In this paper, we propose a methodology for spatio-temporal wardrobe generation for video content. The main goal is to suggest relevant matches between clothes worn by actors and images originating from a set of e-commerce clothing sites. The semi-automatic generation of fine-grained spatial metadata for each video sequence is based on shot detection, keyframe detection, feature matching and clothing type classification based filtering. The result of this annotation process is a spatio-temporal database consisting of videos and the corresponding actor clothing. This database can be queried in various ways depending on the intended target application.
CITATION STYLE
Vandecasteele, F., Vervaeke, J., Vandersmissen, B., De Wachter, M., & Verstockt, S. (2016). Spatio-temporal wardrobe generation of actors’ clothing in video content. In Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics) (Vol. 9733, pp. 448–459). Springer Verlag. https://doi.org/10.1007/978-3-319-39513-5_42
Mendeley helps you to discover research relevant for your work.