A non-transitory computer-readable recording medium stores a label change program for causing a computer to execute a process including: acquiring image data that includes a plurality of areas; setting a label for each of the plurality of areas by inputting the image data to a first machine learning model; specifying a behavior performed by a person located in a first area among the plurality of areas for an object located in a second area; and changing a label set for the second area based on a specified behavior of the person.

A method of detecting a target object performed by a computing device including at least one processor according to an exemplary embodiment of the present disclosure may include: receiving an input image; and generating first result information related to an area corresponding to a target object from the input image based on a trained neural network-based detection model.

A training engine is described which has a memory arranged to access a neural network image classifier, the neural network image classifier having been trained using a plurality of training images from an input space, the training images being labeled for a plurality of classes. The training engine has an adversarial example generator which computes a plurality of adversarial images by, for each adversarial image, searching a region in the input space around one of the training images, the region being one in which the neural network is linear, to find an image which is incorrectly classified into the plurality of classes by the neural network. The training engine has a processor which further trains the neural network image classifier using at least the adversarial images.

Aspects of the invention relate to methods of authenticating a user and user authentication systems. The method comprises classifying an image of the user as authentic or non-authentic by: identifying a separation vector between a user image characteristic vector and a hyperplane generated by a machine learning algorithm; comparing the separation vector with a threshold value; and associating the user image with a classification value if the separation vector exceeds the threshold value. The user may be authenticated based on a classification decision informed by the classification value associated with the user image.

A system including: (a) a network hub or port adapted to detect image files in transit according to their file designations; (b) an object detector configured to identify one or more regions of interest (ROI) in each image file as potentially containing an object of interest (OOI); (c) a feature analyzer adapted to express one or more General Classification Features (GCF) of each ROI as a vector; and (d) a decision module adapted accept or reject each ROI as containing said OOI based upon the one or more GCF vectors.

Various embodiments set forth systems and techniques for training a landmark model. The techniques include determining, using the landmark model, a first landmark in a set of first landmarks associated with a first image; performing, on the first image, a first perturbation to obtain a second image; determining, using the landmark model, a second landmark in a set of second landmarks associated with the second image; determining, based on a first distance between the first landmark and the second landmark, a first loss function; and updating, based on the first loss function, a first parameter of the landmark model.

According to one embodiment, a method, computer system, and computer program product for object detection. The embodiment may include receiving an annotated image dataset comprising rectangles which surround objects to be detected and labels which specify a class to which an object belongs. The embodiment may include calculating areas of high and low probability of rectangle distribution for each class of objects within images of the dataset. The embodiment may include applying a correction factor to confidence values of object prediction results, obtained during validation of a trained object detection (OD) model, depending on a class label and a rectangle location of an object prediction result and calculating an accuracy of the trained OD model. The embodiment may include increasing the correction factor and re-calculating the accuracy of the trained OD model with every increase. The embodiment may include selecting an optimal correction factor which yields a highest accuracy.

An image processing apparatus includes a generation unit that generates feature data based on an image, classification units that classify a predetermined pattern by referring to the feature data, and a control unit that controls operations of the classification units. The classification units include a first classification unit and a second classification unit, processing results of which do not depend on each other, and a third classification unit. The first and the second classification units are operated in parallel. When either of the first and the second classification units determines that a classification condition is not satisfied, the control unit stops operations of all of the classification units. When both the first and the second classification units determine the classification condition is satisfied, the control unit operates the third classification unit by using classification results of the first and the second classification units.

In an example embodiment, a deep convolutional neural network (DCNN) is created to assign a professionalism score to an input image. The professionalism score indicates a perceived professionalism of a subject of the input image. The DCNN is designed to automatically learn features of images relevant to the professionalism through a training process.

Techniques related to improved video coding based on face detection, region extraction, and tracking are discussed. Such techniques may include performing a facial search of a video frame to determine candidate face regions in the video frame, testing the candidate face regions based on skin tone information to determine valid and invalid face regions, rejecting invalid face regions, and encoding the video frame based on valid face regions to generate a coded bitstream.