Abstract: An image processing apparatus acquires a color image including a plurality of primary color signals, performs color space conversion processing on the color image to generate a brightness signal image and a color difference signal image, performs super-resolution processing on the brightness signal image and the color difference signal image to generate a prediction brightness signal image, and uses the prediction brightness signal image to generate a super-resolution color image in which a resolution of the color image is increased.

Abstract: A position and pose determination unit (103) is configured to determine a position and an pose of the video acquisition unit (110) in such a way that a level of hiding caused by overlapping of objects becomes low based on environmental object information indicating an environmental object including a structural object present in the monitored area and a placed object placed in the structural object, and staying characteristic information indicating a staying characteristic of the object determined depending on the environmental object. The analysis and display unit (111) is configured to perform at least one of analyzing a video captured by the video acquisition unit (103) at the position and the pose determined by the position and pose determination unit (110) and displaying the video.

Abstract: A computer-implemented method for automatically detecting events and creating photo-product designs based on the events in a photo-product design system includes automatically identifying an event by an event detection module based on daily numbers of captured photos over a plurality of days, automatically selecting a photo-product type by an intelligent product design creation engine in the photo-product design system, calculating a daily weight for a photo product design in the photo-product type based on the daily numbers of captured photos, automatically determining a number of product photos allocated to each day based on associated daily weight, automatically selecting product photos from the captured photos each day at the event according to the number of product photos allocated to each day, and automatically creating a photo-product design for the event using the selected product photos.

Abstract: Devices, methods, and non-transitory computer readable media are disclosed herein to repair or mitigate the appearance of unwanted reflection artifacts in captured video image streams. These unwanted reflection artifacts often present themselves as brightly-colored spots, circles, rings, or halos that reflect the shape of a bright light source in the captured image. These artifacts, also referred to herein as “ghosts” or “green ghosts” (due to often having a greenish tint), are typically located in regions of the captured images where there is not actually a bright light source located in the image. In fact, such unwanted reflection artifacts often present themselves on the image sensor across the principal point of the lens from where the actual bright light source in the captured image is located. Such devices, methods and computer readable media may be configured to detect, track, and repair such unwanted reflection artifacts in an intelligent and efficient fashion.

Abstract: A system for detecting the presence of pests. In one aspect the system comprises an imaging system including: a housing including one or more pest entrances; a pest attractant arranged within the housing; a pest detection surface; an image capture device, said image capture device being configured to capture one or more images of the pest detection surface; and a triggering sensor arrangement for detecting a target pest that is on or about to enter the pest detection surface, said triggering sensor arrangement being configured to provide a trigger-signal if the presence of an object is detected, the image capture device being configured to take an image of the pest detection surface in response to the trigger-signal.

Abstract: A position and pose determination unit (103) is configured to determine a position and an pose of the video acquisition unit (110) in such a way that a level of hiding caused by overlapping of objects becomes low based on environmental object information indicating an environmental object including a structural object present in the monitored area and a placed object placed in the structural object, and staying characteristic information indicating a staying characteristic of the object determined depending on the environmental object. The analysis and display unit (111) is configured to perform at least one of analyzing a video captured by the video acquisition unit (103) at the position and the pose determined by the position and pose determination unit (110) and displaying the video.

Abstract: A position and pose determination unit (103) is configured to determine a position and an pose of the video acquisition unit (110) in such a way that a level of hiding caused by overlapping of objects becomes low based on environmental object information indicating an environmental object including a structural object present in the monitored area and a placed object placed in the structural object, and staying characteristic information indicating a staying characteristic of the object determined depending on the environmental object. The analysis and display unit (111) is configured to perform at least one of analyzing a video captured by the video acquisition unit (103) at the position and the pose determined by the position and pose determination unit (110) and displaying the video.

Abstract: A method includes receiving a worksite plan to be executed by at least one machine at a worksite from a computing device of a supervising entity, the controller being located at a non-line-of-sight (NLOS) location with respect to the worksite. The worksite plan may include a boundary of the worksite at which the worksite plan is implemented, at least one task defining the worksite plan, and a selection of at least one machine to perform the task. The method may include receiving a validation signal from a device located at the worksite, the validation indicating that the worksite is ready for implementation of the worksite plan based on at least one parameter of worksite readiness. The method may include selecting a first mode of operation of the machine to perform the task and transmitting first instructions to the machine to perform the task based on the first mode of operation.

Abstract: An example apparatus for mining multi-scale hard examples includes a convolutional neural network to receive a mini-batch of sample candidates and generate basic feature maps. The apparatus also includes a feature extractor and combiner to generate concatenated feature maps based on the basic feature maps and extract the concatenated feature maps for each of a plurality of received candidate boxes. The apparatus further includes a sample scorer and miner to score the candidate samples with multi-task loss scores and select candidate samples with multi-task loss scores exceeding a threshold score.

Abstract: An apparatus includes an acquisition unit configured to acquire captured data, a first detection unit configured to detect a first area of an object from the data, a second detection unit configured to detect a second area of the object from the data, and an output unit configured to output a result of detection by the first detection unit and a result of detection by the second detection unit, wherein, in a case where the first detection unit detects the first area in first data, the second detection unit detects the second area in an area corresponding to the first area.

Abstract: A computer-implemented method includes gathering data samples into a data set, correcting for imbalance in the data set to produce a corrected data set by applying active learning to the data set to increase a number of double barreled question data samples occurring in the data set, selecting an optimal machine learning model for the corrected data set, training the optimal machine learning model using the corrected data set, operating the optimal machine learning model on new data to produce a prediction result, and generating a visual representation of at least one prediction results.

Abstract: A position and pose determination unit (103) is configured to determine a position and an pose of the video acquisition unit (110) in such a way that a level of hiding caused by overlapping of objects becomes low based on environmental object information indicating an environmental object including a structural object present in the monitored area and a placed object placed in the structural object, and staying characteristic information indicating a staying characteristic of the object determined depending on the environmental object. The analysis and display unit (111) is configured to perform at least one of analyzing a video captured by the video acquisition unit (103) at the position and the pose determined by the position and pose determination unit (110) and displaying the video.

Abstract: A computer-implemented method of interactive machine learning in which a user is provided with predicted results from a trained machine learning model. The user can take the predicted results and adjust the predicted data to retrain the model.

Abstract: A method for extracting spectral information of a substance under test includes: identifying a pixel region A(x, y) occupied by an object under test from a hyperspectral image acquired; extracting a specular reflection region Aq and a diffuse reflection region Ar from the pixel region A(x, y), and calculating a representative spectrum Iq(?) of the specular reflection region Aq and a representative spectrum Ir(?) of the diffuse reflection region Ar, respectively; by comparing each element in the representative spectrum Iq(?) of the specular reflection region Aq with each element in the representative spectrum Ir(?) of the diffuse reflection region Ar, separating information of a light source from spectral information of the object to obtain a first spectral invariant C(?). This method does not require additional spectral information of the light source, which improves the analysis efficiency.

Abstract: Systems, devices, and methods are disclosed for generating an authenticity score for a creative work using digital image data of one or more creative works. A system receives, via a network, first digital image data of a first creative work. The system determines from the first digital image data, a first set of feature variables, each corresponding to a characteristic of the first creative work. The system determines, via an artificial intelligence model and with input including the first set of feature variables and a comparison dataset, an authenticity score for the first creative work. The artificial intelligence model generates, based on the portion of first feature variables of the first creative work that match corresponding portions of the comparison dataset, an output indicative of an authenticity score. The system communicates, via the network, an authenticity score of the first creative work to a client device.

Abstract: The present embodiments relate to automated memory creation and retrieval from moment content items. In some implementations, the automated memory creation and retrieval system can obtain moment content items from user-designated sources with a single user perspective or multiple user perspectives. The moment content items can be assigned tags and arranged in chronological order. The arranged moment content items can be clustered into memory content items based on clustering conditions. Once memory content items are created, they can be arranged into a memory hierarchy made up of connected nodes. In some implementations, the memory content items are also linked together based on similarity in various dimensions in a memory graph. The automated memory creation and retrieval system can receive search criteria for memories from a user interface and provide the user with memories from matched nodes in the memory hierarchy or linked memories in the memory graph.

Abstract: An improved apparatus and method for enhancing an image, and more particularly an apparatus and method for enhancing an image through cross-talk cancellation in a multiple charged-particle beam inspection are disclosed. An improved method for enhancing an image includes acquiring a first image signal of a plurality of image signals from a detector of a multi-beam inspection system. The first image signal corresponds to a detected signal from a first region of the detector on which electrons of a first secondary electron beam and of a second secondary electron beam are incident. The method includes reducing, from the first image signal, cross-talk contamination originating from the second secondary electron beam using a relationship between the first image signal and beam intensities associated with the first secondary electron beam and the second secondary electron beam. The method further includes generating a first image corresponding to first secondary electron beam after reduction.

Abstract: The present invention relates to a method for processing a plurality of candidate annotations of a given instance of an image, each candidate annotation being defined as a closed shape matching the instance, characterized in that the method comprises performing, by a processing unit (21) of a server (2), steps of: (a) segregating said candidate annotations into a set of separate groups of at least overlapping candidate annotations; (b) selecting a subset of said groups as a function of the number of candidate annotations in each group; (c) building a final annotation of the given instance of said image as a combination of regions of the candidate annotations of said selected groups where at least a second predetermined number of the candidate annotations of said selected groups overlap.

Abstract: Methods, systems, and apparatus, including computer programs encoded on computer storage media, for analyzing images for generating query responses. One of the methods includes determining, using a textual query, an image category for images responsive to the textual query, and an output type that identifies a type of requested content; selecting, using data that associates a plurality of images with a corresponding category, a subset of the images that each belong to the image category, each image in the plurality of images belonging to one of the two or more categories; analyzing, using the textual query, data for the images in the subset of the images to determine images responsive to the textual query; determining a response to the textual query using the images responsive to the textual query; and providing, using the output type, the response to the textual query for presentation.

Abstract: Pallet detection device that acquires point cloud data indicating a point cloud measured by a two-dimensional distance measurement device on a depth map; detects a straight line corresponding to a front surface of a pallet based on the point cloud in a region presumed to include the front surface of the pallet in the point cloud data; detects a line segment indicating the front surface of the pallet based on the straight line; acquires position and orientation of the pallet based on the line segment. Acquires one or more straight line candidates and assigns for each of the one or more straight line candidates a score to the point cloud and selects the straight line from the one or more straight line candidates based on a score accumulated value obtained by accumulating the score for the point cloud in the region presumed to include the front surface of the pallet.

Abstract: A method for ground penetrating radar analysis of a scenario, having a surface, with respect to a cartesian reference system S(x,y) having origin O(0,0) and axes x and y. The method comprises a step of prearranging a gpr apparatus that comprises at least one gpr sensor, a control unit, a centre of reference C(xc,yc) having coordinates xc and yc with respect to the cartesian reference system S(x,y), at least one image acquisition devices, each image acquisition devices having a pointing direction ?j known with respect to the centre of reference C(xc,yc). The method comprises then the steps of handling the gpr apparatus on the surface and detecting possible underground objects by means of gpr technique.

Abstract: Provided is a device for providing motion effects, the device including a first detector configured to detect a movement of a specific object in a video of a content; a second detector configured to detect a sound event from an audio of the content; and a motion synthesizing unit configured to generate a synthesized motion by synthesizing a first motion corresponding to the movement of the specific object and a second motion generated based on the sound event.

Abstract: The disclosure provides a vehicle, a machine vision system, and a robot. In one example, the vehicle includes one or more processing units to implement a machine-learning model and to identify one or more objects depicted in one or more images using the machine-learning model, the machine-learning model being trained using a training dataset that includes a plurality of images procedurally synthesized according to one or more training image definitions.

Abstract: A discrete and safe automated insect monitoring system includes a housing, an interior chamber within the housing, and a light source arranged within the housing to illuminate at least a portion of a floor surface of the interior chamber. A multi-pixel optical sensor is arranged within the housing so that a field of view of the sensor comprehends a substantial portion of the floor surface. A processing circuit arranged within the housing receives optical data from the multi-pixel optical sensor, analyzes the optical data to detect the intrusion of an insect or other object into the interior chamber by comparing most recently received optical data to previously received optical data, and generates an indication in response to detecting the intrusion of an insect or other object. Detection and/or classification results can be wirelessly forwarded to another device, to alert appropriate personnel.

Abstract: A learning device is configured to acquire first training data that is training data regarding a registered category object and second training data regarding an unregistered category object, the registered category object belonging to a registered category which is registered as an identification target, the unregistered category object not belonging to the registered category. The learning device is also configured to calculate, on a basis of the first training data and the second training data, an identification score indicating a degree of certainty that an object to be identified belongs to the registered category. The learning device is also configured to calculate, on a basis of the second training data, an unregistered score indicating a degree of certainty that the object to be identified does not belong to the registered category.

Abstract: The present disclosure relates to a video retrieval method, a method, system and device for generating a video retrieval mapping relationship, and a storage medium. The video retrieval method comprises: acquiring a retrieval instruction, wherein the retrieval instruction carries retrieval information for retrieving a target frame picture; and obtaining the target frame picture according to the retrieval information and a preset mapping relationship.

Abstract: There are provided a medical image processing apparatus, a medical image processing system, a medical image processing method, and a program that allow an observer to grasp a region that contributes to classification in automatic classification of a medical image. A medical image processing apparatus includes an image acquisition unit (40) that acquires a captured image (38) generated through imaging of a living body; a classification unit (48) that classifies the captured image into two or more classes; an image generation unit (50) that generates a region image depicting a location of a region that contributes to classification performed using the classification unit in the captured image subjected to the classification performed using the classification unit; and a display signal transmission unit (44) that transmits, to a display device, a display signal representing the captured image, a classification result derived using the classification unit, and the region image.

Abstract: This disclosure describes methods, non-transitory computer readable storage media, and systems that utilize weakly supervised graph matching to align an ungrounded label graph and a visual graph corresponding to a digital image. Specifically, the disclosed system utilizes a label embedding model to generate label graph embeddings from the ungrounded label graph and a visual embedding network to generate visual graph embeddings from the visual graph. Additionally, the disclosed system determines similarity metrics indicating the similarity of pairs of label graph embeddings and visual graph embeddings. The disclosed system then generates a semantic scene graph by utilizing a graph matching algorithm to align the ungrounded label graph and the visual graph based on the similarity metrics. In some embodiments, the disclosed system utilizes contrastive learning to modify the embedding models.

Abstract: Systems, methods, and articles of manufacture are disclosed for learning models of movies, keywords, actors, and roles, and querying the same. In one embodiment, a recommendation application optimizes a model based on training data by initializing the mean and co-variance matrices of Gaussian distributions representing movies, keywords, and actors to random values, and then performing an optimization to minimize a margin loss function using symmetrical or asymmetrical measures of similarity between entities. Such training produces an optimized model with the Gaussian distributions representing movies, keywords, and actors, as well as shift vectors that change the means of movie Gaussian distributions and model archetypical roles. Subsequent to training, the same similarity measures used to train the model are used to query the model and obtain rankings of entities based on similarity to terms in the query, and a representation of the rankings may be displayed via, e.g., a display device.

Abstract: A method for dynamically updating a machine learning model is provided. The method comprises receiving first data comprising a first number of documents, training a first machine learning model with the first data, and releasing the first machine learning model for use. After receipt of the first data, second data is received comprising a second number of documents. A concept labeling engine, according to the first machine learning model, labels a concept for at least one document in the second data. Responsive to determining that a distribution of words in the second data deviates from a distribution of words in the first data beyond a specified threshold, the first machine learning model is retrained with at least part of the second data, the retraining producing a second machine learning model; and releasing the second machine learning model for use.

Abstract: Disclosed are methods and apparatuses for classifier evaluation. The evaluation involves constructing a ground truth refinement having a degree of error within specified bounds from a malware reference dataset as an approximate ground truth refinement. The evaluation further involves using the approximate ground truth refinement to determine at least one of: a lower bound on precision or an upper bound on recall and accuracy. The evaluation further involves evaluating a classifier by evaluating at least one of a classification method or clustering method by examining changes to the upper bound and/or the lower bound produced by the approximate ground truth refinement.

Abstract: The present disclosure relates to a video retrieval method, a method, system and device for generating a video retrieval mapping relationship, and a storage medium. The video retrieval method comprises: acquiring a retrieval instruction, wherein the retrieval instruction carries retrieval information for retrieving a target frame picture; and obtaining the target frame picture according to the retrieval information and a preset mapping relationship.

Abstract: A method and an apparatus for generating a video fingerprint are disclosed. The method includes: performing shot boundary detection on content of a video; determining a time duration of each shot according to positional points of the shot boundary, and compose the time duration of each shot into a shot boundary time slice sequence; and obtaining video fingerprint information according to the time slice sequence.

Abstract: A method for estimating a pose of a deformable object includes: receiving, by a processor, a plurality of images depicting the deformable object from multiple viewpoints; computing, by the processor, one or more object-level correspondences and a class of the deformable object depicted in the images; loading, by the processor, a 3-D model corresponding to the class of the deformable object; aligning, by the processor, the 3-D model to the deformable object depicted in the plurality of images to compute a six-degree of freedom (6-DoF) pose of the object; and outputting, by the processor, the 3-D model and the 6-DoF pose of the object.

Abstract: An automated teller machine comprising a sensor, a network interface, and a processor configured to receive, during a transaction at the automated teller machine, user data from the sensor configured to monitor user movement information during the transaction, receive, via the network interface, user data from a mobile device, determine a risk of a fraudulent transaction based on a comparison of the user movement information from the sensor and the user data from the mobile device to expected parameters, determine that the risk of a fraudulent transaction is below a threshold, and allow access to a secure resource at the automated teller machine.

Abstract: An image-processing device that includes at least one processor which is configured to: set discrete measurement regions in a standard image and a reference image selected from images acquired in a time series manner and calculate motion vectors in the respective measurement regions; set a region of interest in the standard image; select, from the calculated motion vectors, the motion vectors in the set region of interest; estimate a homographic-transformation matrix that represents a motion in the region of interest by using the selected motion vectors; evaluate an error in the homographic-transformation matrix on the basis of the estimated homographic-transformation matrix and the calculated motion vectors; and set, on the basis of the evaluation result for the homographic-transformation matrix estimated on the basis of a first region of interest, a second region of interest a size of which is increased as compared with that of the first region of interest.

Abstract: Systems, devices, and methods are disclosed for generating an authenticity score for a creative work using digital image data of one or more creative works. A system receives, via a network, first digital image data of a first creative work. The system determines from the first digital image data, a first set of feature variables, each corresponding to a characteristic of the first creative work. The system determines, via an artificial intelligence model and with input including the first set of feature variables and a comparison dataset, an authenticity score for the first creative work. The artificial intelligence model generates, based on the portion of first feature variables of the first creative work that match corresponding portions of the comparison dataset, an output indicative of an authenticity score. The system communicates, via the network, an authenticity score of the first creative work to a client device.

Abstract: There is described a computer-implemented method and system for classifying images, the computer-implemented method comprising: receiving an image to be classified, generating a vector representation of the image to be classified using an image embedding method, comparing the vector representation of the image to predefined vector representations of the predefined image categories, and identifying a relevant category amongst the predefined image categories based on the comparison, the relevant category being associated with the image to be classified and outputting the relevant category.

Abstract: A position and pose determination unit (103) is configured to determine a position and an pose of the video acquisition unit (110) in such a way that a level of hiding caused by overlapping of objects becomes low based on environmental object information indicating an environmental object including a structural object present in the monitored area and a placed object placed in the structural object, and staying characteristic information indicating a staying characteristic of the object determined depending on the environmental object. The analysis and display unit (111) is configured to perform at least one of analyzing a video captured by the video acquisition unit (103) at the position and the pose determined by the position and pose determination unit (110) and displaying the video.

Abstract: A system and method for classification. In some embodiments, the method includes forming a first training dataset and a second training dataset from a labeled input dataset; training a first classifier with the first training dataset; training a variational auto encoder with the second training dataset, the variational auto encoder comprising an encoder and a decoder; generating a third dataset, by feeding pseudorandom vectors into the decoder; labeling the third dataset, using the first classifier, to form a third training dataset; forming a fourth training dataset based on the third dataset; and training a second classifier with the fourth training dataset.

Abstract: The invention provides a computer-implemented method of processing an image, the method comprising the steps of: identifying objects in the image using a content analysis engine; based on information from a relationship database, using a contextual analysis engine to automatically: identify contextual relationships between some or all of the identified objects in the image, derive a background or landscape of said image from one or more of said identified objects; and subsequently, and analyse the identified contextual relationships between the identified objects and the derived background or landscape to thereby associate one or more contexts with said image.

Abstract: A first set of pixels forming a first set of rows and a first set of columns of a first image acquired by an image acquisition device is received. A first corrected set of pixels forming a first corrected set of rows and a first corrected set of columns is generated from the first set of pixels. Successive pixels of each row from the first corrected set of rows correspond to successive parallel lines in a plane defined by the sheet of light that are equally spaced along a first axis of a world coordinate system. Successive pixels of each column from the first corrected set of columns correspond to successive parallel lines in the plane defined by the sheet of light that are equally spaced along a second axis of the world coordinate system. Based on the first corrected set of pixels, a first set of points of the sheet of light is determined. The first set of points of the sheet of light is used to obtain a three-dimensional profile of the object.

Abstract: An unmanned aerial vehicle (UAV) includes a propulsion system, a global navigation satellite system (GNSS) sensor, a camera and a controller. The controller includes logic that, in response to execution by the controller, causes the UAV to in response to detecting a loss of tracking by the GNSS sensor determine an estimated location of the UAV on a map based on a location image captured by the camera, determine a route to a destination using tracking parameters embedded in the map, wherein the map is divided into a plurality of sections and the tracking parameters indicate an ease of determining a location of the UAV using images captured by the camera with respect to each section, and control the propulsion system to cause the UAV to follow the route to the destination.

Abstract: Systems and methods are provided for identifying and recommending electronic content to consumers. In accordance with an implementation, one or more elements of electronic content are identified based on video graph data. In an exemplary method, information associated with a first element of video content is received, and corresponding video graph data is obtained. One or more second elements of video content that are similar to the first element of video content are identified based on the obtained video graph data. A subset the first and second elements of video content is subsequently identified for delivery to the user.

Abstract: An image processing pipeline including a still or video camera includes a first neural network arranged to process and provide a neural network based result for at least one of an image capture setting, sensor processing, global post processing, local post processing, and portfolio post processing. A second neural network is arranged to receive the neural network result and further provide at least one of an image capture setting, sensor processing, global post processing, local post processing, and portfolio post processing, wherein at least one of the first and second neural networks generate data on local processors supported by the still or video camera.

Abstract: A computer-implemented method for searching for a similar image in an image database includes receiving a query image, computing a value of a similarity measure of the query image, comparing the value of the similarity measure of the query image with metadata of the partitions, discarding at least one partition of the image database based on the comparison of the value of the similarity measure of the query image with the metadata, comparing the images of at least one of the remaining partitions with the query image, and selecting at least one image of the at least one partition of the remaining partitions as the similar image based on the comparison of the images of the at least one of the remaining partitions with the query image.

Abstract: Systems and methods are provided for receiving a plurality of images corresponding to a listing in an online marketplace, generating a scene type for each image of the plurality of images, and grouping each image into a scene type group of a set of predefined scene types. Each group of images are input into a respective machine learning model specific to the scene type of the group of images to generate a visual score for each image in each group of images, and an attractiveness score is generated for the listing in the online marketplace based on the visual scores for each image in each group of images.

Abstract: An autonomous mobile device (AMD) follows a user who may move among other people, change orientation, or move in and out view of a camera on the AMD. A predicted location of the user may be determined based on a previous location and previous movement. A proximity value is determined by the distance between a location in physical space of a user and the predicted location. An image from a camera is processed to determine a user depicted in the image and generate a feature vector (FV) of that depiction. A gallery of FVs of the user as viewed from different angles, poses, and so forth is stored. A similarity value is determined between a FV of an unidentified user in an image and the FVs in the gallery. The user may be identified using the proximity value and the similarity value.

Abstract: A method for generating an order includes acquiring by one or more processors, an initial image and a first detection result of a target container before being subjected to a pick-up operation, where the first detection result includes a first detection image with a number of first commodity frames obtained by processing the initial image, and first identification information corresponding to a commodity in each of the first commodity frames, acquiring by one or more processors, a result image of the target container after performing the pick-up operation, generating by one or more processors, an RGB difference image between before-shopping and after-shopping, based on the result image and the initial image, where the RGB difference image is obtained by calculating an absolute value of differences of individual pixels of the initial image and the result image before and after shopping in three RGB channels.

Abstract: Techniques for masking images based on a particular task are described. A system masks portions of an image that are not relevant to a particular task, thus, reducing the amount of data used by applications for image processing tasks. For example, images to be processed using a hair color classification model are masked so that only portions that show the person's hair are available for the model to analyze. The system configures different masker components to mask images for different tasks. A masker component can be implemented at a user device to mask images prior to sending to an application/task-specific model.