Announcing the 2021 November SPRINGEROPEN EURASIP JIVP's Free Web conferencing (Nov. 4, 2021)

3 Novembre 2021

---

Catégorie : Autres événements

---

Abstract: Recent years have witnessed a substantial increase in the deep learning (DL) architectures proposed for visual recognition tasks like person re-identification, where individuals must be recognized over multiple distributed cameras. Although these architectures have greatly improved the state-of-the-art accuracy, the computational complexity of the CNNs commonly used for feature extraction remains an issue, hindering their deployment on platforms with limited resources, or in applications with real-time constraints. There is an obvious advantage to accelerating and compressing DL models without significantly decreasing their accuracy. However, the source (pruning) domain differs from operational (target) domains, and the domain shift between image data captured with different non-overlapping camera viewpoints leads to lower recognition accuracy. In this paper, we investigate the prunability of these architectures under different design scenarios. This paper first revisits pruning techniques that are suitable for reducing the computational complexity of deep CNN networks applied to person re-identification. Then, these techniques are analyzed according to their pruning criteria and strategy, and according to different scenarios for exploiting pruning methods to fine-tune networks to target domains. Experimental results obtained using DL models with ResNet feature extractors, and multiple benchmarks re-identification datasets, indicate that pruning can considerably reduce network complexity while maintaining a high level of accuracy. In scenarios where pruning is performed with large pre-training or fine-tuning datasets, the number of FLOPS required by ResNet architectures is reduced by half, while maintaining a comparable rank-1 accuracy (within 1\% of the original model). Pruning while training larger CNNs can also provide significantly better performance than fine-tuning smaller ones.

Short bio: Md Amran Hossen Bhuiyan received the Bachelor degree in Applied Physics, Electronic & Communication Engineering from the University of Dhaka, Bangladesh in 2009, the M.Sc. degree in Computer Engineering and Information Technology from the Lucian Blaga University of Sibiu, Romania under the Erasmus Mundus external window in 2011 and the Ph.D. degree in Pattern Analysis and Computer Vision from the Istituto Italiano di Tecnologia, Genova, Italy. Currently, he is working as Associate Professor in the Department of Computer Science and Telecommunication Engineering at Noakhali Science and Technology University, Bangladesh. Previously, he was a Mitacs Elevate Postdoctoral Researcher with LIVIA, École de Technologie Supérieure, Université du Québec, Montréal, Canada and the industrial partner organization is SPORTLOGiQ INC. His main research interests include computer vision, machine learning techniques for image and video processing, with applications such as video surveillance, summarizations, and sports analytics, and seeks to match or recognize individuals across non-overlapping views in a multi-camera system.

 

Previous webinar videos are available online at https://vimeo.com/showcase/8005816