HCMUS team at the Multimodal Person Discovery in Broadcast TV Task of MediaEval 2016

We present the method of the HCMUS team participating in Multimodal Person Discovery in Broadcast TV Task at the MediaEval Challenge 2016. There are two main processes in our method. First we identify a list of potential characters of interest from all video clips. Each potential character is defined as a pair of face track, a sequence of face patches, and a name. We use OCR results and face detection to find potential characters. We also apply several simple techniques to check the consistency of linking a name with a face track to reduce potential wrong matching pairs. Then we detect face patches from test video shots with cascade DPM, extract deep features from face patches using a very deep Convolutional Neural Network, and classify faces using SVM.