Abstract: Provided are a system and method for generating an image implemented in the same style as a style of a single image sample based on a result of learning the single image sample, the system including: an image learning unit configured to train a deep learning network model based on an image sample of a specific style to generate a same style generative model network; and an image generating unit configured to input a layout image to the same style generative model network to generate an image having the same style as the specific style image sample while conforming to a layout structure of the layout image.

Abstract: A neural network device for learning dependency of feature data includes: a memory in which at least one program is stored; and a processor that performs a calculation by executing the at least one program, in which the processor is configured to acquire graph information including a data node for a human body; extract feature data corresponding to a plurality of joints constituting the human body from the graph information; acquire a self-attention output corresponding to the feature data based on a self-attention mechanism; and generate result data for a motion of the human body based on the self-attention output, and the self-attention output includes position information acquired based on positional encoding of the feature data and structural information acquired based on geodesic encoding of the feature data.

Abstract: A method of generating a moving viewpoint motion picture, which is performed by a processor that executes at least one instruction stored in a memory, may comprise: obtaining an input image; generating a trimap from the input image; generating a depth map using the input image; generating a foreground mesh/texture map model based on a foreground alpha map obtained based on the trimap and foreground depth information obtained based on the trimap and the depth map; and generating a moving viewpoint motion picture based on the foreground mesh/texture map model.

Abstract: A method and apparatus for generating a face-harmonized image are disclosed. The method of generating a face-harmonized image includes receiving an input image, extracting facial landmarks from a target image and the input image, generating a face-removed image of the target image based on a facial mask region, extracting a user face image from the input image, transforming the user face image to correspond to the facial mask region, generating a face-blended image by blending the transformed user face image with the target image, extracting a feature map of the face-blended image, generating a combined feature map based on the feature map of the face-blended image and a feature map of the target image, generating a face harmonization result image based on the combined feature map, and providing the generated face harmonization result image.

Abstract: Disclosed herein are an apparatus and method for image outpainting based on learning of a single input image. The apparatus generates multiple extended images from an input image using multiple noise vectors, inputs a first extended image, among the multiple extended images, the input image, and an image acquired by horizontally flipping the input image and thereby trains a discriminator network such that the loss of a generative adversarial network is minimized, and learns the multiple extended images such that the sum of two differences is minimized, the first one being the difference between the input image and the center region of a second extended image, and the second one being the difference between images acquired by vertically segmenting the input image and flipping the segments and the left and right regions of a third extended image.

Abstract: Provided are a method and system for training a dynamic deep neural network. The method for training a dynamic deep neural network includes receiving an output of a last layer of the deep neural network and outputting a first loss, receiving an output of a routing module according to an input class of the deep neural network and outputting a second loss, calculating a third loss based on the first loss and the second loss, and updating a weight of the deep neural network by using the third loss.

Abstract: A data learning device in a deep learning network characterized by a high image resolution and a thin channel at an input stage and an output stage and a low image resolution and a thick channel in an intermediate deep layer includes a feature information extraction unit configured to extract global feature information considering an association between all elements of data when generating an initial estimate in the deep layer; a direct channel-to-image conversion unit configured to generate expanded data having the same resolution as a final output from the generated initial estimate of the global feature information or intermediate outputs sequentially generated in subsequent layers; and a comparison and learning unit configured to calculate a difference between the expanded data generated by the direct channel-to-image conversion unit and a prepared ground truth value and update network parameters such that the difference is decreased.

Abstract: Disclosed herein are a method and apparatus for generating a digital surface model using satellite imagery. The method includes correcting a geometric error in input satellite images, generating one or more depth maps using a stereo matching method based on the corrected satellite images, and generating a digital surface model by combining the generated depth maps and by estimating a height value for each location on the ground surface.

Abstract: Disclosed herein are a method and apparatus for matching 3D terrain information based on aerial images captured at different altitudes.

Abstract: Disclosed is a method of reconstructing a three-dimensional color mesh and an apparatus for the same. According to an embodiment of the present disclosure, the method includes: receiving mesh information of an object, multiple multi-view images obtained by photographing the object at different positions, and camera parameter information corresponding to the multiple multi-view images; constructing a texture map with respect to the object on the basis of the received information and setting a texture patch referring to a color value of the same multi-view image; correcting a color value of a vertex included for each texture patch; and performing rendering with respect to the object by applying the corrected color value of the vertex to the texture map.

Abstract: Disclosed herein is an operating method of an apparatus for generating 3D geographic data. The operating method may include receiving multi-view images, generating a 3D mesh and texture for a first distance view from the multi-view images, generating hybrid Digital Surface Model (DSM) data for a second distance view using the 3D mesh and the texture, and generating DSM data for a third distance view using the 3D mesh and the texture.

Abstract: Disclosed herein are a method and apparatus for matching 3D terrain information based on aerial images captured at different altitudes.

Abstract: Disclosed is a method of performing geometric correction on images obtained by multiple cameras and an image obtainment apparatus therefor, the method including: receiving the images obtained by the multiple cameras; extracting feature points of the received images and checking an interrelation between the extracted feature points; estimating locations and directions of the cameras, which respectively obtain the images, by using the interrelation between the feature points; calculating a relative geometric relationship between the multiple cameras from the estimated locations and directions of the cameras when the received images are determined as the images obtained by the multiple cameras with a fixed geometric interrelationship; and performing the geometric correction on each of the images by using the calculated geometric relationship.

Abstract: Disclosed is a method of reconstructing a three-dimensional color mesh and an apparatus for the same. According to an embodiment of the present disclosure, the method includes: receiving mesh information of an object, multiple multi-view images obtained by photographing the object at different positions, and camera parameter information corresponding to the multiple multi-view images; constructing a texture map with respect to the object on the basis of the received information and setting a texture patch referring to a color value of the same multi-view image; correcting a color value of a vertex included for each texture patch; and performing rendering with respect to the object by applying the corrected color value of the vertex to the texture map.

Abstract: An apparatus for generating a 3-dimensional face model includes a multi-view image capturer configured to sense a motion of the mobile device and automatically capture still images from two or more directions; and a 3D model generator configured to generate a 3D face mode using the two or more still images obtained by the multi-view image capturer.

Abstract: Disclosed herein are an apparatus and method for reconstructing a 3D model. The apparatus for reconstructing a 3D model includes an image acquisition unit for acquiring multi-view images by receiving image signals captured by multiple drones using cameras, a geometric calibration unit for estimating motion variables of the drones based on the acquired multi-view images, and a 3D model creation unit for reconstructing a 3D model of a dynamic object from the matched multi-view images using a multi-view stereo method.

Abstract: Disclosed herein are an apparatus and method for reconstructing a 3D model. The apparatus for reconstructing a 3D model includes an image acquisition unit for acquiring multi-view images by receiving image signals captured by multiple drones using cameras, a geometric calibration unit for estimating motion variables of the drones based on the acquired multi-view images, and a 3D model creation unit for reconstructing a 3D model of a dynamic object from the matched multi-view images using a multi-view stereo method.

Abstract: An apparatus for generating a 3-dimensional face model includes a multi-view image capturer configured to sense a motion of the mobile device and automatically capture still images from two or more directions; and a 3D model generator configured to generate a 3D face mode using the two or more still images obtained by the multi-view image capturer.

Abstract: A 3D multifunctional device in which a 3D printer and a 3D scanner are combined and a method for operating the device. The 3D multifunctional device includes a printing unit that includes a nozzle for ejecting material for 3D printing, a scanning unit that includes a light radiation module for radiating light to an object to be scanned and a camera for capturing an image of the object, a head unit for moving the printing unit and the scanning unit in X-axis and Y-axis directions, a bed plate on which the object to be scanned is located or the ejected material is deposited, the bed plate moving in a Z-axis direction, and a control unit for setting a function by selecting 3D scanning or 3D printing, controlling the scanning unit or the printing unit, and controlling the movements of the head unit and bed plate.

Abstract: An apparatus and method for managing a teeth model. The apparatus for managing a teeth model includes a database for storing a 3D intraoral structure image and a 3D teeth model image generated by scanning a healthy intraoral structure and teeth of a user; a generation unit for generating a damaged tooth model image by scanning a damaged tooth when the tooth of the user is damaged; and a modeling unit for modeling any one of a dental prosthesis and an artificial tooth by comparing the damaged tooth model image and the image stored in the database.