ZIAEEFARD, BERGEVIN, MORENCY: TIME-SLICE PREDICTION OF DYADIC ACTIVITIES 1 Time-slice Prediction of Dyadic Human Activities

Recognizing human activities from video data is being leveraged for surveillance and human-computer interaction applications. In this paper, we introduce the problem of time-slice activity recognition which aims to explore human activity at a smaller temporal granularity. Time-slice recognition is able to infer human behaviors from a short temporal window. It has been shown that the temporal slice analysis is helpful for motion characterization and in general for video content representation. These studies motivate us to consider time-slices for activity recognition. To this intent, we propose a new family of spatio-temporal descriptors which are optimized for early prediction with time-slice action annotations. Our predictive spatio-temporal interest point (Predict-STIP) representation is based on the intuition of temporal contingency between time-slices. Furthermore, we introduce a new dataset which is annotated at multiple short temporal windows, allowing the modeling of the inherent uncertainty in time-slice activity recognition. Our experimental results show performance comparable to human annotations.