Neurotechnology & AI · 2026
DreamWalking: Real-Time EEG to 3D
Master's thesis project decoding visual imagination from 64-channel EEG signals into real-time 3D objects rendered in Unreal Engine 5.
Category
Neurotechnology & AI
Year
2026
Highlight
BCI Research
Brain-Computer Interface for 3D Generation
Developed for my M.Sc. Thesis at Politecnico di Torino (105/110 grade), DreamWalking is a real-time closed loop that reads non-invasive 64-channel EEG, decodes visual imagination, and renders volumetric objects in Unreal Engine 5.
- Channel-Agnostic Encoder: Positionally embedded electrode coordinates generalizing across subjects.
- Multimodal Alignment: Soft-contrastive DINOv2 visual feature alignment.
- Validation Benchmark: 74.4% zero-shot two-way identification on unseen concept pairs (p=0.005).