TY - GEN
T1 - RedEye
T2 - 43rd International Symposium on Computer Architecture, ISCA 2016
AU - Likamwa, Robert
AU - Hou, Yunhui
AU - Gao, Yuan
AU - Polansky, Mia
AU - Zhong, Lin
N1 - Publisher Copyright:
© 2016 IEEE.
PY - 2016/8/24
Y1 - 2016/8/24
N2 - Continuous mobile vision is limited by the inability to efficiently capture image frames and process vision features. This is largely due to the energy burden of analog readout circuitry, data traffic, and intensive computation. To promote efficiency, we shift early vision processing into the analog domain. This results in RedEye, an analog convolutional image sensor that performs layers of a convolutional neural network in the analog domain before quantization. We design RedEye to mitigate analog design complexity, using a modular column-parallel design to promote physical design reuse and algorithmic cyclic reuse. RedEye uses programmable mechanisms to admit noise for tunable energy reduction. Compared to conventional systems, RedEye reports an 85% reduction in sensor energy, 73% reduction in cloudlet-based system energy, and a 45% reduction in computation-based system energy.
AB - Continuous mobile vision is limited by the inability to efficiently capture image frames and process vision features. This is largely due to the energy burden of analog readout circuitry, data traffic, and intensive computation. To promote efficiency, we shift early vision processing into the analog domain. This results in RedEye, an analog convolutional image sensor that performs layers of a convolutional neural network in the analog domain before quantization. We design RedEye to mitigate analog design complexity, using a modular column-parallel design to promote physical design reuse and algorithmic cyclic reuse. RedEye uses programmable mechanisms to admit noise for tunable energy reduction. Compared to conventional systems, RedEye reports an 85% reduction in sensor energy, 73% reduction in cloudlet-based system energy, and a 45% reduction in computation-based system energy.
KW - computer vision
KW - continuous mobile vision
KW - pre-quantization processing
KW - programmable analog computing
UR - https://www.scopus.com/pages/publications/84988378495
UR - https://www.scopus.com/inward/citedby.url?scp=84988378495&partnerID=8YFLogxK
U2 - 10.1109/ISCA.2016.31
DO - 10.1109/ISCA.2016.31
M3 - Conference contribution
AN - SCOPUS:84988378495
T3 - Proceedings - 2016 43rd International Symposium on Computer Architecture, ISCA 2016
SP - 255
EP - 266
BT - Proceedings - 2016 43rd International Symposium on Computer Architecture, ISCA 2016
PB - Institute of Electrical and Electronics Engineers Inc.
Y2 - 18 June 2016 through 22 June 2016
ER -