Abstract: Multiple anchors may be utilized for robotic process automation (RPA) of a user interface (UI). The multiple anchors may be utilized to determine relationships between elements in the captured image of the UI for RPA. The results of the anchoring may be utilized for training or retraining of a machine learning (ML) component.

Abstract: The invention provides a decision-making method of comprehensive alumina production indexes based on a multi-scale deep convolutional network. The method mainly consists of several sub-models: a multi-scale deep splicing convolutional neural network prediction sub-model reflecting the influence of bottom-layer production process indexes on the comprehensive alumina production indexes, a full connecting neural network prediction sub-model reflecting the influence of upper-layer dispatching indexes on the comprehensive alumina production indexes, a full connecting neural network prediction sub-model reflecting the influence of the comprehensive alumina production indexes at a past time on current comprehensive alumina production indexes, and a multi-scale information neural network integrated model for collaborative optimization of sub-model parameters.

Abstract: Methods, apparatus, systems and articles of manufacture to provide machine assisted programming are disclosed. An example apparatus includes a feature extractor to convert compiled code into a first feature vector; a first machine leaning model to identify a cluster of stored feature vectors corresponding to the first feature vector; and a second machine learning model to recommend a second algorithm corresponding to a second feature vector of the cluster based on a comparison of a parameter of a first algorithm corresponding to the first feature vector and the parameter of the second algorithm.

Abstract: Various examples of hybrid metric-topological camera-based localization are described. A single image sensor captures an input image of an environment. The input image is localized to one of a plurality of topological nodes of a hybrid simultaneous localization and mapping (SLAM) metric-topological map which describes the environment as the plurality of topological nodes at a plurality of discrete locations in the environment. A metric pose of the image sensor can be determined using a Perspective-n-Point (PnP) projection algorithm. A convolutional neural network (CNN) can be trained to localize the input image to one of the plurality of topological nodes and a direction of traversal through the environment.

Abstract: The disclosure may provide a method for obtaining a document layout, an electronic device, and a storage medium. The method may include: obtaining a plurality of characters and character information of each of the plurality of characters in a target document; generating a character fully-connected graph based on the plurality of characters and the character information of each of the plurality of characters; obtaining a spatial semantic feature of each of the plurality of characters; generating types of character vertices based on the character information of each of the plurality of characters; generating types of edges based on the spatial semantic feature of each of the plurality of characters, and the character information of each of the plurality of characters; and obtaining a document layout of the target document based on the types of the character vertices and the types of the edges.

Abstract: There is disclosed a method and system for determining a predicted state of a traffic signal. A video of the traffic signal is received. Still images of the traffic signal are generated based on the video. A first machine learning algorithm (MLA) outputs a vector for each bulb in each still image, the vector indicating a predicted status of the bulb. A second MLA determines a predicted state of the traffic signal based on the vectors.

Abstract: According to examples, an apparatus may include a processor and a non-transitory computer readable medium on which is stored machine readable instructions that may cause the processor to identify Internet protocol (IP) addresses and connection attributes associated with the IP addresses. The instructions may also cause the processor to train a machine learning model using the IP addresses as inputs to the machine learning model and connection contexts as outputs of the machine learning model. The machine learning model may learn a first weight matrix corresponding to the IP addresses and a second weight matrix corresponding to the connection contexts. In addition, the connection contexts may be concatenations of the connection attributes associated with a corresponding IP address.

Abstract: An eyewear device with camera-based compensation that improves the user experience for user's having partial blindness or complete blindness. The camera-based compensation determines objects, converts determined objects to text, and then converts the text to audio that is indicative of the objects and that is perceptible to the eyewear user. The camera-based compensation may use a region-based convolutional neural network (RCNN) to generate a feature map including text that is indicative of objects in images captured by a camera. Relevant text of the feature map is then processed through a text to speech algorithm featuring a natural language processor to generate audio indicative of the objects in the processed images.

Abstract: Video analysis tool systems and methods are described for a streamlined presentation feedback on a centralized platform to reduce or eliminate a delay time associated with presentation feedback. The video analysis software application tool may be configured to record a presentation to generate a video recording, analyze the video recording of the presentation based on feedback parameters and an associated neural network model, and automatically provide feedback based on the analysis of the video recording.

Abstract: This application provides a gesture recognition method, and relates to the field of man-machine interaction technologies. The method includes: extracting M images from a first video segment in a video stream; performing gesture recognition on the M images by using a deep learning algorithm, to obtain a gesture recognition result corresponding to the first video segment; and performing result combination on gesture recognition results of N consecutive video segments including the first video segment, to obtain a combined gesture recognition result. In the foregoing recognition process, a gesture in the video stream does not need to be segmented or tracked, but phase actions are recognized by using a deep learning algorithm with a relatively fast calculation speed, and then the phase actions are combined, so as to improve a gesture recognition speed, and reduce a gesture recognition delay.

Abstract: A detector for characterizing at least one of a middle ear fluid and a middle ear biofilm includes a handheld probe outputting near-infrared and visible light, an OCT system to obtain A-scans at a plurality of positions on a tympanic membrane, and a camera to obtain surface sub-images at the plurality of positions. A-scans and surface sub-images are synchronized and the surface sub-images are mosaicked to generate a surface image of the tympanic membrane. Cross-sectional scan images or a thickness map are generated from the synchronized A-scans and segmented to extract a plurality of specified features. The specified features are then classified to characterize at least one of the middle ear fluid and the middle ear biofilm. The detector, including handheld probe with camera, OCT system, and a laptop computer, is sized to fit into a handheld, portable, compact, foam-padded briefcase weighing less than 10 kg.

Abstract: Systems, apparatuses, and methods for efficiently and accurately processing an image in order to detect and identify one or more objects contained in the image, and methods that may be implemented on mobile or other resource constrained devices. Embodiments of the invention introduce simple, efficient, and accurate approximations to the functions performed by a convolutional neural network (CNN); this is achieved by binarization (i.e., converting one form of data to binary values) of the weights and of the intermediate representations of data in a convolutional neural network. The inventive binarization methods include optimization processes that determine the best approximations of the convolution operations that are part of implementing a CNN using binary operations.

Abstract: The present disclosure provides various approaches for smart area monitoring suitable for parking garages or other areas. These approaches may include ROI-based occupancy detection to determine whether particular parking spots are occupied by leveraging image data from image sensors, such as cameras. These approaches may also include multi-sensor object tracking using multiple sensors that are distributed across an area that leverage both image data and spatial information regarding the area, to provide precise object tracking across the sensors. Further approaches relate to various architectures and configurations for smart area monitoring systems, as well as visualization and processing techniques. For example, as opposed to presenting video of an area captured by cameras, 3D renderings may be generated and played from metadata extracted from sensors around the area.

Abstract: Embodiments disclose an artificial intelligence chip and a convolutional neural network applied to the artificial intelligence chip comprising a processor, at least one parallel computing unit, and a pooling computation unit. The method comprises: dividing a convolution task into convolution subtasks and corresponding pooling subtasks; executing convolution subtasks at different parallel computing units, and performing convolution, batch normalization, and non-linear computing operation in a same parallel computing unit; sending an execution result of each parallel computing unit from executing the convolution subtask to the pooling computation unit for executing the corresponding pooling subtask; merging executing results of the pooling computation unit from performing pooling operations on the executing results outputted by respective convolution subtasks to obtain an execution result of the convolution task.

Abstract: Embodiments of the present disclosure provide methods, devices, and computer program products for model adaptation. The method for model adaptation comprises: receiving, at a first computing device, a data set to be analyzed from a data collector and determining abnormality of the data set to be analyzed using a machine learning model deployed at the first computing device. The method further comprises transmitting, based on the determined abnormality of the data set, at least a portion of data in the data set to a second computing device, for update of the machine learning model, the second computing device having a higher computing capability than the first computing device. The method further comprises obtaining redeployment of the updated machine learning model from the second computing device.

Abstract: The present disclosure pertains generally to image feature extraction. Both transfer-learning and multi-task training approaches are considered. In one example, a machine learning model is trained to perform a geographic classification task of distinguishing between images captured in different geographic regions based on their visual content. In another example, a machine learning model is trained to perform an order recognition task of determining information about the order of an image sequence based on its visual content, where the order of the image sequence may be different than the order in which its constituent images were captured. A further example combines the two approaches. The knowledge gained by the ML model in learning one or more such tasks can be applied to a desired image recognition task, such as image segmentation, structure detection or image classification, e.g. with a pre-training/fine-tuning framework or a multi-task learning framework.

Abstract: Image data from a camera and depth information from a depth sensor, such as a LiDAR system, are used to segment an image for decoding an optical pattern. The image data is spatially correlated with the depth information. The depth information is used to partition the image into one or more foreground segments and one or more background segments. Scanning for the optical pattern is performed on the one or more foreground segments.

Abstract: Provided is a method of generating a video synopsis of a sports game including: based on a video including a sports game and log information sequentially recording events occurring in the sports game, determining an event section of the video corresponding to a preset event; determining a search section in the video based on log information and the determined event section; detecting a preset object in at least one scene section included in a search section; and generating a first video based on the at least one scene section in which the preset object is detected.

Abstract: A method of planning the correction of spinal deformations of a subject, by performing segmentation on a three dimensional image of the subject's spine in its erect neutral position, such that the positions and orientations of the vertebrae in a region of interest are characterized. Parameters relating to the alignment and position of the vertebrae are derived from the segmentation, followed by determining whether the parameters fall within an acceptable range desired for the spine of the subject. If not within the acceptable range, an alignment optimization is performed on the vertebrae to bring the parameters within the acceptable range, to reduce the spinal deformations of the subject's spine. The alignment optimization is performed by taking into consideration limitations arising from the dynamic range of motion of the vertebrae as determined by analyzing images of the subject's spine, while the subject is in positions of maximum bending.

Abstract: A system and method for managing large numbers of computing devices in a data center are disclosed. The computing devices are configured to flash their indicator lights in a pattern that encodes a device ID, and an image capture device such as a mobile phone or tablet captures the flashes in a series of images/video of the data center. The images/video are processed to create a three-dimensional (3D) model of the data center with computing device IDs positioned therein. The 3D model, including correctly positioned device ID indicators, can be rendered for the user of the mobile device to enable the user to more easily identify computing device locations.