Research map / Image generation / Autoregressive and latent generation
Alleviating the Semantic Gap for Generalized fMRI-to-Image Reconstruction
Tao Fang, Qian Zheng, Gang Pan
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 |
|---|---|---|---|
| Add structural features with equal weighting (no MOE) (ablation) | GESS · GOD | perceptual similarity (avg): 77.1% -> 73.9% | Hurt |
| Add MOE adaptive weighting of structure with CS (ablation) | GESS · GOD | perceptual similarity (avg): 73.9% -> 81.1% | Improved |
| Add structural information weighted by MOE vs semantics only (ablation) | GESS · GOD | perceptual similarity (avg): 77.1% -> 81.1% | Improved |
| Use GG instead of CS guidance (ablation) | GESS · GOD | perceptual similarity (avg): 81.1% -> 78.4% | Hurt |
| Use GG instead of CS guidance on Sub3 (ablation) | GESS · GOD Sub3 | perceptual similarity: 80.4% -> 82.6% | Improved |
| Switch from random split to generalized split | KNN baseline · NSD | perceptual similarity: 81.0% -> 52.0% | Hurt |
| Switch from random split to generalized split | GESS · NSD | perceptual similarity: 88.8% -> 63.3% | Hurt |
| Use GESS instead of KNN baseline under generalized split | GESS · NSD generalized split | perceptual similarity: 52.0% -> 63.3% | Improved |
| Use GESS instead of KNN baseline under generalized split | GESS · NSD generalized split | SAM: 50.8% -> 60.5% | Improved |
| Use GESS instead of KNN baseline under generalized split | GESS · NSD generalized split | RMSE: 48.5% -> 64.5% | Improved |
| Compare GESS with prior methods on GOD (perceptual similarity) | GESS · GOD | perceptual similarity: 3.8%, 13.8% margin over [22], [27] | Improved |
| Compare GESS with prior methods on GOD (SAM) | GESS · GOD | SAM: 7.2%, 5.3% margin over [22], [27] | Improved |
Closest papers by meaning
- Enhancing Cross-Subject fMRI-to-Video Decoding with Global-Local Functional Alignment (ECCV 2024 · similarity 0.86)
- NeuroClips: Towards High-fidelity and Smooth fMRI-to-Video Reconstruction (NeurIPS 2024 · similarity 0.85)
- MindEye2: Shared-Subject Models Enable fMRI-To-Image With 1 Hour of Data (ICML 2024 · similarity 0.85)
- NSD-Imagery: A Benchmark Dataset for Extending fMRI Vision Decoding Methods to Mental Imagery (CVPR 2025 · similarity 0.84)
- CLIP-MUSED: CLIP-Guided Multi-Subject Visual Neural Information Semantic Decoding (ICLR 2024 · similarity 0.84)
- Neuro-Vision to Language: Enhancing Brain Recording-based Visual Reconstruction and Language Interaction (NeurIPS 2024 · similarity 0.84)
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.
