Abstract: Methods, systems, and devices, including computer programs encoded on computer storage media, for selecting a target face are provided. One of the methods includes: obtaining at least one facial area including one or more faces in an image taken by a camera; determining, based on the image, a spatial distance between each of the one or more faces and the camera; and selecting, based on the spatial distance, the target face from the one or more faces.

Abstract: A method for recognizing, identifying, and authenticating persons to be identified from their faces in a preset scenario, includes uploading a first image of users to a server, and receiving the first image on which first facial rectangles and user information have been imposed, and capturing a second image and detecting second facial rectangles therefrom. The method determines whether a number of faces of the second image is equal to a number of faces of the first image, and uploading the second image to the server for identifying third facial rectangles of the second image. A plurality of distances is obtained by calculating at least one distance between third and second facial rectangles of each user shown in the second image and determining whether the users shown in the second image are the same as users shown in the first image based on the plurality of distances.

Abstract: A method for configuring a system for recognizing a class of objects of variable morphology, comprising providing a machine learning system with an initial data set to recognize instances of objects of the class in a sequence of images of a target scene; providing a three-dimensional model specific to the class of objects, the morphology of which can be defined by a set of parameters; acquiring a sequence of images of the scene using a camera; recognizing image instances of objects of the class in the acquired image sequence; conforming the generic three-dimensional model to recognized image instances; storing variation ranges of the parameters resulting from the conformations of the generic model; synthesizing multiple three-dimensional objects from the generic model by varying the parameters in the stored variation ranges; and complementing the data set of the learning system with projections of the synthesized objects in the plane of the images.

Abstract: The present invention relates to a device and method for quality assessment of medical image datasets.

Abstract: Examples for determining a confidence level associated with image segmentation are disclosed. A confidence level associated with a collective image segmentation result can be determined by generating multiple individual segmentation results each from the same image data. These examples can then aggregate the individual segmentation results to form the collective image segmentation result and measure the spread of each individual segmentation result from the collective image segmentation result. The measured spread of each individual segmentation result can then be used to determine the confidence level associated with the collective image segmentation result. This can allow a confidence level associated with the collective image segmentation result to be determined. This confidence level may be determined without needing a ground truth to compare to the collective image segmentation result.

Abstract: A semiconductor pattern detecting apparatus is provided.

Abstract: A compression method for high-resolution light field display is disclosed for applications in which computer memory constraints and latency are critical. The disclosed compression algorithm takes advantage of the 3D structure of a light field to compress the raw light field information with a fixed compression ratio and simple decoding instructions. The compressed high-resolution light field achieves a reduced bandwidth with acceptable quality, and is packed in a way that can be transmitted using common interfaces. In a preferred embodiment, the compression algorithm is used as a post-processing stage after light field information acquisition or after rendering. In a further preferred embodiment, the compression algorithm is incorporated into the acquisition or rendering procedure to reduce memory and rendering processing. These and further embodiments generate a compressed light field with all information required for direct display.

Abstract: Techniques are provided for facial recognition using a high probability group database and a facial network of related persons.

Abstract: An image identification apparatus includes an extraction unit, an excluding unit, and an identification unit. The extraction unit extracts lines from an image. The exclusion unit excludes from objects to be identified a boundary delimiting an entire area of the image among the extracted lines. The identification unit identifies as an object multiple lines that are among the extracted lines and that are not excluded by the exclusion unit if the multiple lines are connected to each other.

Abstract: Devices, systems, and methods obtain training images; generate image-alignment data based on the training images; cluster the training images based at least in part on the image-alignment data, thereby generating clusters of training images; and select one or more representative images from the training images based on the clusters of training images.

Abstract: A method and system for matching an unknown facial image of an individual with an image of a celebrity using facial recognition techniques and human perception is disclosed herein. The invention provides a internet hosted system to find, compare, contrast and identify similar characteristics among two or more individuals using a digital camera, cellular telephone camera, wireless device for the purpose of returning information regarding similar faces to the user. The system features classification of unknown facial images from a variety of internet accessible sources, including mobile phones, wireless camera-enabled devices, images obtained from digital cameras or scanners that are uploaded from PCs, third-party applications and databases. Once classified, the matching person's name, image and associated meta-data is sent back to the user. The method and system uses human perception techniques to weight the feature vectors.

Abstract: A method and system for improving an automated facial recognition software system is provided. The method includes automatically detecting a face of a user via an IOT device. An image of the face is retrieved and image portions are extracted from the image and represented as a vector. The user is classified via determined facial feature attributes with respect to a plurality of user type weights stored in a cache and an initial user type of the user is determined. The vector and data indicating the initial user type are transmitted to a server and a process for inferring with respect to the initial user type, the vector, and images in a specified database associated with the initial user type, a final user type of the user s performed. An identity of the user is determined based on the inferring and the identity is transmitted to the IOT device.

Abstract: In a method for determining a course of lanes (L1, L2) of a road for a vehicle, a distance (D1-D3) of at least one lane boundary (B1-B3, N1-N3) of at least one lane (L1, L2) from a predetermined point of the vehicle (V) is determined (S1). Further, a course of the at least one lane boundary (B1-B3, N1-N3) is predicted (S2) based on information about a course of the road obtained from a database (DB) and based on the determined distance (D1-D3).

Abstract: An eye tracking method includes: constructing, by a processing circuit, an eye model; analyzing, by the processing circuit, a first head center position, according to a plurality of first pupil shape information and the eye model, wherein the plurality of first pupil shape information correspond to a plurality of first gazing vectors; capturing, by a camera circuit, a first image of the eye; analyzing, by the processing circuit, a determined gazing vector, according to the eye model and the first image; and adjusting, by the processing circuit, the first head center position according to an actual pupil shape information group and a plurality of simulated pupil shape information groups.

Abstract: In an example embodiment, a processing device receives a 3D image of a face of a person. The processing device identifies a first plurality of visible landmarks and a second plurality of visible landmarks on the face from the 3D image, wherein the second plurality of visible landmarks are associated with soft facial tissues. The processing device determines a post-treatment arrangement of the first plurality of visible landmarks based on a first mapping between a current dentition of the person and a post-treatment dentition of the person. The processing device determines, based on applying the post-treatment arrangement of the first plurality of visible landmarks to a second mapping between the first plurality of visible landmarks and the second plurality of visible landmarks, a post-treatment shape of the soft facial tissues. The processing device generates a modified version of the 3D image, wherein the soft facial tissues have the post-treatment shape in the modified version of the 3D image.

Abstract: Example methods for adaptive radiotherapy treatment planning using deep learning engines are provided. One example method may comprise obtaining treatment image data associated with a first imaging modality. The treatment image data may be acquired during a treatment phase of a patient. Also, planning image data associated with a second imaging modality may be acquired prior to the treatment phase to generate a treatment plan for the patient. The method may also comprise: in response to determination that an update of the treatment plan is required, transforming the treatment image data associated with the first imaging modality to generate transformed image data associated with the second imaging modality. The method may further comprise: processing, using the deep learning engine, the transformed image data to generate output data for updating the treatment plan.

Abstract: An image processing apparatus is provided with a converter to convert a first color image into lightness information and color information other than the lightness information, a lightness compressor to generate compressed lightness information obtained by decimating a pixel value from the lightness information, a reconverter to combine the compressed lightness information and the color information to generate a second color image, an edge extractor to extract an edge included in the first color image, and a pixel value changer to change a pixel value at a position of the edge in the second color image while maintaining similar colors, to generate a third color image.

Abstract: A method for training a deep learning network based on artificial intelligence is provided. The method includes steps of: a learning device (a) inputting unlabeled data into an active learning network to acquire sub unlabeled data and inputting the sub unlabeled data into an auto labeling network to generate new labeled data; (b) allowing a continual learning network to sample the new labeled data and existing labeled data to generate a mini-batch, and train the existing learning network using the mini-batch to acquire a trained learning network, wherein part of the mini-batch are selected by referring to specific existing losses; and (c) (i) allowing an explainable analysis network to generate insightful results on validation data and transmit the insightful results to a human engineer to transmit an analysis of the trained learning network and (ii) modifying at least one of the active learning network and the continual learning network.

Abstract: A color conversion system includes a first colorimetric input machine having a first color gamut and performing colorimetry of an object to output a first colorimetric value; an input portion to which the first colorimetric value is input; a converting portion converting the first colorimetric value to a second colorimetric value; and an output portion outputting the second colorimetric value. The converting portion stores a first colorimetric value table having color information about each measurement light source used when performing colorimetry by the first colorimetric input machine, and a second colorimetric value table having color information about each measurement light source used when performing colorimetry by a second colorimetric device having a second color gamut, and converts the first colorimetric value to the second colorimetric value by causing colorimetric values of the first colorimetric value table to correspond to colorimetric values of the second colorimetric value table.

Abstract: An apparatus and method are provided to correct motion artifacts in magnetic resonance imaging (MRI) data by finding and removing motion-corrupted encodes. The MRI data non-uniformly sample k-space using a series of shots, each including one or more encodes. The motion-corrupted encodes/shots are identified by omitting respective encodes/shots from the MRI data when reconstructing respective images using a compressed-sensing (CS) method. The image quality is improved for those reconstructed images in which the motion-corrupted encodes are omitted, whereas all other images include the motion-corrupted encodes and exhibit the motion artifact. Assuming a minority of encodes/shots are corrupted by motion, the images improved by omitting the motion-corrupted encodes can be identified as outliers. Once, the motion-corrupted encodes are identified and excluded from the final MRI dataset, a final, high-resolution image is reconstructed using the final MRI dataset.