Adaptive action detection

摘要

Described is providing an action model (classifier) for automatically detecting actions in video clips, in which unlabeled data of a target dataset is used to adaptively train the action model based upon similar actions in a labeled source dataset. The target dataset comprising unlabeled video data is processed into a background model. The action model is generated from the background model using a source dataset comprising labeled data for an action of interest. The action model is iteratively refined, generally by fixing a current instance of the action model and using the current instance of the action model to search for a set of detected regions (subvolumes), and then fixing the set of subvolumes and updating the current instance of the action model based upon the set of subvolumes, and so on, for a plurality of iterations.

主权项

1. In a computing environment, a method performed on at least one processor comprising: (a) processing a target dataset containing video data into a background model; (b) processing the background model into a current action model by processing a source dataset containing video data including action of interest data and at least some labeled video data comprising labels for actions of interests, using the action of interest data in the source dataset as a prior, and employing maximum a posterior estimation to adapt the background model to the target dataset; (c) searching the target dataset using the current action model to find one or more subvolumes containing similar action of interest data, including estimating locations of the actions of interest in the target dataset and differentiating between spatial-temporal interest points in the background model and spatial-temporal interest points for the action of interests; (d) updating the action model into an updated current action model based at least in part on the one or more subvolumes found in (c); and (e) refining the updated current action model by repeating steps (c) and (d) for a plurality of iterations.