… ConvolutionalNeuralNetworks (CNN) algorithms have been achieving outstanding classification capabilities in several complex tasks such as image recognition [2], [3], face …
… ABSTRACT Convolutionalneuralnetworks (CNNs) based deeplearning algorithms require high data flow and computational intensity. For real-time industrial applications, they need to …
K Bong, S Choi, C Kim, S Kang, Y Kim… - … Solid-State Circuits …, 2017 - ieeexplore.ieee.org
… We introduce an ultra-low-power CNN FR processor and a … For ultra-lowpowerconsumption, it adopts 3 key features: 1) … for low-power face detection (FD), 2) an ultra-low-power CNNP …
S Zheng, P Ouyang, D Song, X Li, L Liu… - … on Circuits and …, 2019 - ieeexplore.ieee.org
… This paper proposes an ultra-lowpower speech recognition processor … with energy efficiency of 2.46 pJ/Neuron and powerconsumption of 141 μW. Binary convolutionalneuralnetwork (…
E Park, D Kim, S Kim, YD Kim, G Kim… - … codesign and system …, 2015 - ieeexplore.ieee.org
… Ultralowpower DNN can be realized by dedicated hardware implementations. When … of deeplearning architectures, there are several approaches that utilize multiple neuralnetworks …
H An, S Schiferl, S Venkatesan… - IEEE Journal of Solid …, 2020 - ieeexplore.ieee.org
… Therefore, we propose an ultra-low-power (… power by 26.9×. Second, to understand the scene, we enable hierarchical event recognition through a programmable deepneuralnetwork (…
Q Liu, O Richter, C Nielsen, S Sheik… - Proceedings of the …, 2019 - openaccess.thecvf.com
… abilities with Spiking NeuralNetworks (SNNs). SNNs also enable us to harness the power efficiency of neuromorphic engineering to build ultra-lowpowerneuralnetwork systems. …
A Erdem, C Silvano, T Boesch… - 2018 IEEE 29th …, 2018 - ieeexplore.ieee.org
… efficient mapping of a ConvolutionalNeuralNetwork task graph to the underlying Orlando computing and memory resources required by the execution of convolutional layers by slicing …
… power while [42] has a too big area compared to its computation ability. However, if the targeting application requires ultra-lowpowerconsumption … ability and powerconsumption can …