The classification module receives a list of candidates (rectangular windows in practice) that are likely to contain a pedestrian. In this stage, such candidates are classified as pedestrian or non-pedestrian with the goal of minimizing the number of wrong decisions while maximizing right ones. Broadly speaking, this is a pattern recognition module involving vision and machine learning. The former field dealing with image descriptors and pedestrian models. The later one providing algorithms to learn, mostly from labeled samples, the pedestrian/non-pedestrian decision rule (i.e., the pedestrian classifier) based on the mentioned descriptors and models.
CITATION STYLE
Gerónimo, D., & López, A. M. (2014). Classification. In SpringerBriefs in Computer Science (Vol. 0, pp. 23–71). Springer. https://doi.org/10.1007/978-1-4614-7987-1_3
Mendeley helps you to discover research relevant for your work.