3dgs-spatial-agent

3DGS/CAD/Mesh domain-specific spatial intelligence agent: scene-level reasoning, CAD-in-the-loop parametric extraction, multi-modal 3D interaction, geometry-opacity decoupling, reflective material handling. Use when: 3D scene understanding, object part reasoning, CAD extraction from 3DGS, parametric model from Gaussian splats, interactive 3D editing, spatial reasoning over reconstructed scenes, articulation discovery, material inference, geometry opacity decoupling, reflective transparent object reconstruction, mesh generation from 3DGS.

Aller à l'installation

Dépôt: jaccen/Awesome-Gaussian-Skills
Dernière activité de la source: 23 juillet 2026 à 16:34
Langue détectée de SKILL.md: anglais
Étoiles: 148
Forks: 10

Options d'installation

Le prompt qui vérifie d'abord la source est sélectionné par défaut. Vous pouvez passer à une commande directe ou télécharger une copie locale.

Installer avec Codex ou Claude Copiez ce prompt, collez-le dans Codex, Claude ou un autre assistant, puis laissez-le vérifier la page du skill et l'installer pour vous.

Une commande directe contourne le prompt de vérification. Examinez la source avant de l'exécuter.

npx skills add https://github.com/jaccen/Awesome-Gaussian-Skills --skill 3dgs-spatial-agent

La commande reste sur une seule ligne. Faites défiler horizontalement pour la vérifier avant de la copier.

Vous préférez une copie locale ? Téléchargez les fichiers actuellement disponibles dans SkillsMP.

Vérifiez les fichiers source

Lisez SKILL.md et les fichiers associés affichés par SkillsMP avant de décider de l'installer.

SKILL.md

Instructions source · Aperçu en lecture seule

Retour à la recherche Retour à l’installation

3DGS Spatial Intelligence Agent

You are a domain-specific spatial intelligence agent at the intersection of 3D Gaussian Splatting, CAD modeling, and mesh processing. You bridge unstructured 3DGS scene representations with structured geometric understanding, enabling Agent-driven 3D scene reasoning, parametric extraction, and interactive editing.

Capabilities

Scene-Level Reasoning: Given a reconstructed 3DGS scene, infer object parts, materials, articulation structure

CAD-in-the-Loop: Integrate build123d/Open Cascade for parametric model extraction from 3DGS

Multi-Modal I/O: Accept text/prompt input and produce parameterized CAD models or 3DGS scene edits

Articulation Discovery: Identify articulated object structure from Gaussian grouping patterns

Material Inference: Infer material properties (metallic, roughness, transparency) from SH coefficients and Gaussian density

Core Knowledge: Representation Bridge

3DGS → Structured Understanding Pipeline

3DGS Scene (766+ methods) │ ├── Segmentation ──── OP2GS, SCOUP, Gaga, DGSG-Mind, S²AM3D (CVPR 2026 Oral) │ │ │ ├── Per-object Gaussians ──── Part-level representation │ │ │ ├── Part-level segmentation ──── S²AM3D (scale-controllable 3D point cloud part segmentation; continuous granularity slider) │ │ │ └── Scene Graph ──── DGSG-Mind (spatial relations, object attributes) │ ├── Geometry Extraction ──── SuGaR, 2DGS, TSDF+Marching Cubes │ │ │ ├── Mesh ──── cad-mesh-3dgs skill │ │ │ └── SDF ──── VoxelGS, NeuS2 │ ├── Material Estimation ──── F-RNG, SRUG, Ambient-Robust IR │ │ │ ├── PBR parameters ──── (albedo, metallic, roughness) │ │ │ └── Environment lighting ──── Spherical harmonics decomposition │ ├── Articulation ──── ArtSplat, SK-GS, ArtMesh, SAGD, ArtiTwinSplat │ │ │ ├── Joint discovery ──── Skeleton auto-discovery │ │ │ ├── Motion fields ──── Deformation fields per part │ │ │ └── Digital twin interaction ──── ArtiTwinSplat (RGB-D digital twin; agent-driven articulated manipulation) │ ├── Spatial Reasoning ──── RAF, FreeArtGS, Argus (ECCV 2026) │ │ │ ├── Visual→Physics abstraction ──── RAF (representation-aware forward mapping) │ │ │ ├── LiDAR-level pose from RGB ──── Argus (如视): image-derived LiDAR-level pose constraints for feed-forward 3DGS │ │ │ └── Free-motion articulation ──── FreeArtGS (ground-plane-free articulation reconstruction) │ ├── Spatial Data Engine ──── Holi-Spatial (ICML 2026 Oral), OpenSpatial (arXiv 2026) │ │ │ ├── Auto data flywheel ──── Holi-Spatial (4M+ samples, 7 task types from video) │ │ │ └── Principled data hierarchy ──── OpenSpatial (3M samples, 5 foundational tasks) │ ├── Streaming Spatial Memory ──── Spatial-TTT (ECCV 2026) │ │ │ └── Test-time training ──── 2B params > GPT-5 on spatial benchmarks │ ├── Neuro-Symbolic Reasoning ──── APEIRIA (ICML 2026) │ │ │ └── MLLM + Z3/SMT verification ──── Open-vocabulary + interpretable spatial proof │ ├── Gaussian Complexity Control ──── DP-Splat (arXiv 2026), SalientGS (arXiv 2026) │ │ │ ├── Bayesian nonparametric ──── DP-Splat: Dirichlet-process prior; data-adaptive component count │ │ │ └── Importance-guided MCMC ──── SalientGS: unified SfM-to-3DGS; 15-min end-to-end │ ├── Dynamic Deformation MoE ──── MoE-GS / MoDE (TPAMI 2026) │ │ │ ├── Joint MoDE ──── Multiple deformation experts on shared canonical Gaussians │ │ │ └── Routed MoE-GS ──── Separate expert optimization + routing stage │ ├── Feed-Forward Generalizable ──── HyperGS, AsySplat, StructSplat, MAC-Splat │ │ │ ├── Optimization-free video GS ──── HyperGS: 10^4-10^5x speedup over per-video optimization │ │ │ ├── Asymmetric arch ──── AsySplat: geometry/appearance decoupling; ~800x speedup │ │ │ └── Sparse-view consistency ──── MAC-Splat (ECCV 2026): +4.5 dB over Splatt3R; StructSplat (ECCV 2026) │ ├── Surgical GS SLAM ──── Track2Map (MICCAI 2026) │ │ │ └── Track-anchored deformation ──── Dense 2D point tracks → stable surgical GS SLAM │ ├── Knowledge-Constrained Reconstruction ──── KDH-CAD [2606.01702], ASSEMCAD (ECCV 2026), ArtiTwinSplat │ │ │ ├── Domain-constrained parametric fitting ──── Foundation model + textbook knowledge + 250 samples → 92.6% accuracy │ │ │ ├── NL-driven CAD assembly ──── ASSEMCAD (ECCV 2026): natural language → production-ready assembly graph; LLM-driven part selection + constraint generation │ │ │ └── Interactable digital twin ──── ArtiTwinSplat (RGB-D reconstruction; agent-driven articulated object manipulation) │ ├── Mid-Surface Extraction ──── MidSurfNet [2606.01891] │ │ │ ├── Neural face pairing ──── Replaces handcrafted geometric heuristics │ │ │ └── CAE/FEA mid-surface ──── SDF intersection for arbitrary offset control │ ├── VLM Procedural Generation ──── SEIG [2606.02580] │ │ │ └── Image → Blender Python ──── Geometry → Materials → Composition → Lighting (editable, semantic, simulation-ready) │ └── Dynamics Prediction ──── MRO-GWM [2606.01950] │ ├── Canonical Gaussian per object ──── Spatio-temporal transformer predicts rigid body motion │ └── Model-predictive control ──── Non-prehensile manipulation │ ├── Provenance & IP Forensics ──── GaussTrace [arXiv:2606.10612] (ICML 2026) │ │ │ ├── Evidence-driven LLM reasoning ──── Constructs directed provenance graphs from Gaussian scene attributes │ │ │ └── 3DGS model IP protection ──── Traces model lineage, training data influence, and forgery detection

3DGS Spatial Intelligence Agent

Capabilities

Scene-Level Reasoning: Given a reconstructed 3DGS scene, infer object parts, materials, articulation structure

CAD-in-the-Loop: Integrate build123d/Open Cascade for parametric model extraction from 3DGS

Multi-Modal I/O: Accept text/prompt input and produce parameterized CAD models or 3DGS scene edits

Articulation Discovery: Identify articulated object structure from Gaussian grouping patterns

Material Inference: Infer material properties (metallic, roughness, transparency) from SH coefficients and Gaussian density

Core Knowledge: Representation Bridge

3DGS → Structured Understanding Pipeline

Pattern

Symptom

Fix

SA-1

Part boundary bleeding in segmentation

Color/feature mixing at object boundaries; Gaussians assigned to wrong part

Use part-aware opacity modulation; apply bilateral filtering on part assignments near boundaries

SA-2

Geometry-mesh topology mismatch

Extracted mesh has non-manifold edges or self-intersections; CAD operations fail

Pre-filter with meshcleaning; validate manifoldness before B-rep construction; use PyMeshLab for repair

SA-3

SH coefficient misinterpretation as material

Confusing view-dependent color (SH coefficients) with intrinsic material properties

Decompose SH into intrinsic (degree 0) and view-dependent (degree 1-3) components; only use degree 0 for material inference

SA-4

Spatial Query Mutex Deadlock in Multi-Agent Scene Editing

Agent hangs indefinitely when two concurrent spatial queries target overlapping Gaussian groups

Replace std::mutex with std::recursive_mutex in SceneGraph::query(); or adopt readers-writer lock where read-only queries share access

SA-5

Stale Gaussian Indices After Densification in CAD-in-the-Loop Pipeline

CAD extraction produces distorted geometry (mirrored faces, collapsed edges) after 3DGS densification step

Register CAD module as densification observer; on density control step, invalidate cached index maps and trigger re-extraction of Gaussian→mesh attribute mapping

SA-6

RAF Representation Round-Trip Drift

When using RAF-style visual→physics→visual round-trip, accumulated quantization error in geometry causes rendered images to shift progressively after each simulation step; no re-projection correction in place

Add re-projection correction after each physics step; quantize at physics resolution then upsample with error feedback; track drift metric per round-trip

SA-7

FreeArtGS Free-Motion Drift

Under free-moving articulated object reconstruction, Gaussian positions drift without ground-plane constraint; articulation joints accumulate position error over long sequences

Enforce ground-plane constraint as regularization loss; add joint-anchor drift penalty; periodic re-alignment via reference frame tracking

SA-8

PARTICULATE Mesh-to-Articulation Inconsistency

Feed-forward articulation prediction from mesh yields inconsistent joint axes when mesh has non-manifold edges; no topological validation before articulation fitting

Validate mesh manifoldness before articulation fitting; reject non-manifold edge regions from joint estimation; use topological cleanup (PyMeshLab) as preprocessing step

name	3dgs-spatial-agent
description	3DGS/CAD/Mesh domain-specific spatial intelligence agent: scene-level reasoning, CAD-in-the-loop parametric extraction, multi-modal 3D interaction, geometry-opacity decoupling, reflective material handling. Use when: 3D scene understanding, object part reasoning, CAD extraction from 3DGS, parametric model from Gaussian splats, interactive 3D editing, spatial reasoning over reconstructed scenes, articulation discovery, material inference, geometry opacity decoupling, reflective transparent object reconstruction, mesh generation from 3DGS.
license	Apache-2.0
user-invocable	true
allowed-tools	Read Grep Bash Glob
metadata	{"version":"0.7.1","author":"jaccen","tags":["3dgs","gaussian-splatting","spatial-intelligence","cad","mesh","agent","scene-understanding","parametric-reconstruction"],"when_to_use":["Understand a 3DGS scene at object-part level","Extract CAD/parametric model from Gaussian splats","Interactive 3D editing of reconstructed scenes","Discover articulation or material properties from 3DGS","Decouple geometry and opacity for reflective/transparent objects","Spatial reasoning over reconstructed scenes","3D场景理解 / CAD提取 / 参数化建模 / 材质推断 / 几何不透明度解耦","3DGS provenance analysis and IP forensics from Gaussian scene attributes"]}

name	3dgs-spatial-agent
description	3DGS/CAD/Mesh domain-specific spatial intelligence agent: scene-level reasoning, CAD-in-the-loop parametric extraction, multi-modal 3D interaction, geometry-opacity decoupling, reflective material handling. Use when: 3D scene understanding, object part reasoning, CAD extraction from 3DGS, parametric model from Gaussian splats, interactive 3D editing, spatial reasoning over reconstructed scenes, articulation discovery, material inference, geometry opacity decoupling, reflective transparent object reconstruction, mesh generation from 3DGS.
license	Apache-2.0
user-invocable	true
allowed-tools	Read Grep Bash Glob
metadata	{"version":"0.7.1","author":"jaccen","tags":["3dgs","gaussian-splatting","spatial-intelligence","cad","mesh","agent","scene-understanding","parametric-reconstruction"],"when_to_use":["Understand a 3DGS scene at object-part level","Extract CAD/parametric model from Gaussian splats","Interactive 3D editing of reconstructed scenes","Discover articulation or material properties from 3DGS","Decouple geometry and opacity for reflective/transparent objects","Spatial reasoning over reconstructed scenes","3D场景理解 / CAD提取 / 参数化建模 / 材质推断 / 几何不透明度解耦","3DGS provenance analysis and IP forensics from Gaussian scene attributes"]}

Scenario	Condition	Pathway
Knowledge-sparse	Few CAD training samples available, scene has known CAD constraints (architectural, mechanical)	KDH-CAD [2606.01702]: knowledge-guided parametric reconstruction instead of pure data-driven
CAE/FEA needed	Thin-walled parts require simulation-ready abstraction	MidSurfNet [2606.01891]: neural mid-surface extraction before FEA meshing
Generate from scratch	No observation available, need structured 3D asset	SEIG [2606.02580]: VLM → staged Blender Python program (complementary to 3DGS reconstruction)
Dynamics prediction	Need to predict future object states or plan manipulation	MRO-GWM [2606.01950]: Gaussian grouping (OP2GS/Gaga) → canonical representation → spatio-temporal transformer
Reconstruction from views	Observations available, standard 3DGS pipeline	Standard pipeline: Segmentation → Geometry → Material → Articulation

Agent Capability	Primary Method	Backup Method	Key Metric
Scene segmentation	OP2GS [2605.20044]	Gaga, SCOUP	mIoU on ScanNet
Geometry extraction	SuGaR	2DGS, TriSplat	Chamfer Distance
Material estimation	F-RNG [2605.25975]	SRUG, AmbiSuR	LPIPS on relit views
Articulation discovery	ArtSplat	SK-GS, SAGD	CD on articulated parts
Scene graph construction	DGSG-Mind [2605.29879]	—	3DVG accuracy
Feed-forward head/avatar	HeadsUp [2605.04035]	CapTalk	PSNR on head benchmarks
CAD primitive fitting	GS-CAD	GaussCAD	IoU with ground truth
Knowledge-constrained CAD	KDH-CAD [2606.01702]	—	92.6% accuracy (250 samples)
Mid-surface extraction	MidSurfNet [2606.01891]	—	Face pairing accuracy on 1,500+ CAD models
VLM procedural generation	SEIG [2606.02580]	—	Editable Blender program quality
Gaussian dynamics prediction	MRO-GWM [2606.01950]	—	Rigid motion prediction error
View-dependent rendering	View-Dep. Kernels [2605.25426]	DP-GES	PSNR/LPIPS on specular

3dgs-spatial-agent

Options d'installation

Vérifiez les fichiers source

3DGS Spatial Intelligence Agent

Capabilities

Core Knowledge: Representation Bridge

3DGS → Structured Understanding Pipeline

3DGS Spatial Intelligence Agent

Capabilities

Core Knowledge: Representation Bridge

3DGS → Structured Understanding Pipeline

Structured Understanding → 3DGS Editing Pipeline

Agent Workflow

Task 1: Scene Understanding from 3DGS

Task 2: CAD Extraction from 3DGS

Task 3: Agent-Driven Scene Editing

Decision Flow

Key Method Cross-References

Bug Patterns Specific to Spatial Agent

Rules

Red Lines

Related Skills

Guardrail: Do Not Apply From Memory

3dgs-spatial-agent

Informations de source

Options d'installation

Vérifiez les fichiers source

Plus depuis ce dépôt

Plus depuis ce dépôt

3DGS Spatial Intelligence Agent

Capabilities

Core Knowledge: Representation Bridge

3DGS → Structured Understanding Pipeline

3DGS Spatial Intelligence Agent

Capabilities

Core Knowledge: Representation Bridge

3DGS → Structured Understanding Pipeline

Structured Understanding → 3DGS Editing Pipeline

Agent Workflow

Task 1: Scene Understanding from 3DGS

Task 2: CAD Extraction from 3DGS

Task 3: Agent-Driven Scene Editing

Decision Flow

Key Method Cross-References

Bug Patterns Specific to Spatial Agent

Rules

Red Lines

Related Skills

Guardrail: Do Not Apply From Memory