Abstract: An apparatus for text-to-image generation which is a self-supervised based on one-stage generative adversarial network and uses a discriminator network that extracts an image feature may comprise: a text encoder that extracts a sentence vector from input text; a discriminator that determines whether or not an image matches the text from the sentence vector and the image input from a generator; and a decoder that is connected to an encoder inside the discriminator, wherein the decoder and the encoder form an autoencoder structure inside the discriminator.

Abstract: Disclosed is a method of tracking multiple objects included in an image using a multi-object tracking apparatus including an integrated similarity neural network, the method including setting a tracking area in an input image, extracting at least one object candidate for a target object from the tracking area; extracting reference features for the target object, the object candidate, and the tracking area, selecting two of the target object, the object candidate, and the tracking area to evaluate similarity based on the reference features; allocating the object candidate to the target object on the basis of the evaluated similarity; and tracking the target object on the basis of a location of the allocated object candidate.

Abstract: A method of detecting an object using an object boundary localization uncertainty aware network and an object boundary localization uncertainty attention module may comprise: calculating a convolutional feature from a single image through a convolutional neural network of an existing object detection neural network and then learning a classification and location of an object to detect the object; inputting an obtained localization value and a ground truth value of the object to the object boundary localization uncertainty aware network to learn an object boundary localization uncertainty; calculating, by the object boundary localization uncertainty attention module, an object boundary feature using the obtained object boundary localization uncertainty and training an object detection correction neural network using the object boundary feature; and aggregating an existing object detection result and a correction result to calculate a final object detection result.

Abstract: An early earthquake detection method may comprise acquiring a frame image from a camera; acquiring a vibration signal from the frame image; removing a noise signal due to vibration of the camera from the vibration signal; acquiring a motion signal obtained by magnifying subtle motions from the noise signal-removed vibration signal; extracting vibration characteristics from the motion signal; estimating an occurrence of an earthquake by extracting a peak signal from the vibration characteristics; and determining whether an earthquake occurs by receiving earthquake estimation information from at least one other camera located within a certain range.

Abstract: The present disclosure provides an action recognition method including: acquiring video features for input videos; generating a bounding box surrounding a person who may be a target for an action recognition; pooling the video features based on bounding box information; extracting at least one spatial feature map from pooled video features; extracting at least one temporal feature map from pooled video features; concatenating the at least one spatial feature map and the at least one temporal feature map to generate a concatenated feature map; and performing a human action recognition based on the concatenated feature map.

Abstract: An early earthquake detection method may comprise acquiring a frame image from a camera; acquiring a vibration signal from the frame image; removing a noise signal due to vibration of the camera from the vibration signal; acquiring a motion signal obtained by magnifying subtle motions from the noise signal-removed vibration signal; extracting vibration characteristics from the motion signal; estimating an occurrence of an earthquake by extracting a peak signal from the vibration characteristics; and determining whether an earthquake occurs by receiving earthquake estimation information from at least one other camera located within a certain range.

Abstract: Disclosed is a method of tracking multiple objects included in an image using a multi-object tracking apparatus including an integrated similarity neural network, the method including setting a tracking area in an input image, extracting at least one object candidate for a target object from the tracking area; extracting reference features for the target object, the object candidate, and the tracking area, selecting two of the target object, the object candidate, and the tracking area to evaluate similarity based on the reference features; allocating the object candidate to the target object on the basis of the evaluated similarity; and tracking the target object on the basis of a location of the allocated object candidate.

Abstract: Disclosed are a method of generating a hierarchical structured pattern based descriptor and a method and a device for recognizing an object in an image using the same. The method of generating a hierarchical structured pattern based descriptor may include generating a hierarchical structured pattern by defining a parent node based on a patch region for a feature point of an input image to be analyzed and defining a child node obtained by dividing the parent node to a predetermined depth, calculating a master direction vector of the patch region based on position coordinates and representative pixel values of the parent node and the child node, and calculating a rotation angle of the patch region based on the master direction vector and rotating the hierarchical structured pattern by the rotation angle.

Abstract: A method and mobile terminal for correcting a gaze of a user in an image are provided. The method includes detecting first eye outer points that define a first eye region of the user in an original image; setting second eye outer points different from the first eye outer points; and transforming the first eye region inside of the first eye outer points to a second eye region inside of the second eye outer points.

Abstract: A method of extracting ridge data includes selecting a point in the contour of an object from an image of the object, calculating a chamfer distance of the point in the contour, and determining whether the point in the contour is a ridge point based on relationships between the chamfer distance and chamfer distances of points in the contour adjacent to the point in the contour. A method of tracking a joint motion of an object includes acquiring initial information of the object including information on at least one ridge region representing a distribution of ridge points of the object, and tracking the joint motion of the object from temporally consecutive images of the object based on the initial information.

Abstract: Disclosed are object recognizing method and apparatus. The method may comprise extracting a feature point of an object to be recognized in an input image; generating a plurality of feature images by extracting orientation information of a region adjacent to the feature point and normalizing the region based in the orientation information; deriving a covariance descriptor based on the feature images; and recognizing the object in the input image by comparing the covariance descriptor with a reference covariance descriptor stored in advance. Thus, the object can be recognized fast and robustly to changes on the size of the input image, rotations of the input image, and illumination, whereby the recognition performance can be enhanced.

Abstract: A method and mobile terminal for correcting a gaze of a user in an image are provided. The method includes detecting first eye outer points that define a first eye region of the user in an original image; setting second eye outer points different from the first eye outer points; and transforming the first eye region inside of the first eye outer points to a second eye region inside of the second eye outer points.

Abstract: Disclosed are a method of generating a hierarchical structured pattern based descriptor and a method and a device for recognizing an object in an image using the same. The method of generating a hierarchical structured pattern based descriptor may include generating a hierarchical structured pattern by defining a parent node based on a patch region for a feature point of an input image to be analyzed and defining a child node obtained by dividing the parent node to a predetermined depth, calculating a master direction vector of the patch region based on position coordinates and representative pixel values of the parent node and the child node, and calculating a rotation angle of the patch region based on the master direction vector and rotating the hierarchical structured pattern by the rotation angle.

Abstract: A method and mobile terminal for correcting a gaze of a user in an image includes setting eye outer points that define an eye region of the user in an original image, transforming the set eye outer points to a predetermined reference camera gaze direction, and transforming the eye region of the original image based on the transformed eye outer points.

Abstract: Provided are a method of extracting ridge data in a method of tracking motion of an object and an apparatus and method for tracking a joint motion of an object. The method of extracting ridge data includes selecting a point in the contour of an object from an image of the object, calculating a chamfer distance of the point in the contour, and determining whether the point in the contour is a ridge point based on relationships between the chamfer distance and chamfer distances of points in the contour adjacent to the point in the contour. The method of tracking a joint motion of an object includes acquiring initial information of the object including information on at least one ridge region representing a distribution of ridge points of the object, and tracking the joint motion of the object from temporally consecutive images of the object based on the initial information.

Abstract: In a method of detecting a specific object using a multi-dimensional image including the specific object, with respect to each window slide of the image subjected to window sliding by applying a previously generated 3D cube filter, data of an area corresponding to the window sliding is normalized in a previously defined specific form. After the corresponding part of the normalized data is assigned to each cell in the 3D cube filter, a volume of the cell is then calculated, thereby expressing the volumes of the cells as one volumetric feature vector having a volumetric feature. The volumetric feature vector is applied to a classifier so as to decide whether or not the data of the area corresponding to the window slide corresponds to the specific object.

Abstract: Disclosed are object recognizing method and apparatus. The method may comprise extracting a feature point of an object to be recognized in an input image; generating a plurality of feature images by extracting orientation information of a region adjacent to the feature point and normalizing the region based in the orientation information; deriving a covariance descriptor based on the feature images; and recognizing the object in the input image by comparing the covariance descriptor with a reference covariance descriptor stored in advance. Thus, the object can be recognized fast and robustly to changes on the size of the input image, rotations of the input image, and illumination, whereby the recognition performance can be enhanced.

Abstract: Disclosed herein is a real-time face recognition apparatus and method. A real-time face recognition apparatus includes a face detection unit for detecting a face image by obtaining image coordinates of a face from an input image. An eye detection unit obtains image coordinates of both eyes in the face image. A facial feature extraction unit generates feature histogram data based on parallel processing from the face image. A DB unit stores predetermined comparative feature histograms. A histogram matching unit compares the histogram data generated by the facial feature extraction unit with the comparative feature histograms, and then outputting similarities of face images. The face recognition apparatus may be implemented as internal hardware in which a VGA camera and an exclusive chip interface with each other, thus remarkably reducing a system size and installation cost, and performing face recognition in real time without requiring additional equipment.

Abstract: An image surveillance system and a method of detecting whether an object is left behind or taken away are provided. The image surveillance system includes: a foreground detecting unit which detects a foreground region based on a pixel information difference between a background image and a current input image; a still region detecting unit which detects a candidate still region by clustering foreground pixels of the foreground region, and determines whether the candidate still region is a falsely detected still region or a true still region; and an object detecting unit which determines whether an object is left behind or taken away, based on edge information about the true still region.

Abstract: A method and mobile terminal for correcting a gaze of a user in an image includes setting eye outer points that define an eye region of the user in an original image, transforming the set eye outer points to a predetermined reference camera gaze direction, and transforming the eye region of the original image based on the transformed eye outer points.