Tagged articles
3D reconstruction
46 articles · Page 1 of 1
This article analyzes eight ICLR 2027 papers that achieved unanimous 4/4 reviewer scores, detailing technical innovations from self-managing context language models and hyperbolic reinforcement learning to protein structure fusion and verified proof compression, and reveals how rebuttal responses turned initial rejections into top-tier acceptances.
3D reconstructionICLR 2027amorphous materials
0 likes · 18 min read


Amap TechNov 5, 2026 · Artificial Intelligence
Amap unveiled Phys AI Data at the 2026 World Internet Conference, a three-layer spatial data foundation comprising training data (embodied, spatial, synthetic), a simulation evaluation platform, and Physical AI Map, enabling robot training, validation, and deployment in real-world scenarios like malls and factories.
3D reconstructionAmapEmbodied AI
0 likes · 8 min read

The open-source project 3D Model Anything lets users scan a room with an iPhone's LiDAR, then uses an iterative AI workflow to reconstruct it as an editable Three.js scene that runs in the browser, with all intermediate data openly shared for verification and further development.
3D reconstructionAI modelingLiDAR scanning
0 likes · 12 min read


Amap TechOct 12, 2026 · Artificial Intelligence
Amap had 16 papers accepted at NeurIPS 2026, covering multimodal large models, spatial intelligence, visual reasoning, and reinforcement learning, with highlights including a decomposed verifier for generative world modeling, gradient projection for continual multimodal learning, map-free transit route generation, and benchmarks for coding agents and travel planning.
3D reconstructionEmbodied AIMultimodal Large Models
0 likes · 30 min read

HSImul3R introduces a physics-in-the-loop framework that reconstructs simulation-ready human-scene interactions from sparse views, using scene-targeted reinforcement learning and direct simulation reward optimization to achieve stable physical interactions, validated on HSIBench and deployed on Unitree G1 robot.
3D reconstructionECCV 2026HSIBench
0 likes · 11 min read

PolaFormer++ introduces a polarity-aware linear attention mechanism with channel-wise spiky feature mapping (PaCS) that solves negative value loss and high entropy issues, achieving state-of-the-art results across six vision tasks including classification, detection, segmentation, diffusion, 3D synthesis, and super-resolution.
3D reconstructionADE20KCOCO
0 likes · 16 min read

Atlas, a multimodal autoregressive diffusion Transformer introduced by Fei‑Fei Li's World Labs, can generate pixel‑level camera‑controlled images and videos, perform high‑quality 3D reconstruction from a few photos, simulate space‑time for robotics, and outperforms state‑of‑the‑art models in quantitative evaluations.
3D reconstructionAtlasRobotics
0 likes · 8 min read


21CTOSep 2, 2026 · Artificial Intelligence
World Labs, founded by Fei‑Fei Li and colleagues, launched Atlas—a multimodal autoregressive diffusion transformer that natively handles text, images, video and 3D, offering camera‑control generation, spatial reconstruction, spatiotemporal simulation and image synthesis, and achieving leading scores on human preference and 3D reconstruction benchmarks.
3D reconstructionAIAtlas
0 likes · 6 min read

World Labs' Atlas is a multimodal, camera‑controlled world model that natively handles text, images, video and 3D, offering spatial‑context generation, high‑fidelity 3D reconstruction, and robot simulation, with benchmark results that highlight its advantages over prior models.
3D reconstructionAtlasRobotics
0 likes · 10 min read

Stream3D, a collaborative effort from Harvard, MIT and HKUST, introduces a training‑free, memory‑efficient mechanism that continuously extracts reliable multi‑view evidence from video streams to drive frozen view‑conditioned 3D generators, achieving consistent, complete 3D reconstructions superior to both pure reconstruction and generation baselines.
3D reconstructionEvidence memoryMulti-view learning
0 likes · 14 min read

The article analyzes Real2Sim's approach of constructing simulation environments directly from massive, millimeter‑accurate 3D scans, highlighting its zero‑error reconstruction, multi‑modal data richness, scalable scene generation, and how it surpasses video‑driven methods like SimFoundry for embodied AI training.
3D reconstructionDigital TwinEmbodied AI
0 likes · 8 min read

The Oxygen XR team and Tsinghua University propose DirectFisheye‑GS, a framework that embeds a fisheye camera model directly into the 3D Gaussian splatting pipeline and introduces cross‑view joint optimization, eliminating distortion and view‑inconsistency while achieving SOTA performance on multiple public datasets with efficient rendering and reconstruction.
3D reconstructionComputer VisionCross-View Optimization
0 likes · 11 min read

The DirectFisheye‑GS framework embeds a fisheye camera model directly into the 3D Gaussian splatting pipeline and introduces cross‑view joint optimization, eliminating shape distortion and inter‑view inconsistencies, achieving SOTA performance on multiple public datasets while remaining compatible with existing 3DGS toolchains.
3D reconstructionCVPR 2026Cross-View Optimization
0 likes · 10 min read


AntTechJun 23, 2026 · Artificial Intelligence
This article reviews three CVPR 2026 papers—AVGGT, MVP, and Online3R—detailing how re‑engineered global attention, multi‑view prediction, and online self‑supervised learning each dramatically improve efficiency, stability, or consistency of visual tasks such as multi‑view 3D reconstruction and GUI grounding.
3D reconstructionCVPR 2026GUI grounding
0 likes · 8 min read

CameraSquad introduces a parallel multi‑trajectory video generation framework that delivers precise camera control and cross‑view content consistency, enabling high‑quality 3D point‑cloud reconstruction and superior performance on benchmarks such as WebVid and HumanVid compared with prior camera‑controlled video methods.
3D reconstructionCameraSquadcamera-controlled video
0 likes · 14 min read

Holi‑Spatial introduces a fully automatic pipeline that transforms raw video streams into high‑quality 3D geometry, depth, masks, 3D boxes, instance descriptions, grounding and spatial QA, producing the 4‑million‑item Holi‑Spatial‑4M dataset and substantially improving VLM spatial reasoning performance.
3D reconstructionICML 2026Multimodal Dataset
0 likes · 14 min read

The article reviews three major recent breakthroughs in computer vision—self‑supervised visual foundation models, feed‑forward 3D reconstruction, and unified multimodal models—detailing their underlying methods, key papers, performance characteristics, and practical implications for real‑world AI applications.
3D reconstructionComputer VisionMultimodal Models
0 likes · 22 min read

Xiaomi Auto introduces a novel world‑model framework that tightly couples 3D reconstruction and generative prediction, delivering state‑of‑the‑art performance on Waymo and nuScenes benchmarks while enabling high‑fidelity, long‑duration video synthesis for autonomous‑driving scenarios.
3D reconstructionAutonomous DrivingBenchmark SOTA
0 likes · 10 min read

The VGGT series from KOKONI 3D and collaborators tackles three core 3D perception limits—unbounded sequence memory, dynamic‑static entanglement, and compute‑precision trade‑offs—by introducing StreamCacheVGGT, progressive decoupling, and HD‑VGGT, achieving O(1) memory streaming, 15%+ accuracy gains on dynamic benchmarks, and record‑high AUC on RealEstate10K.
3D reconstructionComputer VisionVGGT
0 likes · 10 min read

Scal3R introduces test‑time training with a global‑context memory and synchronization mechanism that lets models train on and infer over ultra‑long video sequences, achieving accurate camera poses and dense point clouds for kilometer‑scale scenes while outperforming prior SLAM, SfM and streaming baselines on multiple benchmarks.
3D reconstructionComputer VisionScal3R
0 likes · 11 min read

The article examines how 3D reconstruction is evolving from offline, single‑scene pipelines to continuous, streaming workflows that feed web distribution, robot simulation, visual positioning, spatial editing, and world‑generation systems, highlighting recent research, standards, and industry deployments.
3D reconstructionDigital TwinRobotics
0 likes · 10 min read

OmniRoam introduces a panoramic, coarse‑to‑fine framework that generates long, trajectory‑controlled videos with higher spatial consistency and temporal coherence, offering a stable and controllable alternative to short‑clip generation and supporting real‑time preview, high‑resolution refinement, and 3D reconstruction applications.
3D reconstructionGenerative AIOmniRoam
0 likes · 8 min read

HSImul3R introduces a physics‑in‑the‑loop reconstruction pipeline that closes the perception‑simulation gap by jointly optimizing human motion and scene geometry, leveraging reinforcement learning, direct simulation‑reward optimization, and a new HSIBench dataset to produce simulation‑ready 3D human‑scene interactions.
3D reconstructionDSROHSIBench
0 likes · 12 min read

The article presents LoGeR, a long‑context geometric reconstruction framework that combines test‑time‑training memory and sliding‑window attention to achieve minute‑scale, fully‑feedforward 3D reconstruction with superior accuracy on benchmarks such as KITTI and VBR.
3D reconstructionComputer VisionHybrid Memory
0 likes · 11 min read

MIT researchers introduce Wave‑Former, a physics‑aware, generative‑AI framework for mmWave sensing that achieves high‑precision 3D reconstruction of completely hidden objects, raising recall from 54% to 72% while maintaining 85% precision and outperforming existing baselines on real‑world datasets.
3D reconstructionGenerative AIbenchmark
0 likes · 15 min read

LoGeR, a new long‑context geometric reconstruction framework from DeepMind and UC Berkeley, uses a hybrid memory module combining test‑time‑training (TTT) and sliding‑window attention (SWA) to enable feed‑forward 3D reconstruction over sequences of up to tens of thousands of frames, achieving state‑of‑the‑art accuracy on KITTI, VBR, 7‑Scenes, ScanNetV2 and TUM‑Dynamics benchmarks.
3D reconstructionHybrid MemoryLoGeR
0 likes · 11 min read

This thesis investigates the two core challenges of data prior utilization and scene parameterization in multi‑view RGB‑based 3D indoor reconstruction, proposing novel representations and learning‑based methods to improve reconstruction quality, generalization, and applicability across AR, robotics, and autonomous navigation.
3D reconstructionComputer Visiondata priors
0 likes · 8 min read

This roundup surveys five cutting‑edge AI papers—DeepSeekMath‑V2’s self‑verifiable mathematical reasoning, MedSAM‑3’s promptable medical image and video segmentation, SAM 3D’s single‑image 3D reconstruction, Qwen3‑VL’s high‑capacity vision‑language model, and the M² memory‑mesh transformer for image captioning—highlighting their key methods, benchmarks, and code links.
3D reconstructionImage CaptioningMathematical Reasoning
0 likes · 6 min read

Meta’s newly released SAM 3 and SAM 3D models enable single‑image 3D reconstruction and promptable segmentation, outperforming prior methods on benchmarks, introducing a shared perception encoder, a Presence Head to reduce hallucinations, and a two‑stage generation pipeline that produces high‑fidelity geometry and texture.
3D reconstructionMetaSAM 3
0 likes · 12 min read

The Volcano Engine Multimedia Lab showcases cutting‑edge AI‑driven 3D and VR technologies—including volume video, dual‑Gaussian modeling, topological‑aware representations, and the Beaver3D AIGC model—to lower creation barriers, enable real‑time immersive interaction, and bridge research breakthroughs with industry applications.
3D reconstructionAI-generated 3DAIGC3D
0 likes · 15 min read

This article explains how 3D Gaussian Splatting transforms ordinary video footage into high‑quality 3D reconstructions with minimal equipment, outlines the low‑cost workflow using ffmpeg and COLMAP, and discusses practical challenges and future possibilities for the technology.
3D reconstructionCOLMAPComputer Vision
0 likes · 5 min read

SIGGRAPH 2025 announced record‑breaking submissions and awarded five best papers, several honorable mentions, and a Test‑of‑Time prize, highlighting breakthroughs in 3D reconstruction, neural fields, Monte‑Carlo rendering, cloth simulation, and IMU calibration, with detailed author, institution, and technical insights provided.
3D reconstructionAIMonte Carlo rendering
0 likes · 13 min read

The CVPR 2025 awards spotlight groundbreaking work—from the VGGT transformer that predicts full 3D scenes in a single feed‑forward pass to neural inverse rendering that reconstructs geometry from time‑resolved light—offering a comprehensive view of emerging trends, novel architectures, and performance breakthroughs across computer‑vision research.
3D reconstructionCVPR 2025Neural Rendering
0 likes · 11 min read

DP-Recon leverages generative diffusion priors and a visibility‑guided SDS loss to achieve high‑fidelity, compositional 3D scene reconstruction from extremely sparse images, delivering superior geometry, texture, and text‑driven editing capabilities demonstrated on benchmark datasets and real‑world indoor scenarios.
3D reconstructionAIdiffusion models
0 likes · 10 min read

This article explains how to create high‑quality 3D scenes from ordinary video footage by slicing frames with ffmpeg, extracting camera poses with COLMAP, and applying 3D Gaussian Splatting to replace traditional mesh‑texture pipelines, dramatically lowering equipment costs and data size.
3D reconstructionCOLMAPComputer Vision
0 likes · 6 min read

This article introduces Neural Radiance Fields (NeRF) as a cutting‑edge AI technique for high‑quality 3D reconstruction, explains its core principles and advantages, outlines a step‑by‑step building workflow, reviews popular open‑source libraries such as Luma AI, NVIDIA Instant NeRF and NeRFStudio, and offers a forward‑looking summary of its potential and challenges.
3D reconstructionAIComputer Vision
0 likes · 12 min read

Professor Hu Qiang presented recent progress in Neural Radiance Fields—covering implicit/explicit representations, hybrid models, and solutions for dynamic scenes, cloud‑based and edge‑cloud rendering—while also reviewing generative AI advances such as diffusion‑based text‑to‑image/video/3D, LoRA fine‑tuning, and large‑scale story‑book datasets, highlighting applications in virtual‑real content, smart‑city modeling, and 6‑DoF e‑commerce displays.
3D reconstructionGenerative AINeRF
0 likes · 14 min read

This article introduces the PanoViT method for indoor layout estimation from panoramic images, covering research background, the transformer‑based architecture with backbone, vision transformer encoder, boundary‑enhancement and 3D loss modules, experimental results, and step‑by‑step usage in ModelScope.
3D reconstructionindoor layout estimationpanoramic vision transformer
0 likes · 7 min read

This article introduces the PanoViT model, a vision‑transformer‑based approach for indoor layout estimation from panoramic images, covering its research background, architectural components, experimental results on public datasets, and step‑by‑step usage within ModelScope.
3D reconstructionComputer VisionModelScope
0 likes · 8 min read

Alibaba’s Taobao engineers detail how they transformed slow, high‑quality NeRF reconstruction into a real‑time mobile solution by combining an Octree‑Tiny‑MLP architecture, SNeRG optimizations, and a high‑frequency voxel reduction that shrank models to ~5 MB and achieved ~6 FPS on low‑end Android phones, targeting sub‑1 MB models and 50 FPS.
3D reconstructionNeRFmobile optimization
0 likes · 10 min read

NeRF‑Editing introduces an interactive framework that lets users freely deform the geometry of neural radiance fields by coupling an explicit mesh with implicit NeRF representations, propagating mesh vertex changes through tetrahedral ARAP optimization to bend rays during rendering, enabling realistic edits and animations on synthetic and real‑world scenes, a first reported at CVPR 2022.
3D reconstructionARAP deformationComputer Vision
0 likes · 6 min read

This article explains the end‑to‑end workflow of free‑viewpoint video technology—from multi‑camera on‑site capture and hardware setup, through cloud‑based 3D reconstruction, depth estimation and encoding, to mobile SDK rendering—highlighting the technical challenges and optimizations that enable real‑time immersive streaming.
3D reconstructioncloud renderingfree-viewpoint
0 likes · 10 min read

SnowflakeNet introduces a novel Snowflake Point Deconvolution architecture combined with a Skip-Transformer to progressively split seed points, enabling high‑quality point‑cloud completion that preserves fine‑grained geometric details such as smooth surfaces, sharp edges, and corners across dense and sparse datasets.
3D reconstructionComputer VisionSnowflakeNet
0 likes · 10 min read

Camera-space hand mesh recovery (CMR) leverages semantic aggregation of 2D cues and adaptive 2D‑1D registration to predict absolute 3D hand pose and shape directly in camera coordinates, improving accuracy on benchmarks such as FreiHAND, RHD, and Human3.6M.
2D-1D registration3D reconstructionComputer Vision
0 likes · 17 min read

This article explores how 3D digital storebuilding integrates multiple visual sensors, GPU acceleration, and advanced camera calibration to create high‑precision, real‑time digital twins of retail spaces, enabling fine‑grained lifecycle management and immersive customer experiences.
3D reconstructionGPU accelerationcamera calibration
0 likes · 15 min read


Didi TechSep 10, 2020 · Artificial Intelligence
DiDi's AR indoor navigation system addresses GPS unreliability in large indoor venues by using SfM-based 3D reconstruction, robust visual localization with magnetometer/GNSS priors, and sensor fusion with pedestrian dead‑reckoning and deep‑learning heading estimation, cutting passenger pick‑up time by up to 25 % across dozens of airports and malls.
3D reconstructionAR navigationIndoor positioning
0 likes · 19 min read
