A fully annotated thermal face database and its application for thermal facial expression recognition

Abstract

A large number of algorithms for processing faces in regular photographs and videos has been published in recent years, making this field one of the most active research areas in computer vision. Most current algorithms require a sufficiently large, manually annotated database for training. While several large databases for the visible spectrum are available, no sufficiently large and fully annotated database for the emerging thermal infrared modality has been published so far. Instead, algorithms in the thermal spectrum usually rely on specific assumptions regarding image content, making them less robust than their data-driven counterparts that are based on machine learning methods. We address this shortcoming by introducing a novel high-resolution thermal infrared face database with extensive manual annotations. We describe the database in detail and show that it can be used for advanced image processing tasks by training algorithms for facial expression recognition using the database. The full database itself, all annotations and the complete source code are freely available from the authors for research purposes at https://github.com/marcinkopaczka/thermalfaceproject. The code and annotations will be made commonly available under BSD license, the image data will be available for download upon agreeing to the terms and conditions for image data given on the website.