Tamar Back (EE, Technion)
Tuesday, 29.6.2010, 11:30
The Bag-of-Words (BoW) model is often used for recognition of objects, scenes,
actions and more. It achieves impressive results in many diff erent areas,
although it discards the spatial and temporal order of codewords in a labeled
signal. This work is defi ning a new model: Contextual Sequence of Words (CSoW)
which incorporates temporal order in a BoW model for video representation, and
tests it on action recognition tasks. The temporal context is incorporated in
three scales: global, medium and fine scale context. We show that using CSoW
instead of BoW on the same setups achieves signifi cant improvements in action
recognition rates, for 4 di fferent setups.
*MSc thesis under the supervision of Lihi Zelnik Manor.