In the Placing Task of MediaEval 2014, amongst the 7 participating teams, RECOD was ranked 3rd for 10m-precision, and ranked 1st for 100m-precision. RECOD also got a distinctive mention for employing the most diverse multimodal feature set, which included motion.
RECOD strategy was based on re-ranking, and clustering using Optimum Path Forest. The working notes paper with the details is available at the workshop proceedings site.
RECOD team had both current lab members and RECOD alumni that keep cooperating with us. The current members are Prof. Ricardo Torres, post-doc fellow Giovani Chiachia, Ph.D. students Pedro Mendes Jr. and Rodrigo Calumby (also affiliated with UEFS). The alumni are Daniel Pedronette (now at UNESP – Rio Claro), Jurandy Almeida (now at UNIFESP), Lin Tzy Li (now at CPqD) and Otavio Penatti (now at Samsung Research Institute Brazil) .