100 likes | 218 Vues
This paper discusses the integration of natural language interaction in video surveillance systems, showcasing how human-like understanding can improve video analysis. Key techniques explored include detection, segmentation, activity recognition, and scene understanding. The authors, Jordi Gonzàlez, Josep Mª Gonfaus, Carles Fernández, and F. Xavier Roca, provide insights into the MIPRCV project, demonstrating applications like video annotation and retrieval, agent tracking, and augmented reality, aimed at enhancing the context and classification of video data.
E N D
Exploiting Natural-Language Interaction in Video Surveillance Systems Jordi Gonzàlez, Josep Mª Gonfaus, Carles Fernández, F.Xavier Roca V&L Net workshop on Vision and Language Brighton, September 15th, 2011
ISE Lab: Research Lab on Image Sequence Evaluation • UNDERSTAND videos withhumans … (oractions in images) • …toEXPLAINthem in theircontext(orclassify, search…)
ISE Lab: some research we do... • Detection • Segmentation • Agent/body/face tracking • Activityrecognition • Sceneunderstanding • Behaviorrecognition • Video annotation/ retrieval • Augmentedreality • NL descriptions human horse
An example: the MIPRCV project 156 : … 203 : Lo vianant surt per la part inferior dreta... 252 : ...
An example: the MIPRCV demonstrator Servolens with integrated Zoom Giga Ethernet cam 3 Dedicated Servers Pan & Tilt Control Terminal Network Infrastructure
Exploiting Natural-Language Interaction in Video Surveillance Systems Jordi Gonzàlez, Josep Mª Gonfaus, Carles Fernández, F.Xavier Roca http://iselab.cvc.uab.es/