Early and Late Fusion of Classifiers for the MediaEval 2018 Medico Task

AAU/ITEC researchers Mario Taschwer, Klaus Schöffmann, and Jürgen Primus collaborated with Oge Marques (FAU, USA) on a research challenge targeting automatic classification of endoscopic images of the human gastro-intestinal tract. The challenge was posed as the MediaEval 2018 Medico task and called not only for classification effectiveness, but also for run-time efficiency of proposed approaches. The joint AAU-FAU team therefore proposed a combination of CNN-based feature extraction with traditional machine learning algorithms and achieved respectable results in comparison to deep-learning based approaches submitted by other participants, which has been honored with a “distinctive mention” by Medico task organizers. The AAU-FAU approach has been presented remotely by Mario Taschwer at the MediaEval 2018 Workshop, Sophia Antipolis, France, on October 30, 2018.

Paper and slides are available here.