siggraph

# SIGGRAPH Skill **Trit**: 0 (ERGODIC/Coordinator) **Domain**: computer-graphics, research, rendering, animation, simulation **Conference**: ACM SIGGRAPH (Special Interest Group on Computer GRAPHics) --- ## Overview SIGGRAPH is the premier venue for computer graphics research. This skill indexes papers, repos, and techniques from SIGGRAPH 2023-2025. ``` ┌─────────────────────────────────────────────────────────────────────────┐ │ SIGGRAPH RESEARCH DOMAINS │ ├─────────────────────────────────────────────────────────────────────────┤ │ │ │ ┌──────────────┐ ┌──────────────┐ ┌──────────────┐ ┌────────────┐ │ │ │ RENDERING │ │ ANIMATION │ │ GEOMETRY │ │ AI/ML │ │ │ │ │ │ │ │ │ │ │ │ │ │ • NeRF │ │ • Motion │ │ • Meshes │ │ • Diffusion│ │ │ │ • Gaussians │ │ • Rigging │ │ • B-rep │ │ • GAN │ │ │ │ • Ray trace │ │ • Characters │ │ • Splatting │ │ • ControlN │ │ │ └──────────────┘ └──────────────┘ └──────────────┘ └────────────┘ │ │ │ │ ┌──────────────┐ ┌──────────────┐ ┌──────────────┐ ┌────────────┐ │ │ │ SIMULATION │ │ IMAGING │ │ HUMAN │ │ ACCEL │ │ │ │ │ │ │ │ │ │ │ │ │ │ • Physics │ │ • HDR │ │ • Faces │ │ • WebGPU │ │ │ │ • Fluids │ │ • Colorize │ │ • Bodies │ │ • Neural │ │ │ │ • MPM │ │ • Edit │ │ • Motion cap │ │ • Shaders │ │ │ └──────────────┘ └──────────────┘ └──────────────┘ └────────────┘ │ └─────────────────────────────────────────────────────────────────────────┘ ``` --- ## SIGGRAPH 2025 Top Papers | Repo | ★ | Topic | Description | |------|---|-------|-------------| | [VAST-AI-Research/UniRig](https://github.com/VAST-AI-Research/UniRig) | 1274 | Rigging | One Model to Rig Them All | | [XPixelGroup/HYPIR](https://github.com/XPixelGroup/HYPIR) | 1023 | Restoration | Diffusion Score Priors for Image Restoration | | [aigc3d/LAM](https://github.com/aigc3d/LAM) | 891 | Avatars | Large Avatar Model for One-shot Gaussian Head | | [microsoft/renderformer](https://github.com/microsoft/renderformer) | 886 | Rendering | Transformer-based Neural Rendering with GI | | [IGL-HKUST/DiffusionAsShader](https://github.com/IGL-HKUST/DiffusionAsShader) | 796 | Video | 3D-aware Video Diffusion | | [NYU-ICL/image-gs](https://github.com/NYU-ICL/image-gs) | 422 | 2D Gaussians | Content-Adaptive Image Representation | | [PrimitiveAnything](https://github.com/PrimitiveAnything/PrimitiveAnything) | 377 | 3D Gen | Human-Crafted Primitive Assembly | | [3DTopia/LayerPano3D](https://github.com/3DTopia/LayerPano3D) | 305 | Panorama | Layered 3D Panorama Generation | --- ## SIGGRAPH 2024 Top Papers | Repo | ★ | Topic | Description | |------|---|-------|-------------| | [TencentARC/MotionCtrl](https://github.com/TencentARC/MotionCtrl) | 1478 | Motion | Motion Control for Video Generation | | [graphdeco-inria/hierarchical-3d-gaussians](https://github.com/graphdeco-inria/hierarchical-3d-gaussians) | 1351 | Gaussians | Hierarchical 3DGS for Large Datasets | | [hbb1/2d-gaussian-splatting](https://github.com/hbb1/2d-gaussian-splatting) | 2962 | 2DGS | Geometrically Accurate Radiance Fields | | [bytedance/X-Portrait](https://github.com/bytedance/X-Portrait) | 532 | Portraits | Expressive Portrait Animation | | [MisEty/RTG-SLAM](https://github.com/MisEty/RTG-SLAM) | 468 | SLAM | Real-time 3D Reconstruction with Gaussians | | [samxuxiang/BrepGen](https://github.com/samxuxiang/BrepGen) | 378 | CAD | B-rep Generative Diffusion Model | | [AIGAnimation/CAMDM](https://github.com/AIGAnimation/CAMDM) | 286 | Animation | Taming Diffusion for Character Control | | [electronicarts/pbmpm](https://github.com/electronicarts/pbmpm) | 232 | Physics | WebGPU Position Based MPM | --- ## SIGGRAPH 2023 Classics | Repo | ★ | Topic | Description | |------|---|-------|-------------| | [XingangPan/DragGAN](https://github.com/XingangPan/DragGAN) | 36005 | GAN | Interactive Point-based Image Manipulation | | [Doubiiu/ToonCrafter](https://github.com/Doubiiu/ToonCrafter) | 5927 | Animation | Generative Cartoon Interpolation | | [williamyang1991/Rerender_A_Video](https://github.com/williamyang1991/Rerender_A_Video) | 3004 | Video | Zero-Shot Video-to-Video Translation | | [pix2pixzero](https://github.com/pix2pixzero/pix2pix-zero) | 1143 | Image | Zero-shot Image-to-Image Translation | --- ## Key Techniques ### Gaussian Splatting ```python # 3D Gaussian Splatting fundamentals # Each Gaussian: position (μ), covariance (Σ), color (SH), opacity (α) class Gaussian3D: def __init__(self): self.position = np.zeros(3) # μ ∈ R³ self.covariance = np.eye(3) # Σ ∈ R³ˣ³ (positive semi-definite) self.sh_coeffs = np.zeros(48) # Spherical harmonics (RGB × 16) self.opacity = 1.0 # α ∈ [0, 1] def splat(self, camera): # Project to 2D, compute screen-space covariance μ_2d = camera.project(self.position) Σ_2d = camera.project_cov(self.covariance) return μ_2d, Σ_2d ``` ### Neural Radiance Fields (NeRF) ```python # NeRF: F(x, d) → (c, σ) # x = 3D position, d = viewing direction # c = RGB color, σ = volume density def nerf_forward(model, rays_o, rays_d, near, far, n_samples): t = torch.linspace(near, far, n_samples) points = rays_o + t * rays_d # Query MLP rgb, density = model(points, rays_d) # Volume rendering weights = compute_transmittance(density, t) color = (weights * rgb).sum(dim=-1) return color ``` ### Material Point Method (MPM) ```javascript // WebGPU PB-MPM from EA SIGGRAPH 2024 // Position Based Material Point Method struct Particle { position: vec3<f32>, velocity: vec3<f32>, mass: f32, volume: f32, deformation_grad: mat3x3<f32>, } @compute @workgroup_size(256) fn p2g(@builtin(global_invocation_id) id: vec3<u32>) { // Particle to Grid transfer let p = particles[id.x]; let base = floor(p.position / dx); for (var i = 0; i < 27; i++) { let offset = neighbor_offsets[i]; let weight = bspline_weight(p.position, base + offset); atomicAdd(&grid[base + offset].mass, p.mass * weight); atomicAdd(&grid[base + offset].momentum, p.mass * p.velocity * weight); } } ``` --- ## GF(3) Research Classification ``` MINUS (-1): Analysis/Measurement Papers - Perceptual studies - Benchmarks - Quality metrics ERGODIC (0): Method/Algorithm Papers - Novel techniques - Hybrid approaches - Framework design PLUS (+1): Generation/Synthesis Papers - Generative models - Neural rendering - Content creation ``` ### Balanced Research Pipeline ```clojure ;; catp verification for research workflow [:literature-review :method-design :implementation] ; -1 + 0 + 1 = 0 ✓ [:dataset-creation :training :evaluation] ; -1 + 0 + 1 = 0 ✓ [:problem-analysis :algorithm :results] ; -1 + 0 + 1 = 0 ✓ ``` --- ## Resources ### Official - **SIGGRAPH 2025**: https://s2025.siggraph.org/ - **Papers Program**: https://s2025.conference-schedule.org/?filter1=sstype101 - **ACM DL**: https://dl.acm.org/doi/proceedings/10.1145/3721238 ### Curated Lists - **Ke-Sen Huang's Papers**: https://www.realtimerendering.com/kesen/sig2025.html - **Paper Copilot**: https://papercopilot.com/paper-list/siggraph-paper-list/siggraph-2025-paper-list/ - **Paper Digest**: https://www.paperdigest.org/2025/08/siggraph-2025-papers-highlights/ ### Statistics (SIGGRAPH 2025) - **Total Accepted**: 710 - **Technical Papers**: 306 - **TOG Papers**: 24 - **Posters**: 380 - **Location**: Vancouver, Canada --- ## Commands ```bash # Search SIGGRAPH repos gh search repos "siggraph 2025" --sort stars --limit 20 # Clone top paper implementations gh repo clone VAST-AI-Research/UniRig gh repo clone microsoft/renderformer gh repo clone hbb1/2d-gaussian-splatting # Track new SIGGRAPH papers gh api search/repositories -f q="siggraph 2025" --jq '.items[:10] | .[].full_name' ``` --- ## Related Skills | Skill | Trit | Bridge | |-------|------|--------| | `algorithmic-art` | +1 | Procedural generation | | `gay-mcp` | +1 | Color theory for rendering | | `xogot` | +1 | Game engine integration | | `mlx-apple-silicon` | 0 | Neural inference on Metal | | `iroh-p2p` | +1 | Distributed rendering | --- ## SIGGRAPH Asia | Year | Location | Notable Papers | |------|----------|----------------| | 2024 | Tokyo | ToonCrafter, GVHMR, GaussianObject | | 2023 | Sydney | EasyVolcap, Rerender_A_Video | | 2022 | Daegu | VideoReTalking, VToonify | --- **Skill Name**: siggraph **Type**: Research / Computer Graphics **Trit**: 0 (ERGODIC) **GF(3)**: Coordinator role - bridges analysis and synthesis --- ## Autopoietic Marginalia > **The interaction IS the skill improving itself.** Every use of this skill is an opportunity for worlding: - **MEMORY** (-1): Record what was learned - **REMEMBERING** (0): Connect patterns to other skills - **WORLDING** (+1): Evolve the skill based on use *Add Interaction Exemplars here as the skill is used.*