Abstract: Techniques for training an optical character recognition (OCR) model to detect and recognize text in images for robotic process automation (RPA) are disclosed. A text detection model and a text recognition model may be trained separately and then combined to produce the OCR model. Synthetic data and a smaller amount of real, human-labeled data may be used for training to increase the speed and accuracy with which the OCR text detection model and the text recognition model can be trained. After the OCR model has been trained, a workflow may be generated that includes an activity calling the OCR model, and a robot implementing the workflow may be generated and deployed.

Abstract: Provided are systems and methods for processing 360-degree video data. In various implementations, a spherical representation of a 360-degree video frame can be segmented into a top region, a bottom region, and a middle region. The middle region can be mapped into one or more rectangular areas of an output video frame. The top region can be mapped into a first rectangular area of the output video frame using a mapping that converts a square to a circle, such that pixels in the circular top region are expanded to fill the first rectangular region. The bottom region can be mapped into a second rectangular area of the output video frame such that pixels in the circular bottom region are expanded to fill the second rectangular region.

Abstract: A training method for video stabilization and an image processing device using the same are proposed. The method includes the following steps. An input video including low dynamic range (LDR) images is received. The LDR images are converted to high dynamic range (HDR) images by using a first neural network. A feature extraction process is performed to obtain features based on the LDR images and the HDR images. A second neural network for video stabilization is trained according to the LDR images and the HDR images based on a loss function by minimizing a loss value of the loss function to generate stabilized HDR images in a time-dependent manner, where the loss value of the loss function depends upon the features. An HDR classifier is constructed according to the LDR images and the HDR images. The stabilized HDR images are classified by using the HDR classifier to generate a reward value, where the loss value of the loss function further depends upon the reward value.

Abstract: A neural network in multi-task deep learning paradigm for machine vision includes an encoder that further includes a first, a second, and a third tier. The first tier comprises a first-tier unit having one or more first-unit blocks. The second tier receives a first-tier output from the first tier at one or more second-tier units in the second tier, a second-tier unit comprises one or more second-tier blocks, the third tier receives a second-tier output from the second tier at one or more third-tier units in the third tier, and a third-tier block comprises one or more third-tier blocks. The neural network further comprises a decoder operatively the encoder to receive an encoder output from the encoder as well as one or more loss function layers that are configured to backpropagate one or more losses for training at least the encoder of the neural network in a deep learning paradigm.

Abstract: Systems, methods, and software are described herein for removing people distractors from images. A distractor mitigation solution implemented in one or more computing devices detects people in an image and identifies salient regions in the image. The solution then determines a saliency cue for each person and classifies each person as wanted or as an unwanted distractor based at least on the saliency cue. An unwanted person is then removed from the image or otherwise reduced from the perspective of being an unwanted distraction.

Abstract: An encoding apparatus extracts features of an image by applying multiple padding operations and multiple downscaling operations to an image represented by data and transmits feature information indicating the features to a decoding apparatus. The multiple padding operations and the multiple downscaling operations are applied to the image in an order in which one padding operation is applied and thereafter one downscaling operation corresponding to the padding operation is applied. A decoding method receives feature information from an encoding apparatus, and generates a reconstructed image by applying multiple upscaling operations and multiple trimming operations to an image represented by the feature information. The multiple upscaling operations and the multiple trimming operations are applied to the image in an order in which one upscaling operation is applied and thereafter one trimming operation corresponding to the upscaling operation is applied.

Abstract: Methods and systems are described herein for improvements to authenticate users, particularly authenticating a user based on data known to the user. For example, methods and systems allow for users to be securely authenticated based on data known to the users over remote communication networks without storing the data known to the users. Specifically, methods and systems authenticate users by requiring users to select images that are known to the users. For example, the methods and systems may generate synthetic images based on the user's own images and require the user to select the synthetic image, from a set of a set of images, that is known to the user to authenticate the user. Moreover, the methods and systems alleviate storage and privacy concerns by not storing the data known to the users.

Abstract: A device, method, and non-transitory computer readable medium are described. The method includes receiving a dataset including hand written Arabic words and hand written Arabic alphabets from one or more users. The method further includes removing whitespace around alphabets in the hand written Arabic words and the hand written Arabic alphabets in the dataset. The method further includes splitting the dataset into a training set, a validation set, and a test set. The method further includes classifying one or more user datasets from the training set, the validation set, and the test set. The method further includes identifying the target user from the one or more user datasets. The identification of the target user includes a verification accuracy of the hand written Arabic words being larger than a verification accuracy threshold value.

Abstract: Methods, systems, and apparatus, including computer programs encoded on a computer storage medium, that generate from a first pair and a second pair of images of livestock that are within an enclosure and that are taken at different times using a stereoscopic camera, at least two distance distributions of the aquatic livestock within the enclosure. The distance distributions can be used to determine a measure associated with an optical property of the water within the enclosure. A signal associated with the measure can be provided.

Abstract: A system configured to improve the operations associated with generating virtual representations on limited resources of a mobile device. In some cases, the system may utilize viewpoint bundles that include collection of image data with an associated pose in relative physical proximity to each other to render a virtual scene. In other cases, the system may utilize 2.5D manifolds including 2D image data and a weighted depth value to render the 3D environment.

Abstract: Methods for annotating objects within image frames are disclosed. Information is obtained that represents a camera pose relative to a scene. The camera pose includes a position and a location of the camera relative to the scene. Data is obtained that represents multiple images, including a first image and a plurality of other images, being captured from different angles by the camera relative to the scene. A 3D pose of the object of interest is identified with respect to the camera pose in at least the first image. A 3D bounding region for the object of interest in the first image is defined, which indicates a volume that includes the object of interest. A location and orientation of the object of interest is determined in the other images based on the defined 3D bounding region of the object of interest and the camera pose in the other images.

Abstract: Embodiments of the present disclosure include a method that obtains a digital image. The method includes extracting a word block from the digital image. The method includes processing the word block by evaluating a value of the word block against a dictionary. The method includes outputting a prediction equal to a common word in the dictionary when a confidence factor is greater than a predetermined threshold. The method includes processing the word block and assigning a descriptor to the word block corresponding to a property of the word block. The method includes processing the word block using the descriptor to prioritize evaluation of the word block. The method includes concatenating a first output and a second output. The method includes predicting a value of the word block.

Abstract: While a viewer is viewing a first stereoscopic image comprising a first left image and a first right image, a left vergence angle of a left eye of a viewer and a right vergence angle of a right eye of the viewer are determined. A virtual object depth is determined based at least in part on (i) the left vergence angle of the left eye of the viewer and (ii) the right vergence angle of the right eye of the viewer. A second stereoscopic image comprising a second left image and a second right image for the viewer is rendered on one or more image displays. The second stereoscopic image is subsequent to the first stereoscopic image. The second stereoscopic image is projected from the one or more image displays to a virtual object plane at the virtual object depth.

Abstract: Interpretation maps of convolutional neural networks having certifiable robustness using Rényi differential privacy are provided. In one aspect, a method for generating an interpretation map includes: adding generalized Gaussian noise to an image x to obtain T noisy images, wherein the generalized Gaussian noise constitutes perturbations to the image x; providing the T noisy images as input to a convolutional neural network; calculating T noisy interpretations of output from the convolutional neural network corresponding to the T noisy images; re-scaling the T noisy interpretations using a scoring vector ? to obtain T re-scaled noisy interpretations; and generating the interpretation map using the T re-scaled noisy interpretations, wherein the interpretation map is robust against the perturbations.

Abstract: An example apparatus for video frame segmentation includes a receiver to receive a current video frame to be segmented. The apparatus also includes a segmenting neural network to receive a previous mask including a segmentation mask corresponding to a previous frame and generate a segmentation mask for the current frame based on the previous mask and the video frame.

Abstract: While a viewer is viewing a first stereoscopic image comprising a first left image and a first right image, a left vergence angle of a left eye of a viewer and a right vergence angle of a right eye of the viewer are determined. A virtual object depth is determined based at least in part on (i) the left vergence angle of the left eye of the viewer and (ii) the right vergence angle of the right eye of the viewer. A second stereoscopic image comprising a second left image and a second right image for the viewer is rendered on one or more image displays. The second stereoscopic image is subsequent to the first stereoscopic image. The second stereoscopic image is projected from the one or more image displays to a virtual object plane at the virtual object depth.

Abstract: The present disclosure describes systems, non-transitory computer-readable media, and methods for utilizing a machine learning model trained to determine subtle pose differentiations to analyze a repository of captured digital images of a particular user to automatically capture digital images portraying the user. For example, the disclosed systems can utilize a convolutional neural network to determine a pose/facial expression similarity metric between a sample digital image from a camera viewfinder stream of a client device and one or more previously captured digital images portraying the user. The disclosed systems can determine that the similarity metric satisfies a similarity threshold, and automatically capture a digital image utilizing a camera device of the client device. Thus, the disclosed systems can automatically and efficiently capture digital images, such as selfies, that accurately match previous digital images portraying a variety of unique facial expressions specific to individual users.

Abstract: Described herein are computer-implemented methods for analysis of a tissue sample. An example method includes: annotating the whole tumor regions or set of tumorous sub-regions either on a biomarker image or an H&E image (e.g. from an adjacent serial section of the biomarker image); registering at least a portion of the biomarker image to the H&E image; detecting different cellular and regional tissue structures within the registered H&E image; computing a probability map based on the different detected structures within the registered H&E image; deriving nuclear metrics from each of the biomarker and H&E images; deriving probability metrics from the probability map; and classifying tumor nuclei in the biomarker image based on the computed nuclear and probability metrics.

Abstract: A method for operating a distributed neural network having a plurality of intelligent devices and a server includes: generating, by a first intelligent device of the plurality of intelligent devices, a first output using a first neural network model running on the first intelligent device and using a first input vector to the first neural network model; outputting, by the first intelligent device, the first output; receiving, by the first intelligent device, a gesture feedback on the first output from a user; determining, by the first intelligent device, a user rating of the first output from the gesture feedback; labeling, by the first intelligent device, the first input vector with a first label in accordance with the user rating; and training, by the first intelligent device, the first neural network model using the first input vector and the first label.

Abstract: The technology disclosed can provide methods and systems for identifying users while capturing motion and/or determining the path of a portion of the user with one or more optical, acoustic or vibrational sensors. Implementations can enable use of security aware devices, e.g., automated teller machines (ATMs), cash registers and banking machines, other secure vending or service machines, security screening apparatus, secure terminals, airplanes, automobiles and so forth that comprise sensors and processors employing optical, audio or vibrational detection mechanisms suitable for providing gesture detection, personal identification, user recognition, authorization of control inputs, and other machine control and/or machine communications applications. A virtual experience can be provided to the user in some implementations by the addition of haptic, audio and/or other sensory information projectors.