This disclosure describes one or more implementations of systems, non-transitory computer-readable media, and methods that extract multiple attributes from an object portrayed in a digital image utilizing a multi-attribute contrastive classification neural network. For example, the disclosed systems utilize a multi-attribute contrastive classification neural network that includes an embedding neural network, a localizer neural network, a multi-attention neural network, and a classifier neural network. In some cases, the disclosed systems train the multi-attribute contrastive classification neural network utilizing a multi-attribute, supervised-contrastive loss. In some embodiments, the disclosed systems generate negative attribute training labels for labeled digital images utilizing positive attribute labels that correspond to the labeled digital images.

An electronic device includes a camera, a light source unit, a display, and a processor, wherein the processor is configured to obtain a first image of a skin of a user by the camera while a first light is output through the light source unit, obtain a second image of the skin of the user by the camera while the first light is not output through the light source unit, obtain a third image representing a difference between the first image and the second image by converting, into a first difference pixel value, a difference between first brightness values of the first image and second brightness values of the second image, and display, on the display, information on an ultraviolet (UV) protection ingredient applied to the skin of the user, based on the third image.

One variation of a method for segmenting scenes of product units arranged in inventory structures within a store includes: accessing an image based on data captured by a mobile robotic system; detecting a shelving segment in the image; reading a segment identifier from a segment tag, detected in the image, arranged on the shelving segment; accessing a product template representing a product type in the set of product types assigned to the shelving segment based on the segment identifier; detecting a set of product features, in the first region of the image. In response to detecting the set of product features analogous to features of the product template: confirming presence of the unit of the first product type on the shelf in the shelving segment and appending the first product type to a list of product types presently stocked in the shelving segment.

A method and a computing device for clustering phishing web resources based on images of visual content thereof are provided. The method comprises: receiving references to a plurality of phishing web resources; generating, for a given phishing web resource of the plurality of phishing web resources, at least one image of a visual content of the given phishing web resource; analyzing the at least one image associated with the given phishing web resource, the analyzing comprising identifying contours of elements of the visual content of the given phishing web resource within the at least one image; conducting pairwise comparison between the contours associated with the given phishing web resource and contours of stored clusters of visual content images; and storing, in a database, data indicative of an association between the given phishing web resource and a respective cluster of the at least one image.

An information processing device measures a position of the moving object, acquires success or failure information relating to success or failure of the position measurement, acquires at least one of histories, determines to interrupt driving control of the moving object, and decides, if the driving control of the moving object is determined to be interrupted, a restart method. The histories include a history of the position of the moving object, a driving history of the moving object, a history of illuminance around the moving object, and a history of a direction of an imaging element mounted in the moving object. The driving control of the moving object is interrupted on the basis of the success or failure information. The restart method restarts the driving control of the moving object on the basis of at least one of the histories.

Methods and computer program products for training a neural network perform multiple forms of data augmentation on sample waveforms of a training dataset that includes both normal and abnormal samples to generate normal data augmentation samples and abnormal data augmentation samples. The normal data augmentation samples are labeled according to a type of data augmentation that was performed on each respective normal data augmentation sample. The abnormal data augmentation samples are labeled according to a type of data augmentation other than that which was performed on each respective abnormal data augmentation sample. A neural network model is trained to identify a form of data augmentation that has been performed on a waveform using the normal data augmentation samples and the abnormal data augmentation samples.

Apparatuses, systems, and techniques generate poses of an object based on image data of the object obtained from a first viewpoint of the object and a second viewpoint of the object. The poses can be evaluated to determine a portion of the image data usable by an estimator to generate a pose of the object.

There are provided a method and a system for image segmentation utilizing a GAN architecture. A method for training an image segmentation network according to an embodiment includes: inputting an image to a first network which is trained to output a region segmentation result regarding an input image, and generating a region segmentation result; and inputting the region segmentation result generated at the generation step and a ground truth (GT) to a second network, and acquiring a discrimination result, the second network being trained to discriminate inputted region segmentation results as a result generated by the first network and a GT, respectively; and training the first network and the second network by using the discrimination result. Accordingly, region segmentation performance of a semantic segmentation network regarding various images can be enhanced, and a very small image region can be exactly segmented.

Apparatuses, systems, and techniques generate poses of an object based on data of the object observed from a first viewpoint and a second viewpoint. The poses can be evaluated to determine a portion of the data usable by an estimator to generate a pose of the object.

Embodiments of the present invention provide methods, systems, apparatuses, and computer program products for predicting property damage restoration estimates. In one embodiment, a computing entity or apparatus is configured to receive, from a client device, a property damage restoration estimate request comprising one or more digital image files; retrieve policy data associated with a user of the client device, the policy data comprising user identification properties and policy properties; programmatically generate, by fraud detection/prediction circuitry and based on the one or more digital image files, a first predictive value, wherein the first predictive value represents a likelihood that at least one of the digital image files was fraudulently altered; upon identifying that the first predictive value does not exceed a fraud threshold, programmatically generate, by property restoration estimate prediction circuitry and based on the one or more digital image files, a second predictive value, wherein the second predictive value represents a property damage restoration estimate, wherein the second predictive value is based at least on the property properties contained in the policy data and the one or more digital image files; and substantially instantaneously transmit a property damage restoration estimate response comprising the property damage restoration estimate to the client device.