Annotating a video database requires an intensive, time consuming and error prone human effort. However, this is a mandatory task to efficiently describe multimedia contents and train models for automatic content detection. A new selection strategy for active learning methods to minimise human effort in labelling a large database of video sequences is proposed. Formally, active learning is a process where new unlabelled samples are selected iteratively, then presented to users for annotation, and finally added to the training set. The major problem is to then find the best selection function to quickly reach high classification accuracy. It is shown that existing active learning approaches using selective sampling do not maintain their performances when the number of selected samples per iteration increases. The presented selection strategy attempts to provide a solution to this problem. In practice, selecting many samples offers many advantages when dealing with a large amount of data; among them the possibility to share the annotation effort between several users. Finally an attempt to tackle the more realistic and challenging task of multiple label annotation is made. This would reduce to greater extend the human effort for labelling.
展开▼
