Visible to the public A Fast MPEG’s CDVS Implementation for GPU Featured in Mobile Devices

TitleA Fast MPEG’s CDVS Implementation for GPU Featured in Mobile Devices
Publication TypeJournal Article
Year of Publication2018
AuthorsGarbo, A., Quer, S.
JournalIEEE Access
KeywordsCDVS information, Central Processing Unit, Collaboration, Compact Descriptors for Visual Search, comparable precision, composability, computation times, Computer applications, Concurrent computing, CPU-based reference implementation, data structures, descriptor extraction process, Embedded Software, fast MPEG CDVS implementation, feature extraction, GPU data structures, GPU-based approach, graphics processing units, Human Behavior, human factors, Image analysis, image matching, image retrieval, indexing algorithm, Internet, Internet-scale Computing Security, Internet-scale visual search applications, interoperable cross-platform solution, Kernel, main local descriptor extraction pipeline phases, many-cores embedded graphical processor units, matching algorithm, memory access, Metrics, mobile computing, mobile devices, Mobile handsets, Moving Picture Experts Group, MPEG CDVS standard, multiprocessing systems, object detection, parallel algorithms, parallel processing, Policy Based Governance, pubcrawl, resilience, Resiliency, Scalability, Services, Standards, storage management, visual descriptors, visualization
AbstractThe Moving Picture Experts Group's Compact Descriptors for Visual Search (MPEG's CDVS) intends to standardize technologies in order to enable an interoperable, efficient, and cross-platform solution for internet-scale visual search applications and services. Among the key technologies within CDVS, we recall the format of visual descriptors, the descriptor extraction process, and the algorithms for indexing and matching. Unfortunately, these steps require precision and computation accuracy. Moreover, they are very time-consuming, as they need running times in the order of seconds when implemented on the central processing unit (CPU) of modern mobile devices. In this paper, to reduce computation times and maintain precision and accuracy, we re-design, for many-cores embedded graphical processor units (GPUs), all main local descriptor extraction pipeline phases of the MPEG's CDVS standard. To reach this goal, we introduce new techniques to adapt the standard algorithm to parallel processing. Furthermore, to reduce memory accesses and efficiently distribute the kernel workload, we use new approaches to store and retrieve CDVS information on proper GPU data structures. We present a complete experimental analysis on a large and standard test set. Our experiments show that our GPU-based approach is remarkably faster than the CPU-based reference implementation of the standard, and it maintains a comparable precision in terms of true and false positive rates.
Citation Keygarbo_fast_2018