Ajou University Industry-academic Cooperation Foundation

Pulsus Technologies

Other

Ajou University Industry Cooperation Foundation

Myung Hoon Sunwoo

U Seok Lee

Jong Hoon Oh

Jong Ha Moon

Jung Won Lee

Jin Hong Kim

Tae Sun Kim

Se Young Eun

Young Jun Lim

Ji Hoon Bae

Byeong Dae Lee

Min Cheol Kim

Joon-Hyeon Park

Jae Hong Roh

Byung Jun Choi

Chan Hwangbo

Seung Yong Kim

Ho Il Bang

Seo Rin Jung

A convolution method for high speed deep learning according to the present invention includes (a) a step in which a feature map receiving unit of the convolution system receives a feature map configured by N channels; (b) a step in which a main controller of the convolution system selects a '0'-th channel from the feature map configured by N channels; (c) a step in which the main controller confirms a coordinate in which x, y coordinate is “0”, from the feature map of the “0”-th channel; (d) a coarse step in which a convolution calculating unit of the convolution system performs a convolution operation and a rectified linear unit (ReLU) operation while shifting by 2 in a horizontal direction and a vertical direction in the feature map; (e) a step in which the channel switching unit of the convolution system switches the channel to a subsequent channel when the coarse step is completed for the feature map of the “0”-th channel; (g) a step in which the main controller determines whether the switched channel is greater or less than N; and (g) a step in which if the channel switched in step (0 is greater than N, the main controller determines that the convolution operation for all channels has been completed and outputs the feature map by means of a feature map output unit. By doing this, the convolution operation which occupies most of the convolution neural network is reduced to increase inference speed in the deep learning.

Convolution method for high speed deep learning

Disclosed are a method for providing the necessary information for a diagnosis of Alzheimer's disease and an apparatus for executing the method. The apparatus for executing the method for providing the necessary information for a diagnosis of Alzheimer's disease includes one or more processors, a memory, and one or more programs, in which the one or more programs are stored in the memory and configured to be executed by the one or more processors, the one or more programs includes an instruction for acquiring a photographed image of a patient's eyeball, an instruction for preprocessing the photographed image, generating a blood vessel image from the pre-processed photographed image using machine learning-based technology, and providing the necessary information for a diagnosis of Alzheimer's disease based on the generated blood vessel image, and an instruction for generating diagnostic prediction information based on the necessary information for a diagnosis of Alzheimer's disease.

Method for providing necessary information for diagnosis of Alzheimer's disease from photographed image of patient using deep learning and apparatus for executing the method

There are disclosed an apparatus and method for processing images. The apparatus for processing images according to an embodiment includes an image input unit configured to receive a first image of a Bayer pattern including noise and an image output unit configured to output a noise-removed image by removing noise from the first image using a noise removal model, and the noise removal model includes a color correlation block configured to generate a second image of the Bayer pattern including RGB correlation information about the first image from the first image by performing rearrange and convolution operations on the first image, a discrete cosine transform (DCT) block configured to generate a third image in which a high-frequency component of the first image is highlighted by applying a DCT to the first image, and one or more discrete wavelet transform (DWT) blocks configured to generate a low-frequency feature map and one or more high-frequency feature maps by applying a DWT to a first feature map generated based on the first image, the second image, and the third image, and generate a final feature map in which a high-frequency component and a low-frequency component of the first feature map are highlighted based on the low-frequency feature map and the one or more high-frequency feature maps.

Apparatus and method for processing image

An apparatus for generating an HDR image includes an input image generator that generates a first image and a second image and an HDR image generator that generates a high dynamic range (HDR) in which a dynamic range of an original image is extended from the first image and the second image using a pre-trained model including a first neural network, a second neural network, and a third neural network, in which the first neural network is pre-trained to output a third image, the second neural network is pre-trained to output a fourth image and the third neural network is pre-trained to generate the HDR image based on the third image and the fourth image.

Apparatus and method for generating HDR image

The present invention relates to a breast image analysis method with four mammogram images which are input to a convolutional neural network as one input and a system therefor and the system includes an image receiving unit which receives four mammogram images; an image size adjusting unit which adjusts a size of a mammogram image received from the image receiving unit; a preprocessing unit which performs preprocessing on the mammogram image adjusted by the image size adjusting unit; a convolutional neural network (CNN)-based CNN learning unit which generates learning information by learning the mammogram image preprocessed by the preprocessing unit; and a CNN inference unit which receives the learning information learned from the CNN learning unit and a mammogram image to be classified from the image receiving unit to diagnose a breast abnormality.

Convolutional neural network based breast image analysis method using four-channel inputs and system therefor

Disclosed are a method and apparatus for generating a decoding position control signal for decoding using polar codes. The method and apparatus for generating a decoding position control signal for decoding using polar codes according to an embodiment of the present disclosure include generating a decoding tree obtained by forming a plurality of nodes in a hierarchical structure for a polar-encoded codeword, decoding the codeword using a successive cancellation (SC) decoding technique, and generating control signal through a preset operation relationship based on a position of a bit returned during re-decoding among the decoded codeword.

Method and apparatus for generating a decoding position control signal for decoding using polar codes

Disclosed are an apparatus and method for successive cancellation flip decoding of a polar code. The apparatus for successive cancellation flip decoding of a polar code according to an embodiment includes an iterative unit subtotal matrix generator configured to generate an iterative unit subtotal matrix corresponding to a preset iterative unit size based on a portion of an entire subtotal matrix, a selection logic configured to determine one or more selection bits based on a bit string representing a position of a bit returned when re-decoding and generate an auxiliary matrix for generating the entire subtotal matrix from the one or more selection bits, and an entire subtotal matrix generator configured to generate the entire subtotal matrix by using the iterative unit subtotal matrix and the auxiliary matrix.

Apparatus and method for successive cancellation flip decoding of polar code

A polar code decoding apparatus according to an embodiment includes a divider configured to generate a decoding tree in which a plurality of nodes including one or more critical sets for a polar-encoded codeword are formed in a hierarchical structure, and divide the decoding tree into one or more partitions, each partition equally including lowest nodes of the decoding tree, a determiner configured to determine a memory size for storing a primary decoding result based on a specific partition, the specific partition being selected from among the one or more partitions based on the number of critical sets included in each partition, and a decoder configured to decode the codeword primarily by using a successive cancellation (SC) decoding technique.

Apparatus and method for successive cancellation bit-flip decoding of polar code

An apparatus for inserting an error correction code according to an embodiment includes an identifier configured to identify a plurality of critical sets included in a codeword generated using a polar code, based on the number of information bits and a total code length of the codeword, a divider configured to divide the codeword into a plurality of partitions, and an inserter configured to insert an error correction code into each of the divided partitions based on a distribution position of each of the plurality of critical sets.

Method and apparatus for inserting error correction code

A method for decoding a signal encoded with polar codes by a decoding system is provided. The method comprises receiving, from a transmission system, a signal in which a plurality of cyclic redundancy checks (CRCs) are encoded by the polar codes, the plurality of CRCs being inserted into positions determined based on a plurality of information bits, a number of the plurality of information bits and a total code length, and decoding a code section including bits ranging from a first bit of the signal to a position where a last bit of a first CRC is inserted. The method further comprises re-performing successive cancellation flip decoding for the decoded code section, or determining whether to decode a next code section adjacent to the decoded code section, based on whether a CRC is detected in the decoded code section.

Decoding system and method for low latency bit-flipping successive cancellation decoding for polar codes

Disclosed are a method and an apparatus for jaundice diagnosis based on an image. The method for jaundice diagnosis based on an image includes: receiving a jaundice diagnostic image acquired by photographing both a specific body part of a user and a reference object at a place where the user is positioned at present; generating color distortion information indicating a color distortion degree of the reference object included in the jaundice diagnostic image; generating a jaundice diagnostic correction image by correcting color distortion of the jaundice diagnostic image based on the color distortion information; and diagnosing a jaundice symptom for the user by using the jaundice diagnostic correction image.

Method and apparatus for jaundice diagnosis based on an image

Disclosed are image based jaundice analyzing method and apparatus. The image based jaundice analyzing method according to an exemplary embodiment of the present disclosure includes: receiving an image for jaundice diagnosis obtained by photographing a specific body part of a user and a reference object in a location where the user is currently located; generating color distortion information representing a degree of color distortion of the reference object included in the image for jaundice diagnosis; generating a correction image for jaundice diagnosis by correcting the color distortion of the image for jaundice diagnosis based on the color distortion information; and diagnosing a jaundice symptom of the user using the correction image for jaundice diagnosis.

Image based jaundice diagnosing method and apparatus and image based jaundice diagnosis assisting apparatus

The present disclosure a method of decoding a polar code based on a shared node, the method includes extracting an input node from target data that are data to be decoded, by an extractor, sorting the input node as one of a first node of which the pattern of the frozen bit satisfies a predetermined first reference, a second node of which the pattern of the information bit satisfies a predetermined second reference, and a third node that is not the first node and the second node, by a sorter, calculating at least one codeword candidate and at least one path metric that correspond to the input node in accordance with the sorting result by a calculator, finishing decoding the target data by iterating the extracting, the sorting as one, and the calculating of at least one path metric by a controller.

Method and apparatus for decoding polar codes based on shared node

Disclosed is a high efficiency video coding (HEVC) encoding device including a candidate group updater configured to select a plurality of representative modes as a candidate group from among intra-prediction modes and update the candidate group using a plurality of minimum modes selected from the candidate group, the plurality of representative modes each representing a range where there is an optimal mode, and an optimal mode selector configured to select any one mode as an optimal mode from among a plurality of minimum modes selected from the updated candidate group.

HEVC encoding device and method for determining intra-prediction mode using the same

Provided is a method of decoding a low-density parity-check code (LDPC). The decoding method including an initialization process, a check node update process, a variable node update process, a tentative decoding process, and a parity check process, for a plurality of check nodes and a plurality of variable nodes, further includes detecting at least one inactive variable nodes that do not require variable node update among the variable nodes, the variable node update process is performed only on active variable nodes except for the inactive variable node, and the check node update process is performed without using the inactive variable node.

Growing community of inventors

Seoul, South Korea

Myung Hoon Sunwoo