Section: Overall Objectives

Highlights of the Year

• Excellent results at TrecVid MED. This year we participated for the second time in the Multimedia Event Detection (MED) track of TrecVid, one of the major benchmarks in automatic video analysis. In this task 25 event categories (from “making a sandwich” to “attempting a bicycle trick”) have to be detected in a video corpus of 4,000 hours. We ranked first out of 13 participants on the ad-hoc event category task, and 2-nd out of 17 participants for the pre-specified event category task.

• ERC advanced grant. In 2012 Cordelia Schmid was awarded an ERC advanced grant for the ALLEGRO project on Active Large-scale LEarninG for visual RecOgnition. The aim of ALLEGRO is to automatically learn from large quantities of data with weak labels. In 2012 C. Schmid was also nominated IEEE fellow.

• Inria Visual Recognition and Machine Learning Summer School. This year we co-organized the third edition of the Inria Visual Recognition and Machine Learning Summer School in Grenoble. It attracted a total of 182 participants (48 from France, 94 from Europe and 40 from America and Asia).