US 12,010,434 B2
Solid-state imaging apparatus and electronic device to perform arithmetic operation of neural network
Hayato Wakabayashi, Kanagawa (JP); and Ryoji Ikegaya, Kanagawa (JP)
Assigned to SONY SEMICONDUCTOR SOLUTIONS CORPORATION, Kanagawa (JP)
Appl. No. 17/608,764
Filed by SONY SEMICONDUCTOR SOLUTIONS CORPORATION, Kanagawa (JP)
PCT Filed May 14, 2020, PCT No. PCT/JP2020/019272
§ 371(c)(1), (2) Date Nov. 4, 2021,
PCT Pub. No. WO2020/230850, PCT Pub. Date Nov. 19, 2020.
Claims priority of application No. 2019-092323 (JP), filed on May 15, 2019.
Prior Publication US 2022/0264003 A1, Aug. 18, 2022
Int. Cl. H04N 23/80 (2023.01); G06V 10/82 (2022.01)
CPC H04N 23/80 (2023.01) [G06V 10/82 (2022.01)] 11 Claims
OG exemplary drawing
 
1. A solid-state imaging apparatus, comprising:
a pixel array unit, wherein
the pixel array unit includes:
a plurality of first pixels configured to generate electric signals as a plurality of first pixel signals, and
a plurality of second pixels different from the plurality of first pixels, and
the generated electric signals have a logarithmic characteristic with respect to light quantity;
a control unit configured to selectively execute one of a first read processing to read the plurality of first pixel signals from the plurality of first pixels or a second read processing to read a plurality of second pixel signals from the plurality of second pixels; and
a processing unit configured to:
perform, in a case of execution of the first read processing, arithmetic processing for a first neural network based on a plurality of pieces of first input data, and a plurality of logarithmic weighting factors which express strength of a connection between a plurality of first nodes of the first neural network by a logarithm, wherein the plurality of pieces of first input data is based on the plurality of first pixel signals read from the pixel array unit; and
perform, in a case of execution of the second read processing, the arithmetic processing for a second neural network based on a plurality of pieces of second input data and a plurality of weighting factors that indicates strength of a connection between a plurality of second nodes of the second neural network, wherein the plurality of pieces of second input data is based on the plurality of second pixel signals.