Through the Eyes of Emotion - Raw Dataset (Part 1)

This is a PART-1 of the full raw VR eye‑tracking dataset of multimodal recordings collected during controlled emotion elicitation, designed for research in affective computing, emotion recognition, and multimodal learning. It includes synchronized periocular videos (120 fps, 400×400) from both eyes, gaze (~240 Hz), pupil diameter (~120 Hz), IMU signals, scene recordings, and audio, organized by participant (Pxx) and session (e.g., 0a/0b… for Study One and 0c/0d… for Study Two), with each session containing periocular/, scene/, gaze_pupil/, imu/, and audio/ subfolders. Seven basic emotions (Surprise, Happiness, Anger, Fear, Disgust, Sadness, Neutral) are elicited using validated film clips, and segment‑level self‑reported intensity ratings (1–10) are provided via timestamps consolidated in a json file in our repo: https://github.com/MultiRepEyeVR/Through-the-Eyes-of-Emotion, which maps sessions to moments when the intended emotion was reported. The dataset supports tasks such as emotion recognition from eye‑region video, gaze and pupil dynamics, sequence modeling and multimodal fusion, and analysis of gaze behavior under emotional stimuli. Data were collected under informed consent with ethics approval; please cite “Through the Eyes of Emotion: A Multi‑faceted Eye Tracking Dataset for Emotion Recognition in Virtual Reality” when using this resource.