Título:
|
Work-efficient parallel non-maximum suppression for embedded GPU architectures
|
Autor/a:
|
Oro Garcia, David; Fernandez Tena, Carles; Martorell Bofill, Xavier; Hernando Pericás, Francisco Javier
|
Otros autores:
|
Universitat Politècnica de Catalunya. Departament d'Arquitectura de Computadors; Universitat Politècnica de Catalunya. Departament de Teoria del Senyal i Comunicacions; Universitat Politècnica de Catalunya. CAP - Grup de Computació d'Altes Prestacions; Universitat Politècnica de Catalunya. VEU - Grup de Tractament de la Parla |
Abstract:
|
With the emergence of GPU computing, deep neural networks have become a widely used technique for advancing research in the field of image and speech processing. In the context of object and event detection, slidingwindow classifiers require to choose the best among all positively discriminated candidate windows. In this paper, we
introduce the first GPU-based non-maximum suppression (NMS) algorithm for embedded GPU architectures. The obtained results show that the proposed parallel algorithm reduces the NMS latency by a wide margin when compared to CPUs, even clocking the GPU at 50% of its maximum frequency on an NVIDIA Tegra K1. In this paper, we show
results for object detection in images. The proposed technique is directly applicable to speech segmentation tasks such as speaker diarization. |
Abstract:
|
Peer Reviewed |
Materia(s):
|
-Àrees temàtiques de la UPC::Informàtica::Sistemes d'informació -Àrees temàtiques de la UPC::Informàtica -Embedded computer systems -Information display systems -Embedded systems -Graphics processing units -Parallel algorithms -Work-efficient parallel nonmaximum suppression -Embedded GPU architectures -Image processing -Speech processing -Deep neural networks -NMS latency -Positively discriminated candidate windows -Parallel algorithm -NVIDIA Tegra JC1 -Speech segmentation tasks -Speaker diarization -Sistemes incrustats (Informàtica) -Visualització (Informàtica) |
Derechos:
|
|
Tipo de documento:
|
Artículo - Versión publicada Objeto de conferencia |
Editor:
|
Institute of Electrical and Electronics Engineers (IEEE)
|
Compartir:
|
|