Abstract: Methods, systems, and apparatus, including computer programs encoded on a computer storage medium for encoding input data comprising input data values corresponding to respective input data grid points of an input data grid, such as image, video or audio data.

Abstract: Embodiments described herein provide systems and methods for training video models to perform a task from an input instructional video. A procedure knowledge graph (PKG) may be generated with nodes representing procedure steps, and edges representing relationships between the steps. The PKG may be generated based on text and/or video training data which includes procedures (e.g., instructional videos). Using the PKG, a video model may be trained using the PKG to provide supervisory training signals for a number of tasks. Once the model is trained, it may be fine-tuned for a specific task which benefits from the model being trained in a way that makes the model embed procedural information when encoding videos.

Abstract: A plastic item, such as a beverage bottle, conveys two distinct digital watermarks, encoded using two distinct signaling protocols. A first, printed label watermark conveys a retailing payload, including a Global Trade Item Number (GTIN) used by a point-of-sale scanner in a retail store to identify and price the item when presented for checkout. A second, plastic texture watermark conveys a recycling payload, including data identifying the composition of the plastic. The use of two different signaling protocols assures that a point-of-sale scanner will not spend its limited time and computational resources working to decode the recycling watermark, which lacks the data needed for retail checkout. In some embodiments, a recycling apparatus makes advantageous use of both types of watermarks to identify the plastic composition of the item (e.g., relating GTIN to plastic type using an associated database), thereby increasing the fraction of items that are correctly identified for sorting and recycling.

Abstract: A learning data generation device for generating learning data for learning a recognizer capable of estimating a contour of a sphere making spinning motion, with high accuracy, the sphere being recorded in a single camera video image, is provided.

Abstract: This document describes techniques and systems for generating a fused object bounding box based on uncertainty. At least two bounding boxes, each associated with a different sensor, is generated. A fused center point and yaw angle as well as length, width, and velocity can be found by mixing the distributions of the parameters from each bounding box. A discrepancy between the center points of each bounding box can be used to determine whether to refine the fused bounding box (e.g., find an intersection between at least two bounding boxes) or consolidate the fused bounding box (e.g., find a union between at least two bounding boxes). This results in the fused bounding box having a confidence level of the uncertainty associated with the fused bounding box. In this manner, better estimations of the uncertainty of the fused bounding box may be achieved to improve tracking performance of a sensor fusion system.

Abstract: Techniques for determining a probability that a first sensor is miscalibrated with respect a second sensor are discussed herein. For example, a computing device may receive calibrated extrinsics of a camera to a lidar, determine a plurality of sets of perturbed extrinsics based on the calibrated extrinsics, determine respective costs for perturbed extrinsics of the plurality of sets of perturbed extrinsics based on image data captured by the camera, the plurality of sets of perturbed extrinsics, and lidar data captured by the lidar, and determine a local maxima score for the calibrated extrinsics based at least in part on the respective costs for the perturbed extrinsics of the plurality of sets of perturbed extrinsics and a cost of the calibrated extrinsics. The computing device may then determine a probability that the camera is miscalibrated based on a Bayes probability and the local maxima score.

Abstract: A method for detecting defects in products from images thereof and an electronic device applying the method inputs a defect image repair data set into an autoencoder to train the autoencoder, and generates a reconstructed image, calculates a reference error value between the sample image and the reconstructed image by a preset error function, and set a threshold value based on the reference error value. The electronic device inputs an image possibly revealing a defect into the autoencoder and generates the reconstructed image corresponding to the image to be detected, and uses the preset error function to calculate the reconstruction error between the image and the reconstructed image, thereby determining whether the image being analyzed does reveal defects. When the reconstruction error is greater than the threshold value, a determination is made that a defect is revealed.

Abstract: The present disclosure relates to systems, methods, and non-transitory computer readable media for accurately enhancing optical character recognition with a machine learning approach for determining words from reverse text, vertical text, and atypically-sized text. For example, the disclosed systems segment a digital image into text regions and non-text regions utilizing an object detection machine learning model. Within the text regions, the disclosed systems can determine reverse text glyphs, vertical text glyphs, and/or atypically-sized text glyphs utilizing an edge based adaptive binarization model. Additionally, the disclosed systems can utilize respective modification techniques to manipulate reverse text glyphs, vertical text glyphs, and/or atypically-sized glyphs for analysis by an optical character recognition model.

Abstract: An image processing method and apparatus based on machine learning are disclosed. The image processing method based on machine learning, according to the present invention, may comprise the steps of: generating a first corrected image by inputting an input image to a first convolution neural network; generating an intermediate image on the basis of the input image; performing machine learning on a first loss function of the first convolution neural network on the basis of the first corrected image and the intermediate image; and performing machine learning on a second loss function of the first convolution neural network on the basis of the first corrected image and a natural image.

Abstract: There is provided a recognition system adaptable to a portable device or a wearable device. The recognition system senses a body heat using a thermal sensor, and performs functions such as the living body recognition, image denoising and body temperature prompting according to detected results.

Abstract: There is provided a recognition system adaptable to a portable device or a wearable device. The recognition system senses a body heat using a thermal sensor, and performs functions such as the living body recognition, image denoising and body temperature prompting according to detected results.

Abstract: Disclosed is an image processing device comprising a processor configured to estimate the scale of image features by the steps of: processing multiple images of a scene by means of a first trained model to identify features in the images and to estimate the depths of those features in the images; processing the multiple images by a second trained model to estimate a scaling for the images; and estimating the scales of the features by adjusting the estimated depths in dependence on the estimated scaling. A method for training an image processing model is also disclosed.

Abstract: Systems and methods for multi-task joint training of a neural network including an encoder module and a multi-headed attention mechanism are provided. In one aspect, the system includes a processor configured to receive input data including a first set of labels and a second set of labels. Using the encoder module, features are extracted from the input data. Using a multi-headed attention mechanism, training loss metrics are computed. A first training loss metric is computed using the extracted features and the first set of labels, and a second training loss metric is computed using the extracted features and the second set of labels. A first mask is applied to filter the first training loss metric, and a second mask is applied to filter the second training loss metric. A final training loss metric is computed based on the filtered first and second training loss metrics.

Abstract: A method and system for training a neural network are provided. The method includes receiving an input image, selecting at least one data augmentation method from a pool of data augmentation methods, generating an augmented image by applying the selected at least one data augmentation method to the input image, and generating a mixed image from the input image and the augmented image.

Abstract: A method for querying data obtained from a distributed sensor network, comprising: receiving sensor data representing an aspect of an environment with a sensor of the distributed sensor network; communicating a representation of the sensor data to a fog node through an automated communication network; determining, by the fog node, a correspondence of a query received through the automated communication network to characteristics of the representation of the sensor data; and selectively communicating, in response to the query, at least one of: the sensor data having the determined characteristics corresponding to the query, an identification of the sensor data having the determined characteristics corresponding to the query, and the data representing the sensor data having the determined characteristics corresponding to the query.

Abstract: A method for qualitative and/or quantitative characterization of a coating surface is provided, comprising: providing a program recognizing coating surface defect types; determining, by the program, whether a camera(s) coupled to the program is within a predefined distance range and/or within a predefined image acquisition angle range relative to a currently presented coating surface; depending on the determination: generating a feedback signal indicative of whether adjustment of the position of the camera(s) is within predefined distance range and/or within the predefined image acquisition angle range; and/or automatically adjusting the relative distance of the camera and and/or automatically adjusting the angle of the camera; enabling the camera to acquire an image of the coating surface only when the camera(s) is/are within the predefined distance range and/or image acquisition angle range; processing the digital image for recognizing coating surface defects; and outputting a characterization of the

Abstract: Systems and methods for determining attributes in media content. A computing device may be configured to obtain the media content based on a received media content identifier and segment the media content into scenes. The computing device may analyze viewer engagement metrics to identify a scene associated with a viewer engagement score that exceeds a threshold, select video frames from the identified scene, and identify primary objects in the series of images in the scene. The computing device may add a bounding box around the identified primary objects in one or more selected frames and perform text extraction within the bounding box. The computing device may determine object attributes of the identified primary objects, querying a database to identify topics of interest (ToIs) based on the extracted text and the determined object attributes, and performing a responsive action in response to identifying the one or more ToIs.

Abstract: A system produces a dolly zoom effect by utilizing side view information. The system first captures a main image at a main location. The main image includes at least a foreground object of a given size and a background. The system calculates one or more side view locations based on a zoom-in factor to be applied to the background and an estimated size of the foreground object. The system then guides a user to capture one or more side view images at the one or more side view locations. The foreground object of the given size is superimposed onto a zoomed-in background. Then the side view information is used by the system to perform image inpainting.

Abstract: A computing device is provided, including a processor configured to receive imaging relevance data for a geographic area. The processor may be further configured to generate, based at least in part on the imaging relevance data, image mask instructions specifying a region of interest included in the geographic area. The processor may be further configured to transmit the image mask instructions to a satellite. The processor may be further configured to receive, from the satellite, filtered satellite image data of the region of interest. One or more deprioritized regions of the geographic area outside the region of interest may be excluded from the filtered satellite image data.

Abstract: In implementations of systems for generating indications of relationships between electronic documents, a processing device implements a relationship system to segment text of electronic documents included in a document corpus into segments. The relationship system determines a subset of the electronic documents that includes electronic document pairs having a number of similar segments that is greater than a threshold number. The similar segments are identified using locality sensitive hashing. The electronic document pairs are classified as related documents or unrelated documents using a machine learning model that receives a pair of electronic documents as an input and generates an indication of a classification for the pair of electronic documents as an output. Indications of relationships between particular electronic documents included in the subset are generated based at least partially on the electronic document pairs that are classified as related documents.