A Landmark Detection and Iris Prediction Dataset for Gaze Tracking Research
Eye-tracking technology enables acquiring the user’s gaze and using it as an input for a variety of tasks. Primarily, this is realized with external devices (i.e., eye trackers) that incorporate infrared sensors. However, requiring users to have dedicated eye-tracking equipment limits the potential applications of this technology. Therefore, in the last decade, several research groups have pursued the development of gaze tracking solutions that leverage standard RGB cameras such as the webcams embedded in laptops. Unfortunately, these systems have lower accuracy and reliability than eye trackers. Nonetheless, novel landmark detection algorithms and computer vision pipelines based on machine learning might represent a more viable alternative. In this paper, we introduce and share an annotated dataset that can be utilized for developing, evaluating, and optimizing gaze tracking solutions. Our dataset incorporates features predicted using MediaPipe and specifically, Facemesh and Iris, two models designed for real-time image segmentation and object detection. Furthermore, we labeled each sample using an eye-tracking device, which provides a benchmark for studies aimed at training and testing novel gaze tracking algorithms.
7th International Conference on Human Interaction & Emerging Technologies: Artificial Intelligence & Future Applications
Publication Date
April 2024
Document Object Identifier / Link
N/A
{"head\/og\/url":"podcasts\/a-landmark-detection-and-iris-prediction-dataset-for-gaze-tracking-research-4zXyLE1GA","head\/title":" A Landmark Detection and Iris Prediction Dataset for Gaze Tracking Research","head\/description":"Eye-tracking technology enables acquiring the user\u2019s gaze and using it as an input for a variety of tasks. Primarily, this is realized with external devices (i.e., eye trackers) that incorporate infrared sensors. However, requiring users to have dedicated eye-tracking equipment limits the potential applications of this technology. Therefore, in the last decade, several research groups have pursued the development of gaze tracking solutions that leverage standard RGB cameras such as the webcams embedded in laptops. Unfortunately, these systems have lower accuracy and reliability than eye trackers. Nonetheless, novel landmark detection algorithms and computer vision pipelines based on machine learning might represent a more viable alternative. In this paper, we introduce and share an annotated dataset that can be utilized for developing, evaluating, and optimizing gaze tracking solutions. Our dataset incorporates features predicted using MediaPipe and specifically, Facemesh and Iris, two models designed for real-time image segmentation and object detection. Furthermore, we labeled each sample using an eye-tracking device, which provides a benchmark for studies aimed at training and testing novel gaze tracking algorithms.","head\/og\/image":"https:\/\/media.podscholars.com\/f\/podcasts\/4zXyLE1GA.jpg","podcast":{"ID":5,"status":40,"date_created":"2022-09-09 02:59:44","owner_ID":1,"article_ID":5,"audience_type":null,"stats_view":1997,"stats_like":19,"stats_save":0,"stats_play":0,"title":" A Landmark Detection and Iris Prediction Dataset for Gaze Tracking Research","abstract":"Eye-tracking technology enables acquiring the user\u2019s gaze and using it as an input for a variety of tasks. Primarily, this is realized with external devices (i.e., eye trackers) that incorporate infrared sensors. However, requiring users to have dedicated eye-tracking equipment limits the potential applications of this technology. Therefore, in the last decade, several research groups have pursued the development of gaze tracking solutions that leverage standard RGB cameras such as the webcams embedded in laptops. Unfortunately, these systems have lower accuracy and reliability than eye trackers. Nonetheless, novel landmark detection algorithms and computer vision pipelines based on machine learning might represent a more viable alternative. In this paper, we introduce and share an annotated dataset that can be utilized for developing, evaluating, and optimizing gaze tracking solutions. Our dataset incorporates features predicted using MediaPipe and specifically, Facemesh and Iris, two models designed for real-time image segmentation and object detection. Furthermore, we labeled each sample using an eye-tracking device, which provides a benchmark for studies aimed at training and testing novel gaze tracking algorithms.","duration":947,"file_audio":null,"file_video":"631a8ffee8dac.mp4","compressed":1,"author_firstname":"Nicholas","author_lastname":"Caporusso","img_profile":"f\/users\/63da6eaea34b9_th.jpg","article_date":2022,"article_DOI":null,"publication_ID":5,"publication_name":"7th International Conference on Human Interaction & Emerging Technologies: Artificial Intelligence & Future Applications","author_slug":"nicholas-caporusso-4zXyLE1Gw","publication_slug":"7th-international-conference-on-human-interaction-emerging-technologies-artificial-intelligence-future-applications-4zXyLE1GA","keywords":[{"name":"human-computer interaction","slug":"human-computer-interaction-4zXyLE1GE"},{"name":"gaze tracking","slug":"gaze-tracking-4zXyLE1GF"},{"name":"machine learning","slug":"machine-learning-4zXyLE1GG"},{"name":"tensorflow","slug":"tensorflow-4zXyLE1GH"},{"name":"mediapipe","slug":"mediapipe-4zXyLE1GI"},{"name":"landmark detection","slug":"landmark-detection-4zXyLE1GJ"},{"name":"facemesh","slug":"facemesh-4zXyLE1GK"}],"disciplines":[],"publication_authors":[],"authors_links":[],"slug":"a-landmark-detection-and-iris-prediction-dataset-for-gaze-tracking-research-4zXyLE1GA"}}
