Research map / 3D vision / Gaussian splatting
CryoSplat: Gaussian Splatting for Cryo-EM Homogeneous Reconstruction
Suyi Chen, Haibin Ling
Read the paperCheck your idea against 137,217 papers
AI-extracted. The measured effects below were read by AI from the paper's full text (no summary yet). Numbers and claims can be wrong or missing context, so check the paper before you cite or build on it.
Measured effects reported in the paper
| What was changed | Where | Measured | Effect |
|---|---|---|---|
| increase number of Gaussians in cryoSplat (ablation) | cryoSplat · FSC across EMPIAR datasets | FSC resolution: best FSC at 10,000 and 30,000 Gaussians, 4.3 on EMPIAR-10180 | Improved |
| increase Gaussian count from 32,768 to 524,288 (ablation) | cryoSplat · RAG complex | FSC | Improved |
| use anisotropic instead of isotropic Gaussians (N=30,000) (ablation) | cryoSplat · four EMPIAR datasets | FSC | Improved |
| use cryoSplat instead of cryoDRGN on EMPIAR-10049 | cryoDRGN · EMPIAR-10049 | resolution (A): 4.07 -> 2.49 | Improved |
| use cryoSplat instead of cryoSPARC on EMPIAR-10049 | cryoSPARC · EMPIAR-10049 | resolution (A): 4.23 -> 2.49 | Improved |
| use cryoSplat instead of baselines on EMPIAR-10028 | cryoDRGN and cryoSPARC · EMPIAR-10028 | resolution (A): 3.80 for all methods | No clear effect |
| use cryoSplat instead of cryoDRGN for speed | cryoDRGN · runtime (2,048-3,072 Gaussians) | FPS: 2-3x higher FPS; 5 vs 50 epochs; up to 30x overall | Improved |
| use fewer than 10,000 Gaussians (ablation) | cryoSplat · EMPIAR-10028 | FSC: resolution 3.8 under all settings | Hurt |
| use cryoSplat instead of cryoDRGN (batch 8) for memory at D=256 | cryoDRGN · GPU memory D=256 | GPU memory (MiB): 4,906 -> 378 | Improved |
| increase exponential decay gamma from 0.1 to 0.5 (ablation) | cryoSplat · corrected FSC, 30,000 Gaussians | half-map vs corrected FSC gap: gamma 0.1 -> 0.5 | Improved |
| use a single unified learning rate instead of per-parameter rates (ablation) | cryoSplat · optimization stability | stability | Improved |
| lower SNR to -15 dB and -20 dB (ablation) | cryoSplat · synthetic 80S (5,120 Gaussians) | FSC: SNR -15 dB high resolution kept; -20 dB satisfactory | No clear effect |
Closest papers by meaning
- Reconstructing Heterogeneous Biomolecules via Hierarchical Gaussian Mixtures and Part Discovery (NeurIPS 2025 · similarity 0.94)
- CryoBench: Diverse and challenging datasets for the heterogeneity problem in cryo-EM (NeurIPS 2024 · similarity 0.89)
- CryoFM: A Flow-based Foundation Model for Cryo-EM Densities (ICLR 2025 · similarity 0.89)
- CryoSPIN: Improving Ab-Initio Cryo-EM Reconstruction with Semi-Amortized Pose Inference (NeurIPS 2024 · similarity 0.89)
- cryoSPHERE: Single-Particle HEterogeneous REconstruction from cryo EM (ICLR 2025 · similarity 0.89)
- CryoGEM: Physics-Informed Generative Cryo-Electron Microscopy (NeurIPS 2024 · similarity 0.89)
Writing a paper in this area? scime draft is a free LaTeX editor with AI included, real-time co-authors and a check of your references and venue rules before you submit. It also imports OpenAI Prism projects.
