tianwenf/mindtoface-downsampled

These are the dataset and models released for the ECCV 2026 Spotlight Paper Mind-to-Face: Neural-Driven Photorealistic Avatar Synthesis via EEG Decoding. The folders contain:

• gs_models: the GaussianAvatar models for the two subjects
• gs_training_datasets: preprocessed face and 3D dataset for training the GS models
• processed_EEG_pointmaps: preprocessed EEG pointmaps
• raw_EEG_recording: raw EEG recording captured, in OpenBCI format
• raw_capture_video_810p: the downsampled multicamera captures of the subjects, synchronized with EEG. We are still uploading the full resolution one, and we'll update the link as soon as it finishes uploading.