Modeling Scene Activities From Event Relationships and Global Rules“

This material has two sections. The first section details the sampling equations used for inference in the model presented in the submitted paper. In the second section we provide more experimental results that could not be included in the main paper due to space restrictions. 1. Generative model and notations The graphical model is reproduced in Fig. 1. To make this a self-contained, we first recall the generative model, its notations and then provide the inference derivation. It corresponds to the following set of equations: Figure 1: Mixed Event Relationship Model: Shaded cicles represent observed variables. The arrows represent dependencies between variables. Colors green-blue-red are used to indicate dependencies used for a)all events, b)dependent and c)independent events respectively. The method of generating the event occurrence matrix is described as follows.