VEOS: vision-based vertical electrooculography inference from monocular periocular video for ocular artefact suppression in EEG.

On VEOS-Dev, VEOS achieves median Pearson correlationr=0.81to ground-truth vertical EOG and median blink-onset timing error of 18 ms.

On Eye-BCI, the video-only model achieves medianr=0.74, outperforming an eye-tracker baseline.

Cleaning preserves task-relevant electroencephalography structure, including motor imagery (motor imagery) mu/beta rhythms, steady-state visual evoked potentials (steady-state visual evoked potentials) spectral signal-to-noise ratio, and P300 morphology.