Research map / Video / Video segmentation and tracking
Brain Netflix: Scaling Data to Reconstruct Videos from Brain Signals
Camilo L Fosco, Benjamin Lahner, Bowen Pan, Alex Andonian, Emilie L Josephs, Alex Lascelles, Aude Oliva
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 CLIP alignment (contrastive) after masked brain modeling (ablation) | Brainflix stage 1 · BMD test set | SSIM: 0.138 -> 0.178 | Improved |
| add masked brain modeling pretraining (stage 1) before regression (ablation) | MLP regressor · BMD test set | SSIM: 0.123 -> 0.138 | Improved |
| regress BLIP embedding to generate a caption instead of regressing raw conditioning vector (ablation) | Brainflix stage 2 · BMD test set | SSIM: 0.178 -> 0.180 | Improved |
| add HAD dataset to stage 1 pretraining (ablation) | Brainflix · BMD test set | 2-way accuracy: 0.721 -> 0.767 | Improved |
| train on all subjects instead of a single subject (ablation) | Brainflix · BMD subject 1 | SSIM: 0.123 -> 0.129 | Improved |
| replace linear regression with MLP regressor (no stage 1) (ablation) | Brainflix stage 2 · BMD test set | MSE: 0.828 -> 0.721 | Improved |
| multi-subject training versus single-subject Brainflix | Brainflix · CC2017 | 2-way accuracy: 0.871 -> 0.888 | Improved |
| add data from more subjects and datasets for zero-shot regression | Brainflix regressor · BMD and CC2017 subject 1 zero-shot | latent MSE | Improved |
| Brainflix compared to Mind-Video on human evaluation | CC2017 · CC2017 | human 6-way accuracy (%): 83.9 -> 84.7 | No clear effect |
Closest papers by meaning
- CortexVideo: A Semantic-Spatial Dual-Anchor Framework for High-Fidelity fMRI-to-Video Reconstruction (ECCV 2026 · similarity 0.91)
- MindEye2: Shared-Subject Models Enable fMRI-To-Image With 1 Hour of Data (ICML 2024 · similarity 0.90)
- NeuroClips: Towards High-fidelity and Smooth fMRI-to-Video Reconstruction (NeurIPS 2024 · similarity 0.89)
- NeuralFlix: A Simple While Effective Framework for Semantic Decoding of Videos from Non-invasive Brain Recordings (AAAI 2025 · similarity 0.89)
- SIM: Surface-based fMRI Analysis for Inter-Subject Multimodal Decoding from Movie-Watching Experiments (ICLR 2025 · similarity 0.88)
- Neuro-Vision to Language: Enhancing Brain Recording-based Visual Reconstruction and Language Interaction (NeurIPS 2024 · similarity 0.88)
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.
