The outstanding performance achieved by Convolutional Neural Network (CNN) comes at the cost of extremely high computational requirements, making them out of reach for most low-power embedded devices [1]. Most of the energy cost during CNN inference comes from memory access as analyzed in [1]. This work focuses on reducing the memory footprint of CNNs to improve energy efficiency. The Approximate Computing (AxC) paradigm leverages the inherent error-resilience of CNNs to improve energy efficiency by relaxing the need for fully accurate operations. CNNs have a high degree of redundancy in terms of their structure and parameters [2], and this redundancy is not always necessary for an accurate prediction. This observation has paved the way for...
La vision par ordinateur est un domaine interdisciplinaire étudiant la manière dont les ordinateurs ...
The impressive breakthroughs of the last two decades in the field of machine learning can be in larg...
Bien qu'ils soient capables de représenter des fonctions très complexes, les réseaux de neurones pro...
The outstanding performance achieved by Convolutional Neural Network (CNN) comes at the cost of extr...
Recently, Convolutional Neural Networks have become the state-of-the-art soluion(SOA) to most comput...
RÉSUMÉ: Les méthodes basées sur l'apprentissage profond, et en particulier les réseaux de neurones c...
Computer vision is an interdisciplinary field that investigates how computers can gain a high level ...
A primary goal of computer vision is to equip machines with the ability to extract information from ...
National audienceRecently, deep neural networks have proven their ability to achieve excellent resul...
This thesis studies empirical properties of deep convolutional neural networks, and in particular th...
Recent development in deep learning have achieved impressive results on image understanding tasks. H...
La capacité des méthodes d'apprentissage profond à exceller en vision par ordinateur dépend fortemen...
Video content now occupies about 82% of global internet traffic. This large percentage is due to the...
Resource de calcul : INSA de Rouen Normandie, cluster de calcul de l'UFR Sciences etTechniques, Le C...
Following the successful use of deep learning (DL) in the field of computer vision and natural langu...
La vision par ordinateur est un domaine interdisciplinaire étudiant la manière dont les ordinateurs ...
The impressive breakthroughs of the last two decades in the field of machine learning can be in larg...
Bien qu'ils soient capables de représenter des fonctions très complexes, les réseaux de neurones pro...
The outstanding performance achieved by Convolutional Neural Network (CNN) comes at the cost of extr...
Recently, Convolutional Neural Networks have become the state-of-the-art soluion(SOA) to most comput...
RÉSUMÉ: Les méthodes basées sur l'apprentissage profond, et en particulier les réseaux de neurones c...
Computer vision is an interdisciplinary field that investigates how computers can gain a high level ...
A primary goal of computer vision is to equip machines with the ability to extract information from ...
National audienceRecently, deep neural networks have proven their ability to achieve excellent resul...
This thesis studies empirical properties of deep convolutional neural networks, and in particular th...
Recent development in deep learning have achieved impressive results on image understanding tasks. H...
La capacité des méthodes d'apprentissage profond à exceller en vision par ordinateur dépend fortemen...
Video content now occupies about 82% of global internet traffic. This large percentage is due to the...
Resource de calcul : INSA de Rouen Normandie, cluster de calcul de l'UFR Sciences etTechniques, Le C...
Following the successful use of deep learning (DL) in the field of computer vision and natural langu...
La vision par ordinateur est un domaine interdisciplinaire étudiant la manière dont les ordinateurs ...
The impressive breakthroughs of the last two decades in the field of machine learning can be in larg...
Bien qu'ils soient capables de représenter des fonctions très complexes, les réseaux de neurones pro...