35.11% to 69.15% (under the 2m-2° threshold) and reduces the positional error from 9.89m to 2.62m. Concurrently, it releases SemCityLockeD, the first real UAV localization benchmark with centimeter-level precision.
ShellMaker: Language-Guided Exterior Completion under Structural Constraints: ShellMaker proposes a language-guided building exterior completion framework: given a building scaffold (wall + door/window opening layouts output by indoor generators/CityGML/CAD) and a style prompt, it produces a complete PBR exterior facade mesh through four modules: parametric roof generation, LLM part-aware prompt refinement, compatibility-aware joint wall-roof texture retrieval, and geometry-aware assembly. While strictly preserving footprints and opening layouts, it matches any architectural style, achieving a footprint IoU of 0.992 and an opening center error of only 0.091.
SK-Adapter: Skeleton-Based Structural Control for Native 3D Generation: SK-Adapter proposes 3D skeletons as a first-class control signal, injecting them into a frozen 3D flow matching generative backbone (Trellis) via a lightweight adapter network (GRPE topological encoding + skeletal cross-attention). This approach achieves precise structural control in native 3D space while preserving pre-trained generative priors, and is accompanied by the construction of Objaverse-TMS, a dataset containing 24k text-mesh-skeleton triplets.
SKEL-CF: Coarse-to-Fine Biomechanical Skeleton and Surface Mesh Recovery: SKEL-CF proposes a coarse-to-fine framework to recover biomechanical skeletons and surface meshes from a single RGB image. Under this framework, the model is first trained on the high-quality HMR-SKEL dataset, followed by explicit camera intrinsic modeling to eliminate depth ambiguity. Finally, through an encoder coarse prediction + decoder layer-wise iterative refinement strategy, it significantly outperforms the previous SKEL method HSMR on challenging datasets such as MOYO (reducing MPJPE by 18.6%), while achieving comparable or even superior numerical accuracy to SMPL-based methods, and outputting anatomically more plausible poses.
SPECSIA: Stylization Dataset for Novel-View Enhancement in Drawing-based 3D Animation: This paper proposes the SPECSIA-15K paired stylization dataset (14,980 pairs of artifact-laden projections and clean targets) and a lightweight plug-and-play post-correction module DraViE. By replacing instance-level frame-by-frame optimization with a data-level pre-training prior, it effectively eliminates projection artifacts under novel views in drawing-driven 3D animation while preserving the original drawing style.
SpectralSplats: Robust Differentiable Tracking via Spectral Moment Supervision: This work shifts the supervision target of 3DGS differentiable tracking from the spatial domain (pixel-wise photometric error) to the frequency domain (spectral moments). Combined with a frequency annealing schedule strictly derived from the principle of phase wrapping, the method enables Gaussian assets to obtain non-zero directional gradients even when completely non-overlapping with the target at initialization, smoothly "flowing" to the correct pose—thereby enabling robust tracking without manual alignment or category priors.
StereoGS: Sparse-View 3D Gaussian Splatting via Stereo Priors: StereoGS addresses the inherent scale ambiguity and cross-view inconsistency in monocular depth priors of sparse-view 3DGS by constructing virtual stereo camera pairs, introducing stereo depth regularization, and applying gradient-aware opacity decay, achieving state-of-the-art novel view synthesis on four datasets: LLFF, DTU, Mip-NeRF360, and Blender.
TORA: Topological Representation Alignment for 3D Shape Assembly: During the training of a flow-matching-based 3D assembly model, TORA additionally aligns its feature representation with the "who-is-similar-to-whom" relational structure of a frozen, pre-trained 3D encoder (matching Gram matrices via CKA loss). This distills geometric interaction priors into the network, accelerating convergence by up to \(6.9\times\), significantly enhancing cross-domain robustness, and incurring zero extra cost during inference.
Towards in-the-wild Egocentric 3D Hand-Object Pose Estimation: This work constructs the first in-the-wild egocentric hand-object contact dataset, EPIC-Contact, using a low-cost annotation pipeline of "hand contact area \(\to\) bijective transfer to object \(\to\) optimization fitting." It also proposes HOPformer, an end-to-end network that injects pre-trained hand priors into object features via cross-attention to jointly regress the 3D poses of both hands and the object in a single forward pass. This improves the success rate from 76.2% to 82.4% on ARCTIC, while nearly doubling the success rate and reducing the contact deviation by 75% on in-the-wild data.
UniPR-3D: Towards Universal Visual Place Recognition with Visual Geometry Grounded Transformer: UniPR-3D introduces the 3D geometry-aware tokens of VGGT into visual place recognition for the first time. It designs tailored aggregation strategies for the distinct characteristics of 2D and 3D tokens and supports both single-frame and variable-length sequence matching, outperforming existing single-view and multi-view methods on multiple benchmarks.
UniTriSplat: A Unified 3D Gaussian Splatting Framework with Uniform Spherical Rasterization for Universal Cameras: This work shifts the rasterization of 3D Gaussian Splatting from "planar projection for a specific camera class" to the "HEALPix equal-area grid on the unit sphere". By utilizing a single camera-agnostic spherical rasterizer to universally support perspective, fisheye, and panoramic cameras, it simultaneously eliminates polar oversampling in equirectangular projections and achieves stable cross-camera generalization.
URoPE: Universal Relative Position Embedding across Geometric Spaces: URoPE lifts key image patches into 3D points along camera rays at fixed depth anchors, then reprojects them onto the query view image plane. This enables standard 2D RoPE to directly encode relative positional relationships across views or geometric spaces. It serves as a parameter-free, intrinsic-aware, coordinate-system-agnostic, and FlashAttention-compatible universal position embedding.
ViewSplat: View-Adaptive 3D Gaussian Splatting for Feed-Forward Synthesis: ViewSplat introduces a "view-adaptation" step to pose-free feed-forward 3DGS. It first predicts a set of base Gaussians and scene-conditioned View MLPs. During rendering, these MLPs take target view coordinates to dynamically calculate residual offsets for all Gaussian attributes (position, scale, rotation, opacity, color) on the fly. While maintaining real-time rendering (the largest backbone still achieves 90 FPS), this approach reconstructs high-frequency view-dependent effects such as specular highlights and sharp reflections that are typically lost in static representations, achieving a new state-of-the-art in pose-free NVS fidelity.
Vitality-Aware Compression for Efficient Image-to-Shape Diffusion Transformers: For image-to-3D Diffusion Transformers (DiTs), this paper introduces a "layer vitality" metric based on Earth Mover's Distance to measure each layer's contribution to geometric synthesis. Based on this, a pipeline of structured pruning, layered mixed-precision quantization, and targeted distillation that fine-tunes only the lowest-vitality layers is developed. This compresses backbones by up to 66% on state-of-the-art models like Step1X-3D and Hunyuan3D while maintaining geometric fidelity.
VOCA: Visual Odometry with Codec Awareness: VOCA leverages the built-in motion vectors from the H.264 video encoding process as an initialization prior for KLT optical flow tracking. Built upon the Basalt framework, it achieves high-accuracy causal visual odometry on videos compressed by up to 100x, comprehensively outperforming existing methods on the EuRoC, TUM-VI, and MSD datasets.
VolSplat: Rethinking Feed-Forward 3D Gaussian Splatting with Voxel-Aligned Prediction: VolSplat shifts the primitive prediction paradigm of feed-forward 3D Gaussian Splatting from pixel-aligned to voxel-aligned: it first projects multi-view 2D features back to 3D space to form a voxel feature grid through depth maps, refines them using residual learning via a sparse 3D U-Net, and then predicts Gaussian parameters per voxel, thus fundamentally decoupling the 3D representation from the rigid binding of the input image resolution to achieve adaptive Gaussian density allocation.
Walking in the Implicit: Interactive World Exploration via Neural Scene Representation: NeuWorld replaces the "rollout state" of interactive world exploration from an ever-growing sequence of video frame latent variables with a set of fixed-length, renderable, and locally anchored neural implicit scene (NIS) tokens. This allows the diffusion model to transition only within this compact state space, while a frozen decoder renders observations based on queried camera poses. Consequently, without leveraging any pre-trained video foundation models and training entirely from scratch, NeuWorld achieves excellent long-term pose consistency, loop-revisit consistency, and inference efficiency.
Zero-Shot Depth from Defocus: Zero-Shot Metric Depth Estimation from Focus Stacks: This paper proposes FOSSA, a Transformer-based zero-shot depth-from-defocus network that efficiently aggregates defocus cues from a focus stack via a novel stack attention layer. Trained on large-scale synthetic data, it achieves strong zero-shot generalization across datasets and scenes. Additionally, this work releases ZEDD, a high-quality real-world depth-from-defocus benchmark containing 100 scenes, 4K resolution, and LiDAR ground truth.