Research map / 3D vision / Face and shape
Clean Implicit 3D Structure From Noisy 2D STEM Images
Hannah Kniesel, Timo Ropinski, Tim Bergner, Kavitha Shaga Devan, Clarissa Read, Paul Walther, Tobias Ritschel, Pedro Hermosilla
Check 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 |
|---|---|---|---|
| jointly model noise with a Normalizing Flow and reconstruct an implicit 3D field using MLE | STEM reconstruction · Synthetic | 3D MSE: 1.064 -> 0.669 | Improved |
| use a Normalizing Flow instead of Gaussian noise | noise distribution modeling · Noise synthesis | JSD: 1.07 -> 0.58 | Improved |
| reduce implicit MLP capacity to 32 features (ablation) | Ours · Synthetic | 3D MSE: 0.669 -> 1.181 | Hurt |
| reduce implicit MLP capacity to 64 features (ablation) | Ours · Synthetic | 3D MSE: 0.669 -> 1.017 | Hurt |
| reduce implicit MLP capacity to 128 features (ablation) | Ours · Synthetic | 3D MSE: 0.669 -> 1.166 | Hurt |
| increase implicit MLP capacity from 32 to 256 features (ablation) | Ours · Synthetic | 3D MSE: 1.181 -> 0.669 | Improved |
| use 2 rather than 8 noise-flow layers (ablation) | Ours · Synthetic | 3D MSE: 0.669 -> 1.054 | Hurt |
| use 4 rather than 8 noise-flow layers (ablation) | Ours · Synthetic | 3D MSE: 0.669 -> 0.866 | Hurt |
| increase noise-flow capacity to 16 layers (ablation) | Ours · Synthetic | 3D MSE: 0.669 -> 1.141 | Hurt |
| account for out-of-focus blur during reconstruction (ablation) | L2Blur · Synthetic, out-of-focus data | 3D MSE: 0.965 -> 0.909 | Improved |
| use an implicit reconstruction instead of an explicit volume (ablation) | STEM reconstruction · Synthetic | 3D reconstruction quality | Improved |
Closest papers by meaning
- Noise Calibration and Spatial-Frequency Interactive Network for STEM Image Enhancement (CVPR 2025 · similarity 0.90)
- Spatial-Frequency Learning for Scanning Transmission Electron Microscopy Image Enhancement (TPAMI 2026 · similarity 0.89)
- A Gaussian Filter-Based 3D Registration Method for Series Section Electron Microscopy (AAAI 2025 · similarity 0.84)
- DRACO: A Denoising-Reconstruction Autoencoder for Cryo-EM (NeurIPS 2024 · similarity 0.82)
- V2V3D: View-to-View Denoised 3D Reconstruction for Light Field Microscopy (CVPR 2025 · similarity 0.81)
- Volume Tells: Dual Cycle-Consistent Diffusion for 3D Fluorescence Microscopy De-noising and Super-Resolution (CVPR 2025 · similarity 0.81)
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.
