Abstract: An apparatus for retraining an object detector according to an exemplary embodiment includes an inputter configured to receive an undetected image, a style transferer configured to generate one or more first augmented images that have the same content attribute as an object area of the undetected image, but a different style attribute, a content transferer configured to generate one or more second augmented images that have the same style attribute as the object area, but a different content attribute, and an influence analyzer configured to analyze a cause of non-detection of the undetected image by comparing object detection reliabilities of the undetected image, the first augmented image, and the second augmented image.

Abstract: An apparatus for retraining an object detector according to an exemplary embodiment includes an inputter configured to receive an undetected image, a style transferer configured to generate one or more first augmented images that have the same content attribute as an object area of the undetected image, but a different style attribute, a content transferer configured to generate one or more second augmented images that have the same style attribute as the object area, but a different content attribute, and an influence analyzer configured to analyze a cause of non-detection of the undetected image by comparing object detection reliabilities of the undetected image, the first augmented image, and the second augmented image.

Abstract: Provided is a machine learning-based object detection method performed by an object detection apparatus. the method comprises constructing an object detection model by performing machine learning on a training image set, wherein the object detection model is a model for detecting a target object in an input image based on the result of comparing a confidence score for the target object with a threshold value, obtaining an input image given a detection result for the target object, wherein the obtained input image is an image not included in the training image set, predicting one or more object regions, in which the target object exists, in the obtained input image by using the object detection model, classifying a region not matching the detection result among the predicted object regions as a false detection region; and adjusting the threshold value of the object detection model based on a confidence score of the false detection region.

Abstract: Provided is a machine learning-based object detection method performed by an object detection apparatus. the method comprises constructing an object detection model by performing machine learning on a training image set, wherein the object detection model is a model for detecting a target object in an input image based on the result of comparing a confidence score for the target object with a threshold value, obtaining an input image given a detection result for the target object, wherein the obtained input image is an image not included in the training image set, predicting one or more object regions, in which the target object exists, in the obtained input image by using the object detection model, classifying a region not matching the detection result among the predicted object regions as a false detection region; and adjusting the threshold value of the object detection model based on a confidence score of the false detection region.

Abstract: Disclosed is a method for reducing noise. The method comprises decoding a bitstream encoded with a first codec, acquiring Quantization Parameter (QP) values determined in a process of encoding the bitstream with the first codec, detecting a frame with noise by using the QP values and inputting the frame with noise to a predefined noise filter.

Abstract: According to a first aspect of the present invention, a method for setting a region of interest may include: detecting a mark in an image frame photographed by a camera; and setting a region of interest using the detected mark.

Abstract: Disclosed is a method for reducing noise. The method comprises decoding a bitstream encoded with a first codec, acquiring Quantization Parameter (QP) values determined in a process of encoding the bitstream with the first codec, detecting a frame with noise by using the QP values and inputting the frame with noise to a predefined noise filter.

Abstract: According to a first aspect of the present invention, a method for setting a region of interest may include: detecting a mark in an image frame photographed by a camera; and setting a region of interest using the detected mark.

Abstract: A method of creating a texture map is provided. The method includes extracting feature points of a particular object from one or more image frames captured by a camera; selecting one of the image frames as an image frame to be used in the creation of a texture map of the particular object based on information regarding the extracted feature points; and creating the texture map of the particular object using the selected image frame.

Abstract: The present invention relates to a method of creating an animatable digital clone includes receiving input multi-view images of an actor captured by at least two cameras and reconstructing a three-dimensional appearance therefrom, accepting shape information selectively based on a probability of photo-consistency in the input multi-view images obtained from the reconstruction and transferring a mesh topology of a reference human body model onto a shape of the actor obtained from the reconstruction. The method further includes generating an initial human body model of the actor via transfer of the mesh topology utilizing sectional shape information of the actor's joints, and generating a genuine human body model of the actor from learning genuine behavioral characteristics of the actor by applying the initial human body model to multi-view posture learning images where performance of a predefined motion by the actor is recorded.

Abstract: An SOI device includes an SOI substrate having a stacked structure including a buried oxide layer and a first silicon layer sequentially stacked on a silicon substrate. The SOI substrate possesses grooves having a depth that extends from an upper surface of the first silicon layer to a partial depth of the buried oxide layer. An insulation layer is formed on the lower surfaces of the grooves and a second silicon layer is formed filling the grooves having the insulation layer formed thereon. Gates are formed on the second silicon layer and junction regions are formed in the first silicon layer on both sides of the gates to contact the insulation layer.

Abstract: An apparatus for transforming an image in a mobile device is disclosed. The apparatus comprises a pen style transformation unit for transforming an original image to a pen style image using a separated Gaussian filter and outputting pen style image, a cartoon style transformation unit for transforming the original image to a cartoon style image and outputting the cartoon style image, a mosaic creation unit for creating a photomosaic to which the original image is applied based on an optimum photo image set and outputting the created photo mosaic by applying the created photo mosaic to the original image, and an image output unit for outputting the images output from the pen style transformation unit, the cartoon style transformation unit, and the mosaic creation unit to an output unit.

Abstract: An SOI device includes an SOI substrate having a stacked structure including a buried oxide layer and a first silicon layer sequentially stacked on a silicon substrate. The SOI substrate possesses grooves having a depth that extends from an upper surface of the first silicon layer to a partial depth of the buried oxide layer. An insulation layer is formed on the lower surfaces of the grooves and a second silicon layer is formed filling the grooves having the insulation layer formed thereon. Gates are formed on the second silicon layer and junction regions are formed in the first silicon layer on both sides of the gates to contact the insulation layer.

Abstract: An apparatus for producing digital cartoons includes: an image collection unit for collecting photorealistic image data; and a cartooning processor for performing a rendering process including image abstraction and edge generation on the photorealistic image data collected by the image collection unit to generate cartooned images. The apparatus further includes a stylization unit for producing a cartoon page having various shapes of cartoon frames, resizing the cartooned images to place the resized images into the cartoon frames and adding cartoon elements to the images placed in the cartoon frames.

Abstract: An apparatus for processing an effect using style lines, includes a contour line creation unit for creating contour lines using polygons of a three-dimensional (3D) object and location information of a camera; a style line creation unit for putting edge lists, extracted at the time of creating the contour lines, into groups, and creating one or more style lines for each of the groups; and an effect processing unit for representing a line style by inserting the created style lines inside and outside a contour line corresponding to the group. Therefore, representation to appear to have been drawn by a human hand is enabled by adding style lines to existing contour lines in order to represent a line style that belongs to various styles of non-photo realistic rendering.

Abstract: A method for producing cartoon animation using character animation and mesh deformation is provided. The system includes a motion analysis module, a mesh deformation module, a motion deformation module, and a skinning module. The motion analysis module receives existing motion data having information about non-deformed motions of a character, and extracts parameters from the existing motion data by analyzing an animation value that a character's each joint has. The mesh deformation module receives existing mesh data having information about an external appearance of a character and existing skinning data having information for cohering the parameters or bones with mesh, and generates deformed mesh data. The motion deformation module receives the existing motion data and deforms motion using the parameters. The skinning module receives the deformed mesh data, the deformed motion data, and the existing skinning data and generates character animation data having cartoon like motion.

Abstract: An SOI device includes an SOI substrate composed of a stack structure of a silicon substrate, a buried oxide layer, and a silicon layer. Grooves are defined in the silicon layer each exposing the buried oxide layer. A barrier layer is formed on the lower portion of the sidewall of each of the grooves. An epi-silicon layer is formed to fill the grooves and cover the barrier layer. Gates are formed on the epi-silicon layer, and junction areas are formed in the silicon layer on both sides of the gates.

Abstract: The present invention relates to a method of creating an animatable digital clone includes receiving input multi-view images of an actor captured by at least two cameras and reconstructing a three-dimensional appearance therefrom, accepting shape information selectively based on a probability of photo-consistency in the input multi-view images obtained from the reconstruction and transferring a mesh topology of a reference human body model onto a shape of the actor obtained from the reconstruction. The method further includes generating an initial human body model of the actor via transfer of the mesh topology utilizing sectional shape information of the actor's joints, and generating a genuine human body model of the actor from learning genuine behavioral characteristics of the actor by applying the initial human body model to multi-view posture learning images where performance of a predefined motion by the actor is recorded.

Abstract: Provided is an image-based volume data carving method for rapidly carving a specific area of dimensional volume data based on images. The method includes the steps of: generating a mask image to be carved from an input image; dividing a viewing transform matrix of the mask image into a shear transform matrix and a warp transform matrix, and calculating a scale factor from the shear transform matrix; modifying the mask image to be parallel to an axis of the volume data; shearing a volume slice in such a manner that the volume data can be parallel to viewing rays passing through a volume, and scaling the size of the volume slice; and carving part of the volume slice through an operation between the mask image and each volume slice.

Abstract: There are provided a unified framework based on extensible styles for 3D non-photorealistic rendering and a method of configuring the framework. The unified framework includes: 3D model data processing means for generating a scene graph by converting a 3D model input into 3D data and organizing the scene graph using vertexes, faces, and edges; face painting means for selecting a brusher to paint faces (interiors) of the 3D model using the scene graph; line drawing means for extracting line information from the 3D model using the scene graph and managing the extracted line information; style expressing means for generating a rendering style for the 3D model and storing the rendering style as a stroke, the rendering style being equally applied to a face-painting method and a line-drawing method; and rendering means for combining the stroke and the selected brusher to render the 3D model using both the face-painting method and the line-drawing method.