Аннотация
We present a novel dual-stage object-detection method. In the first
stage, an object detector based on appropriate visual features is
used to find object candidates. In the second stage, the object candidates
are assigned a confidence value based on local-contextual information.
Our context-based method is called COBA, for COntext BAsed object
detection. At a given detection rate COBA is able to lower the false-detection
rate. Experiments in which frontal human faces are to be detected
show that the number of false positives is lowered by a factor 8.7
at a detection rate of 80% when compared to the current high-performance
object detectors. Moreover, COBA is capable of flexibly using other
new object-detection algorithms as `plug-ins' in the second stage.
Hence, object detection can be straightforwardly improved by our
method a soon as new insights emerge and are available in algorithmic
form.
Пользователи данного ресурса
Пожалуйста,
войдите в систему, чтобы принять участие в дискуссии (добавить собственные рецензию, или комментарий)