Gwangju Institute of Science and Technology

Disclosed herein is a method for generating a grasp map including receiving a prompt generated to specify a target object together with an image obtained by capturing the target object, extracting a feature map corresponding to the image using an image encoder, inputting the prompt to a prompt encoder to generate a token for the target object determined to be grasped in the image, and inputting the feature map and the token to a mask decoder to generate a mask indicating an area corresponding to the target object in the image and a grasp map indicating an area where the target object is graspable.

Disclosed herein is a fine-tuning method including: confirming training data composed of input data and ground-truth data labeled for the input data; training a linear layer parameter of a pre-trained foundation model using the training data; extracting, from the training data, filtering data in which a pair of the input data and the ground-truth data satisfies a predefined matching condition, based on the foundation model in which the linear layer parameter is trained according to the training data; and training an adapter parameter for the foundation model using the filtering data.

An electronic device for removing noise when rendering images and an operation method thereof are provided. The operation method may include generating, based on a view point of a current frame, a first image, a second image, and geometry (G)-buffer images; generating, based on a relationship between the first image and the second image, a first intermediate image and a second intermediate image, from which a portion of noise in the first image and the second image is removed, respectively; determining bandwidth parameters by inputting, to an artificial neural network, the first intermediate image, the second intermediate image, the G-buffer images, and a reprojected image obtained when an output image of a previous frame is reprojected, and generating, using a first filter and a second filter to which the bandwidth parameters are applied, a target image by removing noise from the first intermediate image and the second intermediate image.

A problem solving method, including: confirming an example problem and a test problem included in an abstraction and inference problem; generating a hierarchical knowledge graph having a plurality of layers by hierarchically analyzing the example problem from a pixel level to a problem level, and specifying core knowledge appearing in the example problem based on the hierarchical knowledge graph; generating the hierarchical knowledge graph having the plurality of layers by hierarchically analyzing the test problem from the pixel level to the problem level, and specifying a candidate component corresponding to the core knowledge based on the hierarchical knowledge graph for the test problem; and generating a solution to the test problem by synthesizing the core knowledge into the candidate component.

The present invention provides a device including a region of interest setting unit that sets a region of interest including a third molar and a periodontal region, a parameter calculation unit that calculates at least one parameter for periodontal disease prognosis evaluation on the basis of image information within the region of interest, a tooth extraction difficulty evaluation unit that evaluates the tooth extraction difficulty for the third molar on the basis of a previously created deep learning algorithm with the parameter as an input, and a complications prediction unit that predicts complications depending on a degree of invasion between the third molar and an IAN, and the parameter calculation unit calculates a parameter among an impaction depth of the third molar, the distance between the third molar and a lower jawbone, an angle of the third molar, and the degree of invasion between the third molar and the IAN.

Provided is a method for generating an initial guess for simulating a semiconductor device. The method includes; (a) setting 1D direction for a 3D semiconductor device to be simulated and 2D cross sections perpendicular to the 1D direction; (b) setting a charge intensity-gate voltage model for the 2D cross sections; (c) numerically solving an electron or hole continuity equation along the 1D direction to detect integrated electron charge density and quasi-Fermi potential for the 2D cross sections; and (d) calculating a physical quantity for each position of the 2D cross sections by applying the integrated electron charge density and the quasi-Fermi potential for the 2D cross sections to the charge density-gate voltage model, and providing the calculated physical quantity as the initial guess for the semiconductor device simulation. Therefore, the initial solution according to the present invention can accelerate the speed of the semiconductor device simulation.

In a process for producing formate, a mixed enzyme by mixing hydrogenase (H 2 ase) with oxygen tolerance and formate dehydrogenase (FDH) with oxygen tolerance is prepared, and the mixed enzyme and a gas including H 2 , CO 2 and NAD + are mixed such that formate may be produced even in the presence of oxygen, and thereby utilizing hydrogen sources including oxygen, such as coke oven gas.

The present inventive concept relates to a hydride ion conductor represented by Formula 1, having a cubic crystal structure, and having the P 4 3m space group (No. 215). According to the present inventive concept, without relying on conventional highly oxidative elements, such as oxygen (O), fluorine (F), or iodine (I), a hydride ion conductor having a stable crystal structure and capable of transferring hydride ions was synthesized by simply mixing hydrides of transition metals, hydrides of alkali metals, and metal borohydrides in a specific molar ratio, followed by mechanical milling. The synthesized hydride ion conductor exhibits ionic conductivity more than 100 times higher than that of conventional ion conductors containing no borohydrides and no hydrogen (H) vacancies, making it highly suitable for use as a solid electrolyte in various energy storage devices.

Disclosed herein is a system for generating an anticancer drug candidate compound based on genotype. The system for generating an anticancer drug candidate compound based on genotype of the present invention includes: a processor configured to determine compound information corresponding to a generation condition including genotype information; and a memory, wherein the memory includes: a first encoder configured to determine an encoding vector corresponding to the generation condition; a diffusion model configured to determine a representation vector generated from the encoding vector; and a decoder configured to determine compound information by decoding the generated representation vector.

Disclosed herein is a depth map estimation method including acquiring multiple feature maps corresponding to a given image using a pre-trained depth model, generating a first feature map in a Euclidean space by fusing the multiple feature maps, and generate a second feature map by mapping the first feature map to a hyperbolic space, estimating a bidirectional kernel filter using the first feature map and the second feature map, and generating an initial depth map using the bidirectional kernel filter and an effective depth map, and calculating a hyperbolic affinity map based on a curvature in the hyperbolic space, and correcting the initial depth map based on the hyperbolic affinity map.