Samsung Electronics Co., Ltd.

Sanghyuck Ha

Byeoungsu Kim

Sangsoo Ko

Kyoungyoung Kim

Jaegon Kim

Hyoungjun Jeon

Changgwun Lee

Doyun Kim

Sangmin Suh

Juhee Park

An operation method of a processor configured to operate based on an artificial neural network including a first neural network and a second neural network is provided. The operation method includes: analyzing similarity of a structure of the first neural network and a structure of the second neural network, the first neural network including a plurality of first neural network layers and the second neural network including a plurality of second neural network layers; selecting, from the plurality of first neural network layers and the plurality of second neural network layers, sharing layers capable of being commonly used, based on a result of the analysis; and reconstructing the structure of the first neural network or the structure of the second neural network based on the sharing layers.

Processor for reconstructing artificial neural network, electrical device including the same, and operating method of processor

The inventive concept relates to a neural network device, an operation method of the neural network device, and a neural network system including the neural network device, the operation method of the neural network device includes, receiving raw data from a sensor, converting the raw data to a feature map suitable for a first hidden layer among a plurality of hidden layers of a target neural network using a sensor response network adaptively trained to the sensor, inputting the feature map to the first hidden layer of the target neural network, and driving the neural network on a basis of the feature map.

Neural network device, operation method thereof, and neural network system including the same

An advanced driver assist system (ADAS) includes a processing circuit and a memory storing instructions executable by the processing circuit. The processing circuit executes the instructions to cause the ADAS to: obtain, from a vehicle, a video sequence including a plurality of frames captured while driving the vehicle, where each of the frames corresponds to a stereo image including a first viewpoint image and a second viewpoint image; determine depth information in the stereo image based on reflected signals received while driving the vehicle; fuse the stereo image and the depth information to generated fused information, and detect at least one object included in the stereo image based on the fused information.

Advanced driver assist system, method of calibrating the same, and method of detecting object in the same

An image sensor module includes an image sensor configured to generate image data; a memory including a processor in memory (PIM) circuit, of the PIM circuit including banks configured to store the image data and a plurality of processing elements; and a signal processor, wherein the memory is configured to read the image data from the banks, perform a first image processing operation on the image data by using the processing elements corresponding to the banks, and store image-processed image data on which the first image processing operation is performed in the banks or output the image-processed image data.

Image sensor module and method of operating the same

A method of operating a neural network device including a plurality of layers, includes receiving sensing data from at least one sensor, determining environmental information, based on the received sensing data, determining multiple layers corresponding to the determined environmental information, and dynamically reconstructing the neural network device by changing at least two layers, among the plurality of layers, to the determined multiple layers.

Method and system for learning neural network and device

An integrated circuit for generating a product of an input and a constant includes a lookup table memory configured to store seeds corresponding to multiples of the constant, processing circuitry configured to generate a plurality of addresses respectively corresponding to a plurality of parts extracted from the input, configured to receive a plurality of seeds from the lookup table memory based on the plurality of addresses, and configured to generate a plurality of partial products based on the plurality of seeds. The processing circuitry is configured to sum the plurality of partial products. A number of the plurality of seeds stored in the lookup table memory is less than a number of possible values of the plurality of partial products.

Integrated circuit for constant multiplication and device including the same

ADAS includes a processing circuit and a memory which stores instructions executable by the processing circuit. The processing circuit executes the instructions to cause the ADAS to receive, from a vehicle that is in motion, a video sequence, generate a position image including at least one object included in the stereo image, generate a second position information associated with the at least one object based on reflected signals received from the vehicle, determine regions each including at least a portion of the at least one object as candidate bounding boxes based on the stereo image and the position image, and selectively adjusting class scores of respective ones of the candidate bounding boxes associated with the at least one object based on whether a respective first position information of the respective ones of the candidate bounding boxes matches the second position information.

Advanced driver assist system and method of detecting object in the same

A streaming accelerator includes a first pool, a first switch bus, a second pool and a second switch bus. The first pool includes neural processing unit (NPU) bundles, and each of NPU bundles includes a plurality of NPUs. The first switch bus provides a first streaming data to a first selected NPU bundle and a second selected NPU bundle respectively. The second pool includes network interface card (NIC) bundles, and each of the NIC bundles includes an encoder and a NIC. The second switch bus provides a first intermediate streaming data and a second intermediate streaming data to a first selected NIC bundle and a second selected NIC bundle. The first selected NIC bundle encodes the first intermediate streaming data to generate a first encoded streaming data. The second selected NIC bundle encodes the second intermediate streaming data to generate a second encoded streaming data.

Streaming accelerators and streaming systems including the same

An electronic device configured to control a host vehicle includes: an image sensor configured to photograph a surrounding environment of the host vehicle; and a processor configured to perform an image processing operation based on a first image captured by the image sensor, and control the host vehicle based on the processing result, wherein the processor determines whether to use a high speed performance of the image processing operation based on a speed of the host vehicle, and the electronic device is configured such that when the high speed performance is not used, the processor performs the image processing operation by using a first image processing module, and when the high speed performance is used, the processor performs the image processing operation by using a second image processing module having less data throughput than the first image processing module.

Electronic device for controlling driving vehicle and operation method of the electronic device

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 image sensor module includes an image sensor configured to generate image data and memory including at least a memory bank storing the image data and a processor-in-memory (PIM) circuit, the PIM circuit including a plurality of processing elements. The memory is configured to read the image data from the memory bank; generate optical flow data and pattern density data using the plurality of processing elements, the optical flow data indicating time-sequential motion of at least one object included in the image data, and the pattern density data indicating a density of a pattern of the image data; and output the image data, the optical flow data, and the pattern density data.

Image sensor module, image processing system, and operating method of image sensor module

A neural network system for processing a neural network model including an operation processing graph that includes a plurality of operations, includes an operation processor including an internal memory storing a first module input feature map. The operation processor is configured to obtain a first branch output feature map by performing a first operation among the plurality of operations, based on the stored first module input feature map, and obtain a second branch output feature map by performing a second operation among the plurality of operations after the first operation is performed, based on the stored first module input feature map. The internal memory maintains storage of the first module input feature map while the first operation is performed.

Growing community of inventors

Yongin-si, South Korea

Sanghyuck Ha