Abstract: The subject technology trains a neural network based on a training process. The subject technology selects a frame from an input video, the selected frame comprising image data including a representation of a face and hair, the representation of the hair being masked. The subject technology determines a previous predicted frame. The subject technology concatenates the selected frame and the previous predicted frame to generate a concatenated frame, the concatenated frame being provided to the neural network. The subject technology generates, using the neural network, a set of outputs including an output tensor, warping field, and a soft mask. The subject technology performs, using a warping field, a warp of the selected frame and the output tensor. The subject technology generates a prediction corresponding to a corrected texture rendering of the selected frame.

Abstract: A method for processing images of an appliance includes capturing, via an imaging device, an image of a cavity of the appliance. The method further includes identifying a support member in the cavity based on the image via a processor in communication with the imaging device. The support member is configured to support an object thereon. The method further includes calculating a histogram of gradients based on the image via the processor. The method further includes receiving dimensional data of the appliance. The method further includes comparing the histogram of gradients to the dimensional data. The method further includes determining a position of the support member based on the comparison of the histogram of gradients to the dimensional data.

Abstract: A fire protection system according to an embodiment of the inventive concept includes a plurality of sensors having different address values, detecting fire occurrence, generating a fire alarm, and performing Radio Frequency (RF) communication with each other, a first server configured to perform RF communication with each of the plurality of sensors, and a second server configured to receive big data from the outside and communicate with the first server, wherein the second server includes a data collection unit configured to collect sensor data measured from each of the plurality of sensors and the big data, a data extraction unit configured to extract fire data based on information collected by the data collection unit, and a complex event processing unit configured to process complex events based on the fire data.

Abstract: Techniques are described herein for determining an active traffic light signal. The techniques involve, determining active signals for first and second traffic light heads of a group of traffic light heads, wherein members of the group of traffic light heads share an active signal at a point in time. A first vote is allocated for the active signal for the first traffic light head indicating a signal state of the first traffic light head and a second vote is allocated for the active signal for the second traffic light head indicating a signal state of the second traffic light head. A signal indicated by the group of traffic light heads is determined based at least in part on the first and second votes.

Abstract: A vehicle interior monitoring system monitors the behavior of a passenger, based on the image of the interior of the vehicle, captured on a camera. A masking unit of the processor masks a portion of the image captured on the camera, the portion corresponding to the outside of the vehicle.

Abstract: An automated method includes detecting a tooth of the scan data using a first artificial intelligence neural network, extracting a tooth scan data from the scan data based on a result of a tooth detection, generating a tooth mapped data corresponding to a predetermined space based on the tooth scan data, generating the tooth boundary curve by inputting the tooth mapped data to a second artificial intelligence neural network and mapping the tooth boundary curve to the scan data.

Abstract: An object detection device for detecting an object includes a light receiving section, an image acquisition part, and a malfunction detection part. The light receiving section has a light receiving surface in which light receiving elements capable of receiving incident light including reflected light of emitted radiated light are arranged in a plane, and outputs a first light receiving signal according to a light receiving state of the light receiving element in a first area, when a predetermined usable area in the usable areas on the light receiving surface is referred to as a first area. The image acquisition part acquires an ambient light image indicating a light receiving intensity of ambient light using the first light receiving signal. A malfunction detection part detects a malfunction of the light receiving element in the first area using the ambient light image.

Abstract: A method for point cloud compression of an intelligent cooperative perception (iCOOPER) for autonomous air vehicles (AAVs) includes: receiving a sequence of consecutive point clouds; identifying a key point cloud (K-frame) from the sequence of consecutive point clouds; transforming each of the other consecutive point clouds (P-frames) to have the same coordinate system as the K-frame; converting each of the K-frame and P-frames into a corresponding range image; spatially encoding the range image of the K-frame by fitting planes; and temporally encoding each of the range images of the P-frames using the fitting planes.

Abstract: An image processing method includes: acquiring a standard image matched with an image processing model; determining light field variation parameters of a use environment of the image processing model based on the standard image; determining an image training sample set based on the light field variation parameters, the image training sample set including images of different light field backgrounds; training a correction network of the image processing model through the image training sample set to obtain a model updating parameter adapted to the correction network, and generating a trained image processing model based on the model updating parameter; acquiring a single image in the use environment, and performing prediction processing on the single image through the trained image processing model to obtain light field information corresponding to the single image; and correcting a light field of the single image based on the light field information.

Abstract: Provided are method and system of calculating cross-section area and included angle of three-dimensional blood vessel branch, comprising: obtaining the contours of a main blood vessel and a branch blood vessel; determining connecting points of the main blood vessel and the branch blood vessel, a three-dimensional branch outer contour, and an initial normal vector of a branch section; determining the position of the branch section at the branch connecting part and an intersecting cross-section of the branch section and the branch outer contour; rotating the three-dimensional branch cross-section contour to obtain an equivalent two-dimensional branch cross-section contour, and calculating the area of the two-dimensional branch cross-section contour; obtaining branch cross-section areas and corresponding normal vectors thereof; and taking the minimum branch cross-section area as the cross-section area of this segment of branch blood vessel, and determining the included angle between the branch blood vessel and t

Abstract: Systems and methods for a vision-based machine learning model for autonomous driving with adjustable virtual camera. An example method includes obtaining images from a multitude of image sensors positioned about a vehicle. Features associated with the images are determined, with the features being output based on a forward pass through a first portion of a machine learning model. The features are projected into a vector space associated with a virtual camera at a particular height. The projected features are aggregated with other projected features associated with prior images. A plurality of objects which are positioned according to the virtual camera are determined.

Abstract: Techniques are generally described for user-customized computer vision event detection. A camera device may capture a first frame of image data. A first machine learning model may generate first embedding data for the first frame of image data. The first embedding data may be input into a second machine learning model associated with the camera device. The second machine learning model and the first embedding data may be used to determine that an area-of-interest represented by the first frame of image data is in a first state. State data stored in memory may be determined. The state data may indicate that the area-of-interest was previously in a second state. An alert may be sent to a device associated with the camera device indicating that the area-of-interest has changed from the second state to the first state.

Abstract: Systems and methods for brightfield inspection of a circular rotating wafer are provided. A method includes: acquiring a plurality of images of a circular wafer, that is rotating, by using a plurality of line cameras; obtaining a plurality of synchronized images, based on the plurality of images, by synchronizing a motion of the circular wafer, that is rotating, with at least one line camera from among the plurality of line cameras; obtaining a single wafer map by integrating together the plurality of synchronized images; obtaining an in-focus image of the single wafer map while the circular wafer is moving; and performing brightfield inspection of the circular wafer based on the in-focus image of the single wafer map.

Abstract: Provided is a marker detection apparatus able to detect the position of a marker with high accuracy and a robot teaching system using the same.

Abstract: An object detection model training apparatus, method and non-transitory computer readable storage medium thereof are provided. The apparatus performs a first object detection on a plurality of training images to generate a piece of first label information corresponding to each of the training images by a first teacher model. The apparatus trains a student model based on the training images and the first label information. The apparatus performs a second object detection on the training images to generate a piece of second label information corresponding to each of the training images by a second teacher model. The apparatus trains the student model based on the training images and the second label information.

Abstract: A method and device for 3D object detection. The method comprises the steps of: generating one or more fused images(201) based on a pair of images(530), the pair of images (530) including a left view image(101,710a) and a right view image(102,710b); extracting one or more fused features from the fused images(201) by a single backbone network(210) of a share network with feature unmixing (SNFU,540); unmixing the fused features to a left view-aware feature and a right view-aware feature by a feature unmixing sub-network(220) of the SNFU(540); predicting the 3D object based on the left view-aware feature and the right view-aware feature; and determining spatial features of the predicted 3D object. The proposed method and device can reduce the computation complexity to the level of that in monocular based 2D object detection so as to improve the computation efficiency while keeping high precision.

Abstract: Examples described herein relate to rendering an image using a reduced fidelity of shading for portions of a triangle that are potentially occluded by other objects. A computer device may pre-populate a potentially occluded depth map in a graphics processing unit (GPU) for a current frame based on depth data of a previous frame, a low resolution depth pre-pass, or a rasterizing occlusion geometry. The GPU may determine, at a rasterization stage, that a depth of a portion of a triangle being processed is further than a depth of a corresponding tile in the potentially occluded depth map. The GPU may decrease a fidelity of shading applied to at least the portion of the triangle in comparison to a specified fidelity for the triangle in response to determining that the depth of the portion of the triangle is further than the depth of the corresponding tile.