Towards large scale multimedia indexing: A case study on person discovery in broadcast news

Nam Le (1) , Hervé Bredin (2) , Gabriel Sargent (3) , Miquel India (4) , Paula Lopez-Otero (5) , Claude Barras (2) , Camille Guinaudeau (2) , Guillaume Gravier (3) , Gabriel Barbosa da Fonseca (6) , Izabela Lyon Freire (6) , Zenilton Patrocínio Jr (6) , Silvio Jamil F. Guimarães (6) , Gerard Martí (4) , Josep Ramon Morros (4) , Javier Hernando (4) , Laura Docio-Fernandez (5) , Carmen Garcia-Mateo (5) , Sylvain Meignier (7) , Jean-Marc Odobez (1)

1 IDIAP Research Institute
2 LIMSI - Laboratoire d'Informatique pour la Mécanique et les Sciences de l'Ingénieur
3 LinkMedia - Creating and exploiting explicit links between multimedia fragments
4 UPC - Universitat Politècnica de Catalunya [Barcelona]
5 Universidade de Vigo
6 PUC Minas
7 LIUM - Laboratoire d'Informatique de l'Université du Mans

The rapid growth of multimedia databases and the human interest in their peers make indices representing the location and identity of people in audiovisual documents essential for searching archives. Person discovery in the absence of prior identity knowledge requires accurate association of audiovisual cues and detected names. To this end, we present 3 different strategies to approach this problem: clustering-based naming, verification-based naming, and graph-based naming. Each of these strategies utilizes different recent advances in unsupervised face / speech representation, verification, and optimization. To have a better understanding of the approaches, this paper also provides a quantitative and qualitative comparative study of these approaches using the associated corpus of the Person Discovery challenge at MediaEval 2016. From the results of our experiments, we can observe the pros and cons of each approach, thus paving the way for future promising research directions.

Keywords: Multimedia; Multimodal analysis; person discovery

Domains: Computer Science [cs]; Multimedia [cs.MM]; Computer Science [cs]; Sound [cs.SD]; Computer Science [cs]; Signal and Image Processing; Engineering Sciences [physics]; Signal and Image processing; Computer Science [cs]; Document and Text Processing; Computer Science [cs]; Computer Vision and Pattern Recognition [cs.CV]; Computer Science [cs]; Machine Learning [cs.LG]

Nam Le, Hervé Bredin, Gabriel Sargent, Miquel India, Paula Lopez-Otero, et al.. Towards large scale multimedia indexing: A case study on person discovery in broadcast news. Content-Based Multimedia Indexing CBMI, Jun 2017, Firenze, Italy. ⟨10.1145/3095713.3095732⟩. ⟨hal-01551690⟩