We introduce a new large-scale video dataset designed to assess the performance of diverse visual event recognition algorithms with a focus on continuous visual event recognition (CVER) in outdoor areas with wide coverage. Previous datasets for action recognition are unrealistic for real-world surveillance because they consist of short clips showing one action by one individual [15,8]. Datasets have been developed for movies [11] and sports [12], but, these actions and scene conditions do not apply effectively to surveillance videos. Our dataset consists of many outdoor scenes with actions occurring naturally by non-actors in continuously captured videos of the real world. The dataset includes large numbers of instances for 23 event types distributed throughout 29 hours of video. This data is accompanied by detailed annotations which include both moving object tracks and event examples, which will provide solid basis for large-scale evaluation. Additionally, we propose different types of evaluation modes for visual recognition tasks and evaluation metrics along with our preliminary experimental results. We believe that this dataset will stimulate diverse aspects of computer vision research and help us to advance the CVER tasks in the years ahead.

No abstract

No abstract

No abstract

No abstract

No abstract

No abstract

Existing methods for video scene analysis are primarily concerned with learning motion patterns or models for anomaly detection. We present a novel form of video scene analysis where scene element categories such as roads, parking areas, sidewalks and entrances, can be segmented and categorized based on the behaviors of moving objects in and around them. We view the problem from the perspective of categorical object recognition, and present an approach for unsupervised learning of functional scene element categories. Our approach identifies functional regions with similar behaviors in the same scene and/or across scenes, by clustering histograms based on a trajectory-level, behavioral codebook. Experiments are conducted on two outdoor webcam video scenes with low frame rates and poor quality. Unsupervised classification results are presented for each scene independently, and also jointly where models learned on one scene are applied to the other.