Compact convolutional neural network cascade for face detection

I. A. Kalinovskii, V. G. Spitsyn

Результат исследований: Материалы для журналаСтатья

1 цитирование (Scopus)

Выдержка

This paper presents a new solution to the frontal face detection problem based on a compact convolutional neural networks cascade. Test results on an FDDB dataset show that it is able to compete with state-of-the-art algorithms. This proposed detector is implemented using three technologies: SSE/AVX/AVX2 instruction sets for Intel CPUs, Nvidia CUDA, and OpenCL. The detection speed of our approach exceeds considerably all the existing CPU-based and GPU-based algorithms. Thanks to its high computational efficiency, our detector can process 4K Ultra HD video stream in real time (up to 27 fps) on mobile platforms while searching objects with a dimension of 60×60 pixels or higher. At the same time, its processing speed is almost independent of the background and the number of objects in a scene. This is achieved by asynchronous computation of stages in the cascade.

Язык оригиналаАнглийский
Страницы (с-по)375-387
Число страниц13
ЖурналCEUR Workshop Proceedings
Том1576
СостояниеОпубликовано - 2016

Отпечаток

Face recognition
Program processors
Detectors
Neural networks
Computational efficiency
Pixels
Processing
Graphics processing unit

ASJC Scopus subject areas

  • Computer Science(all)

Цитировать

Compact convolutional neural network cascade for face detection. / Kalinovskii, I. A.; Spitsyn, V. G.

В: CEUR Workshop Proceedings, Том 1576, 2016, стр. 375-387.

Результат исследований: Материалы для журналаСтатья

@article{71bb1bdf9148450ea85d8420ffd070c7,
title = "Compact convolutional neural network cascade for face detection",
abstract = "This paper presents a new solution to the frontal face detection problem based on a compact convolutional neural networks cascade. Test results on an FDDB dataset show that it is able to compete with state-of-the-art algorithms. This proposed detector is implemented using three technologies: SSE/AVX/AVX2 instruction sets for Intel CPUs, Nvidia CUDA, and OpenCL. The detection speed of our approach exceeds considerably all the existing CPU-based and GPU-based algorithms. Thanks to its high computational efficiency, our detector can process 4K Ultra HD video stream in real time (up to 27 fps) on mobile platforms while searching objects with a dimension of 60×60 pixels or higher. At the same time, its processing speed is almost independent of the background and the number of objects in a scene. This is achieved by asynchronous computation of stages in the cascade.",
keywords = "Cascade classifiers, Convolutional neural networks, Deep learning, Face detection",
author = "Kalinovskii, {I. A.} and Spitsyn, {V. G.}",
year = "2016",
language = "English",
volume = "1576",
pages = "375--387",
journal = "CEUR Workshop Proceedings",
issn = "1613-0073",
publisher = "CEUR-WS",

}

TY - JOUR

T1 - Compact convolutional neural network cascade for face detection

AU - Kalinovskii, I. A.

AU - Spitsyn, V. G.

PY - 2016

Y1 - 2016

N2 - This paper presents a new solution to the frontal face detection problem based on a compact convolutional neural networks cascade. Test results on an FDDB dataset show that it is able to compete with state-of-the-art algorithms. This proposed detector is implemented using three technologies: SSE/AVX/AVX2 instruction sets for Intel CPUs, Nvidia CUDA, and OpenCL. The detection speed of our approach exceeds considerably all the existing CPU-based and GPU-based algorithms. Thanks to its high computational efficiency, our detector can process 4K Ultra HD video stream in real time (up to 27 fps) on mobile platforms while searching objects with a dimension of 60×60 pixels or higher. At the same time, its processing speed is almost independent of the background and the number of objects in a scene. This is achieved by asynchronous computation of stages in the cascade.

AB - This paper presents a new solution to the frontal face detection problem based on a compact convolutional neural networks cascade. Test results on an FDDB dataset show that it is able to compete with state-of-the-art algorithms. This proposed detector is implemented using three technologies: SSE/AVX/AVX2 instruction sets for Intel CPUs, Nvidia CUDA, and OpenCL. The detection speed of our approach exceeds considerably all the existing CPU-based and GPU-based algorithms. Thanks to its high computational efficiency, our detector can process 4K Ultra HD video stream in real time (up to 27 fps) on mobile platforms while searching objects with a dimension of 60×60 pixels or higher. At the same time, its processing speed is almost independent of the background and the number of objects in a scene. This is achieved by asynchronous computation of stages in the cascade.

KW - Cascade classifiers

KW - Convolutional neural networks

KW - Deep learning

KW - Face detection

UR - http://www.scopus.com/inward/record.url?scp=84978513217&partnerID=8YFLogxK

UR - http://www.scopus.com/inward/citedby.url?scp=84978513217&partnerID=8YFLogxK

M3 - Article

AN - SCOPUS:84978513217

VL - 1576

SP - 375

EP - 387

JO - CEUR Workshop Proceedings

JF - CEUR Workshop Proceedings

SN - 1613-0073

ER -