Samsung Electronics Co., Ltd.

Juhee Park

Sanghyuck Ha

Jeongwon Lee

Kyongtae Park

Doyun Kim

Hongseob Shin

A neural network device includes a quantization parameter calculator configured to quantize parameters of a neural network that is pre-trained, so that the quantized parameters are of mixed data types, analyze a statistical distribution of parameter values of an M-bit floating-point type, the parameter values being associated with at least one layer of the neural network, M being a natural number greater than three, obtain a quantization level of each of the parameters statistically covering a distribution range of the parameter values, based on the analyzed statistical distribution, and quantize input data and weights of the M-bit floating-point type into asymmetric input data of an N-bit fixed-point type and weights of an N-bit floating-point type, using quantization parameters that are obtained based on the obtained quantization level of each of the parameters, N being a natural number greater than one and less than M.

Neural network device and method of quantizing parameters of neural network

An electronic device is disclosed. The electronic device includes at least one processor electrically connected with an image sensor and a memory. The memory stores instructions, when executed, causing the processor to obtain an image through the image sensor, segment the obtained image into the plurality of regions, calculate values of a first parameter for each of the plurality of regions based on the reference color components indicating a representative color of each of the plurality of regions, calculate values of a second parameter for each of the plurality of regions based on first pixel values output from the plurality of first light receiving elements included in each of the plurality of regions and second pixel values output from the plurality of second light receiving elements, determine a type of a light source corresponding to each of the plurality of regions, based on a value of the second reference parameter included in data about the correlationship corresponding to the calculated values of the first parameter and values of the second parameter calculated for each of the plurality of regions, and determine a type of a light source of the image based on the determined types of the light source. In addition, various other embodiments recognized from the specification are also possible.

Growing community of inventors

Gyeonggi-do, South Korea

Juhee Park