Abstract: Various embodiments include methods for identifying inconsistencies in images that could be due to malicious attacks. Various embodiments may include receiving a plurality of camera images from one or more cameras of an apparatus (e.g., a vehicle), performing a plurality of different processes on the plurality of images to detect different types of image inconsistencies, using results of the plurality of different processes on the plurality of images to recognize a vision attack and performing one or more mitigation actions in response to recognizing a vision attack. The plurality of different processes may include temporal consistency checks on the plurality of images spanning a period of time, inconsistency counter checks on the plurality of images that determine whether a number of inconsistencies in camera images satisfies a threshold, past history checks on the plurality of images comparing objects previously recognized to objects recognized in currently obtained images.

Abstract: Systems and techniques are provided for performing scene segmentation and object tracking. For example, a method for processing one or more frames. The method may include determining first one or more features from a first frame. The first frame includes a target object. The method may include obtaining a first mask associated with the first frame. The first mask includes an indication of the target object. The method may further include generating, based on the first mask and the first one or more features, a representation of a foreground and a background of the first frame. The method may include determining second one or more features from a second frame and determining, based on the representation of the foreground and the background of the first frame and the second one or more features, a location of the target object in the second frame.

Abstract: Systems and techniques are provided for performing an accurate object count using monocular three-dimensional (3D) perception. In some examples, a computing device can generate a reference depth map based on a reference frame depicting a volume of interest. The computing device can generate a current depth map based on a current frame depicting the volume of interest and one or more objects. The computing device can compare the current depth map to the reference depth map to determine a respective change in depth for each of the one or more objects. The computing device can further compare the respective change in depth for each object to a threshold. The computing device can determine whether each object is located within the volume of interest based on comparing the respective change in depth for each object to the threshold.

Abstract: System and techniques are described herein for processing images to detect objects in the provided images. In one illustrative example, a method of processing image data includes obtaining an image including at least a first object. The method can include generating a feature map based on providing the image to a neural network. The method can further include identifying a plurality of objects based on the feature map, the plurality of objects including a first part of the first object. The method can include identifying a first set of object parts within the plurality of objects corresponding to the first object.

Abstract: Systems and techniques are provided for performing scene segmentation and object tracking. For example, a method for processing one or more frames. The method may include determining first one or more features from a first frame. The first frame includes a target object. The method may include obtaining a first mask associated with the first frame. The first mask includes an indication of the target object. The method may further include generating, based on the first mask and the first one or more features, a representation of a foreground and a background of the first frame. The method may include determining second one or more features from a second frame and determining, based on the representation of the foreground and the background of the first frame and the second one or more features, a location of the target object in the second frame.

Abstract: Disclosed are systems, apparatuses, processes, and computer-readable media to implement a heterogenous biometric authentication process in a control system. For example, a method may include detecting the presence of a first person at a first time period and in an area associated with a function controlled by a control system. The method may include transmitting an authentication request to a first device detected by the control system, and receiving an authentication response from the first device. The authentication response includes information related to a biometric authentication performed at the first device. The method may further include authenticating the first person in the control system based on the information related to the biometric authentication. The method may then perform the function based on the authentication.

Abstract: A method for picture processing is described. A first tracking area is obtained. A second tracking area is also obtained. The method includes beginning to track the first tracking area and the second tracking area. Picture processing is performed once a portion of the first tracking area overlapping the second tracking area passes a threshold.

Abstract: A method for picture processing is described. A first tracking area is obtained. A second tracking area is also obtained. The method includes beginning to track the first tracking area and the second tracking area. Picture processing is performed once a portion of the first tracking area overlapping the second tracking area passes a threshold.

Abstract: This disclosure is directed to methods and systems for protecting a deep learning model from piracy and unauthorized uses. The protection may be implemented by embedding an ownership detection mechanism such that unauthorized use of the model may be detected using a detection input data and corresponding model signature. In addition, the deep learning model may be used in conjunction with a secret or license protected data encoder such that the deep learning model may generate meaningful output only when processing encoded input data. An unauthorized user who does not have access to the secret data encoder may not be able to use a pirated copy of the deep learning model to generate meaningful output. Under such a scheme, a deep learning model itself may be widely distributed without restriction and without license-protection.

Abstract: A method includes receiving a user input (e.g., a one-touch user input), performing segmentation to generate multiple candidate regions of interest (ROIs) in response to the user input, and performing ROI fusion to generate a final ROI (e.g., for a computer vision application). In some cases, the segmentation may include motion-based segmentation, color-based segmentation, or a combination thereof. Further, in some cases, the ROI fusion may include intraframe (or spatial) ROI fusion, temporal ROI fusion, or a combination thereof.

Abstract: This disclosure relates to digital image segmentation, region of interest identification, and object recognition. This disclosure describes a method, a system, for image segmentation based on fully convolutional neural network including an expansion neural network and contraction neural network. The various convolutional and deconvolution layers of the neural networks are architected to include a coarse-to-fine residual learning module and learning paths, as well as a dense convolution module to extract auto context features and to facilitate fast, efficient, and accurate training of the neural networks capable of producing prediction masks of regions of interest. While the disclosed method and system are applicable for general image segmentation and object detection/identification, they are particularly suitable for organ, tissue, and lesion segmentation and detection in medical images.

Abstract: A method for picture processing is described. A first tracking area is obtained. A second tracking area is also obtained. The method includes beginning to track the first tracking area and the second tracking area. Picture processing is performed once a portion of the first tracking area overlapping the second tracking area passes a threshold.

Abstract: A method for picture processing is described. A first tracking area is obtained. A second tracking area is also obtained. The method includes beginning to track the first tracking area and the second tracking area. Picture processing is performed once a portion of the first tracking area overlapping the second tracking area passes a threshold.

Abstract: This disclosure is directed to a system and a method for providing enhanced real-time or near-real-time response to request for detached data analytics services. In one implementation, a system is disclosed for predicting a data analytics service that may be requested by a user based on real-time user interactive operations, and for pre-loading/pre-configuring a pipeline of data analytics components for performing the predicted data analytics service before an actual request is made. Additionally, at least some intermediate data may be calculated by the pre-configured pipeline and may be pre-cached in memory. Upon actual user request for the data analytics service, only data analytics that require additional input data concurrently provided with the request would need to be performed. In such a manner, user-perceived delay in completing the detached data analytics service is reduced.

Abstract: To determine real-world information about objects moving in a scene, the camera capturing the scene is typically calibrated to the scene. Automatic scene calibration can be accomplished using people that are found moving about in the scene. During a calibration period, a video content analysis system processing video frames from a camera can identify blobs that are associated with people. Using an estimated height of a typical person, the video content analysis system can use the location of the person's head and feet to determine a mapping between the person's location in the 2-D video frame and the person's location in the 3-D real world. This mapping can be used to determine a cost for estimated extrinsic parameters for the camera. Using a hierarchical global estimation mechanism, the video content analysis system can determine the estimated extrinsic parameters with the lowest cost.

Abstract: The system and method disclosed herein provides an integrated and automated workflow, sensor, and reasoning system that automatically detects breaches in protocols, appropriately alarms and records these breaches, facilitates staff adoption of protocol adherence, and ultimately enables the study of protocols for care comparative effectiveness. The system provides real-time alerts to medical personnel in the actual processes of care, thereby reducing the number of negative patient events and ultimately improving staff behavior with respect to protocol adherence.

Abstract: A method includes receiving information that identifies a reference position in a location space. The method also includes receiving data that identifies one among a plurality of candidate geometrical arrangements. The method also includes producing a representation that depicts a plurality of objects which are arranged, relative to the reference position in the location space, according to the identified candidate geometrical arrangement.

Abstract: Automatic adaptive zoom enables computing devices that receive video streams to use a higher resolution stream when the user enables zoom, so that the quality of the output video is preserved. In some examples, a tracking video stream and a target video stream are obtained and are processed. The tracking video stream has a first resolution, and the target video stream has a second resolution that is higher than the first resolution. The tracking video stream is processed to define regions of interest for frames of the tracking video stream. The target video stream is processed to generate zoomed-in regions of frames of the target video stream. A zoomed-in region of the target video stream corresponds to a region of interest defined using the tracking video stream. The zoomed-in regions of the frames of the target video stream are then provided for display on a client device.

Abstract: This disclosure relates to digital image segmentation, region of interest identification, and object recognition. This disclosure describes a method, a system, for image segmentation based on fully convolutional neural network including an expansion neural network and contraction neural network. The various convolutional and deconvolution layers of the neural networks are architected to include a coarse-to-fine residual learning module and learning paths, as well as a dense convolution module to extract auto context features and to facilitate fast, efficient, and accurate training of the neural networks capable of producing prediction masks of regions of interest. While the disclosed method and system are applicable for general image segmentation and object detection/identification, they are particularly suitable for organ, tissue, and lesion segmentation and detection in medical images.

Abstract: In accordance with one aspect of the present technique, a method includes receiving one or more videos from one or more image capture devices. The method further includes generating a video-loop of the person from the one or more videos. The video-loop depicts the person in the commercial site. The method also includes generating an action clip from the video-loop. The action clip includes a suspicious action performed by the person in the commercial site. The method further includes generating an activity summary of the person including the video-loop and the action clip.