-A A +A
Effective and meaningful performance evaluation and validation is an important open challenge for activity analysis and event recognition algorithms applied to security and surveillance applications (the so called smart CCTV). This challenge is mainly due to the limited number of commonly accessible datasets and the lack of standard evaluation protocols. Because the accuracy of activity analysis and event recognition results is highly data dependent, algorithms need to be evaluated with large test corpora containing significant statistical data variability.In a period in which the demand for effective security solutions is increasingly growing we have been witnessing a considerable increase in datasets availability; however their size is still too limited for guaranteeing that the results can lead to a conclusive validation of the performance of an algorithm. Moreover, the use of these dataset for formal performance …
Publication date: 
1 Jan 2007

Andrea Cavallaro, P Hosmer, B Fisher, J Ferryman, AG Burger, WK Krebs, F Brémond, J Orwell, A Colombo, V Leung, SA Velastin, R Vezzani, R Cucchiara, C Distante, G Diraco, A Leone, F Yin, D Makris, S Velastin, C Marchessoux, X Desurmont, S Boughorbel, R Koeleman, W Favoreel, C Machy, E Jaspers, S Blunsden

Biblio References: 
Queen Mary, University of London