Abstract: Exemplary embodiments provide distributed parallel training of a machine learning model. Multiple processors may be used to train a machine learning model to reduce training time. To synchronize trained model data between the processors, data is communicated between the processors after some number of training cycles. To improve the communication efficiency, exemplary embodiments synchronize data among a set of processors after a predetermined number of training cycles, and synchronize data between one or more processors of each set of the processors after a predetermined number of training cycles. During the first synchronization among a set of processors, compressed model gradient data generated after performing the training cycles may be communicated. During the second synchronization between the set of processors, trained models or full model gradient data generated after performing the training cycles may be communicated.

Abstract: In one embodiment, a method of signaling individual layers in a transport stream includes: determining a plurality of layers in a transport stream, wherein each layer includes a respective transport stream parameter setting; determining an additional layer for the plurality of layers in the transport stream, wherein the additional layer enhances one or more of the plurality of layers including a base layer and the respective layer parameter settings for the plurality of layers do not take into account the additional layer; and determining an additional transport stream parameter setting for the additional layer, the additional transport stream parameter setting specifying a relationship between the additional layer and at least a portion of the plurality of layers, wherein the additional transport stream parameter setting is used to decode the additional layer and the at least a portion of the plurality of layers.

Abstract: Methods, systems, and apparatus for training model parameters stored in shared memory to predict risk. The method may include obtaining training data that includes a plurality of training data structures that each represent attributes of an entity, wherein each training data structure represents (i) features derived from a first set of categories defined by a first model and from a second set of categories defined by a second model, and (ii) a risk-level associated with the entity. For each respective training data structure, providing the training data structure as an input to the model, receiving an output from the model based on the model's processing of the training data structure, determining an amount of error between the output of the model and the risk-level of the training data structure, and adjusting a parameter value of the model stored in a shared memory based on the determined error.

Abstract: Disclosed are methods, apparatuses, and systems for encoding and decoding an image. The present invention provides an intra prediction unit receives an input image, removes high frequency ingredients by low pass filtering an encoded luma pixel value in the input image during intra prediction, and generates a prediction block by predicting a chroma pixel value by using a low pass filter (LPF) LM chroma mode for applying an LM chroma mode, which is an extended chroma mode technique for generating a prediction block by predicting the chroma pixel value by applying a correlation between color planes to the luma pixel value having removed therefrom the high frequency ingredients.

Abstract: An embodiment may include a machine learning based classifier that maps input observations into respective categories and a database containing a corpus of training data for the classifier. The training data includes a plurality of entries, each entry having an observation respectively associated with a ground truth category thereof. A computing device may be configured to select, from the training data, a plurality of subsets each containing a different number of entries. The computing device may also be configured to, for each particular subset: (i) divide the particular subset into a training portion and a validation portion, (ii) train the classifier with the training portion, (iii) provide the validation portion as input to the classifier as trained, and (iv) based on how entries of the validation portion are mapped to the categories, determine a respective precision for the particular subset.

Abstract: A method of automatically geolocating a visual target. The method comprises operating a flying vehicle in a search region including the visual target. The method further includes affirmatively identifying a visual target in an aerial photograph of the search region captured by the flying vehicle. The method further includes automatically correlating the aerial photograph of the search region to a geo-tagged photograph of the search region, wherein the geo-tagged photograph is labelled with pre-defined geospatial coordinates. Based on such automatic correlation, a geospatial coordinate is determined for the visual target in the search region.

Abstract: An audio/video recording and communication device (A/V device) including one or more motion sensors arranged in a first row, a first camera, one or more second cameras arranged in a second row, a first cover disposed over the one or more motion sensors, a second cover disposed over the first camera, and a third cover disposed over the one or more second cameras. The one or more motion sensors may include a first motion sensor having a first orientation, a second motion sensor having a second orientation, and a third motion sensor having a third orientation. Additionally, the one or more second cameras may include a camera having a first orientation, a camera having a second orientation, and a camera having a third orientation. In some instances, the first camera is located between the one or more motion sensors and the one or more second cameras.

Abstract: Activating security system alarms based on data generated by audio/video (A/V) recording and communication devices in accordance with various embodiments of the present disclosure are provided. In one embodiment, an network device communicatively coupled to a security system in an armed mode is provided, the network device comprising: one or more processors; a communication module; and a non-transitory machine-readable memory storing a program comprising sets of instructions for: receiving sensor data indicative of an entry/exit event at an entry point; determining, based on at least one of motion data and image data generated by an A/V recording and communication device having a field of view, whether a person was present in the field of view of the A/V recording and communication device prior to the entry/exit event; determining whether the person is authorized; and upon determining that the person is not authorized, activating an alarm action of the security system.

Abstract: The present disclosure provides a video processing method, applied to an unmanned aerial vehicle (UAV) equipped with a camera device for capturing videos. The video processing method includes in response to the UAV moving in accordance with a flight trajectory, controlling the camera device of the UAV to obtain a first video segment when reaching a first photography point; in response to reaching a second photography point as the UAV continues moving, controlling the camera device of the UAV to capture environmental images to obtain a panoramic image, and generating a second video segment based on the panoramic image; and generating a target video based on the first video segment and the second video segment.

Abstract: Sensor information and map information may be obtained. The sensor information may characterize positions of objects in an environment of a sensor. The map information may characterize a road configuration in an environment of a vehicle. A sensor range configuration for the vehicle may be determined based on the road configuration in the environment of the vehicle. A portion of the sensor information may be processed for vehicle navigation based on the sensor range configuration.

Abstract: This disclosure describes, in part, an audio/video (A/V) device that includes a first camera for generating first image data and one or more second cameras for both motion detection and generating second image data. For instance, the A/V device may generate and then store the second image data in one or more buffers. The A/V device may then detect an event, such as possible motion of an object. Based on detecting the event, A/V device may cause the one or more second cameras to cease generating the second image data and cause the first camera to begin generating the first image data. The A/V device may then process the second image data and send the processed second image data to a computing system. Next, the A/V device may process the first image data and send the processed first image data to the computing system.

Abstract: An imaging apparatus that generates a wide-angle image on the basis of images from a plurality of imaging units and is capable of decreasing a likelihood of a subject becoming a double image. The imaging apparatus includes a determination unit configured to determine an imaging range of at least one of a first imaging unit and a second imaging unit on the basis of information relating to a subject designated by a user in a case in which a switching instruction from a non-composition mode to a composition mode and information relating to a subject designated by the user are received from an external device.

Abstract: A method of detecting an object in a hallway proximate a door including: receiving a status of an access control operably connected to a door of a room, the status indicating that the door has been opened; determining that an object is located in a hallway proximate the door; determining whether the door was opened from inside of the room or from outside of the room using a credential to actuate the access control; determining who opened the door; and generating an alert in response to who opened the door and whether the door was opened from inside of the room or from outside of the room using a credential to actuate the access control.

Abstract: Various embodiments relate to a method of producing a machine learning model with a fingerprint that maps an input value to an output label, including: selecting a set of extra input values, wherein the set of extra input values does not intersect with a set of training labeled input values for the machine learning model; selecting a first set of artificially encoded output label values corresponding to each of the extra input values in the set of extra input values, wherein the first set of artificially encoded output label values are selected to indicate the fingerprint of a first machine learning model; and training the machine learning model using a combination of the extra input values with associated first set of artificially encoded output values and the set of training labeled input values to produce the first learning model with the fingerprint.

Abstract: [Problem] To provide an information processing system, an information processing method, and a program capable of suitably tracking a moving object even when there is variation in the speed of the moving object to be tracked. [Solution] Provided are: a video acquisition unit (201) for accepting input of a video taken by a plurality of photographing devices; a reference object selection unit (207) for selecting, from among moving objects seen in the video taken by a first photographing device among the plurality of photographing devices, another moving object different from the moving object to be tracked and seen in the video taken by the first photographing device; and a target object appearance prediction unit (213) for predicting, based on whether the other moving object appeared in the video taken by a second photographing device among the plurality of photographing devices, a time of day at which the moving object to be tracked will appear.

Abstract: A TDI scanner including a dynamically programmable focal plane array including a two-dimensional array of detectors arranged in a plurality of columns and a plurality of rows, the array being divided into a plurality of banks separated from one another by gap regions, each bank including a plurality of sub-banks, and each sub-bank including at least one row of detectors, a ROIC coupled to the focal plane array and configured to combine in a TDI process outputs from detectors in each column of detectors in each sub-bank, and a controller configured to program the focal plane array to selectively and dynamically set characteristics of the focal plane array, the characteristics including a size and a location within the two-dimensional array of each of the plurality of sub-banks and the gap regions, the size corresponding to a number of rows of detectors included in the respective sub-bank or gap region.

Abstract: A method of identifying a behavior direction recognition based service requester includes a plurality of intelligent robot devices arranged in the airport and a server controlling movement of one or more intelligent robot devices of the plurality of intelligent robot devices. The intelligent robot includes a communication unit transmitting movement information of an airport user moving in the airport from the server, and a processor configured to receive the movement information of the airport user from the communication unit, learn movement of the airport user, recognize a wandering state of the airport user based on the learned movement of the airport user, and move to the airport user based on a result of the recognition. The intelligent robot device may be associated with an artificial intelligence module, an unmanned aerial vehicle (UAV), a robot, an augmented reality (AR) device, a virtual reality (VR) device, and devices related to 5G services.

Abstract: Methods, systems, and apparatus, including computer programs encoded on a storage device, for monitoring a vehicle access point. In one aspect, a monitoring system is disclosed that includes a processor and a computer storage media storing instructions that, when executed by the processor, cause the processor to perform operations. The operations may include obtaining, by the monitoring system, monitoring system data that is generated by a first monitoring system component, determining, by the monitoring system and based on the monitoring system data, a current occupancy of a vehicle, determining, by the monitoring system and based on the current occupancy of the vehicle, whether a vehicle access point is to be monitored, and in response to determining, by the monitoring system and based on the current occupancy of the vehicle, that the vehicle is not occupied by a human occupant, using one or more second monitoring system components to monitor the vehicle access point.

Abstract: Techniques are disclosed relating to increasing the amount of training data available to machine learning algorithms. A computer system may access an initial set of training data that specifies a plurality of sequences, each of which may define a set of data values. The computer system may amplify the initial set of training data to create a revised set of training data. The amplifying may include identifying sub-sequences of data values in ones of the plurality of sequences in the initial set of training data and using an inheritance algorithm to create a set of additional sequences of data values, where each one of the set of additional sequences may include sub-sequences of data values from at least two different sequences in the initial set of training data. The computer system may process the set of additional sequences using the machine learning algorithm to train a machine learning model.

Abstract: Flight based infrared imaging systems and related techniques, and in particular UAS based thermal imaging systems, are provided to improve the monitoring capabilities of such systems over conventional infrared monitoring systems. An infrared imaging system is configured to compensate for various environmental effects (e.g., position and/or strength of the sun, atmospheric effects) to provide high resolution and accuracy radiometric measurements of targets imaged by the infrared imaging system. An infrared imaging system is alternatively configured to monitor regulatory limitations on operation of the infrared imaging system and adjust and/or disable operation of the infrared imaging systems accordingly.