Abstract:
Billions of online video ads are viewed every month. We aim present a large scale dataset of the response of Indian audience to the online advertisements. The project is an attempt to find a causality between popularity of a multimedia object and the expressions displayed by the subject while viewing it. We also want to investigate the relation between sharing and salient content of advertisements. As an additional step, we would attempt to identify the object of interest in the advertisements and use this as an index for collaborative filtering for estimating the propensity of sharing. The object of interest would be estimated using user behavior data from visual saliency augmented with multi-source modality information of advertisements.