TOSCA-MP and HHI at TRECVID 2013: Semantic Indexing

We participated in the semantic indexing task, and submitted the following runs. We experimented with dynamic classifier selection, using runs from the HHI and JRS teams as input. All runs runs were of type M, using parts of the IACC1 data for training, and IACC1.C as a validation set for fusion. The four runs use different methods for selecting the best classifier and determining the resulting score, thus the runs achieve better score when their MAP is determined independently rather than when the binary classification are used to select a classifier in fusion. • TOSCA1: best in terms of AP, max. score of all agreeing classifiers • TOSCA2: best in terms of number of correct classification, max. score of all agreeing classifiers • TOSCA3: best in terms of AP, max. AP as score • TOSCA4: same as TOSCA2, with slightly updated input runs (did not change the fused result) The fused result does not outperform the best of the input classifiers. We found that the main reason for this is that our input classifiers yield better results in terms of ranking than in terms of decision boundary.