Network in network based weakly supervised learning for visual tracking |
| |
Institution: | 1. Istituto di Matematica Applicata e Tecnologie Informatiche ‘E. Magenes’ - CNR Italy;2. Department of Computer Science, University of Verona Italy;3. University of Science, Ho Chi Minh city, Vietnam;4. Center for Mathematical Morphology - Mines ParisTech - PSL France;5. Department of Engineering, Pontifical Catholic University of Peru, PUCP Peru;6. Millennium Institute Foundational Research on Data, Department of Computer Science, University of Chile, Chile;7. Electrical and Computer Engineering Department, University of Patras, Rion-Patras, Greece;8. Department of Computer Science Aberystwyth University, Aberystwyth, SY23 3DB, UK;9. Department of Computer Science Edge Hill University, Ormskirk, L39 4QP, UK;10. Technology Innovation Institute, Abu Dhabi, UAE;11. KUCARS, Department of Electrical Engineering and Computer Sciences, Khalifa University, UAE;12. University of Florence, Florence, Italy;13. Centre de Recherche et Restauration des Musees de France, Paris, France;14. Vietnam National University, Ho Chi Minh city, Vietnam;15. Université Bourgogne Franche-Comté, Dijon, France;p. Norwegian University of Science and Technology, Gjovik, Norway;1. Machine Learning Research Group, School of Engineering, University of Guelph, Canada;2. Center for Vision Technologies, SRI International, USA;3. Canadian Institute for Advanced Research, Canada |
| |
Abstract: | One of the key limitations of the many existing visual tracking method is that they are built upon low-level visual features and have limited predictability power of data semantics. To effectively fill the semantic gap of visual data in visual tracking with little supervision, we propose a tracking method which constructs a robust object appearance model via learning and transferring mid-level image representations using a deep network, i.e., Network in Network (NIN). First, we design a simple yet effective method to transfer the mid-level features learned from NIN on the source tasks with large scale training data to the tracking tasks with limited training data. Then, to address the drifting problem, we simultaneously utilize the samples collected in the initial and most previous frames. Finally, a heuristic schema is used to judge whether updating the object appearance model or not. Extensive experiments show the robustness of our method. |
| |
Keywords: | Little supervision Handcrafted features Visual tracking Deep network Network in network Object appearance model Large scale training data Drifting problem |
本文献已被 ScienceDirect 等数据库收录! |
|