Abstract: A device for employing an efficient neural network architecture through partial convolution is provided, including a fast network module, a data input module, and an outcome module. A fast neural network including multiple fast neural network blocks with at least one PConv layer and at least two PWConv layers are integrated in the fast network module. The data input module is responsible for loading and providing input data to the fast network module. The PConv layer is applied for partial convolution of the input data with achieving standard convolution operations on partial channels while preserving other channels unaffected and selectively convolves only a portion of input channels by leveraging redundant information in feature maps. The two PWConv layers following the PConv layer are configured to transform and integrate features. The outcome module is configured to receive results generated by the fast network module.

Abstract: A system for flow-count includes one or more cameras and a processing system. The one more cameras are configured to capture image data comprising a sequence of frames. The processing system is configured to extract spatial features and temporal features based on the sequence of frames; construct, based on the spatial features and temporal features, a density map and/or a flow map corresponding to a frame of the sequence of frames, wherein the density map comprises density values for the plurality of pixels, wherein the flow map comprises two-dimensional (2D) vectors for the plurality of pixels, and wherein a 2D vector indicates a speed and direction for a pixel of the plurality of pixels in the frame; and determine, based on the density map and/or the flow map, a count, a distribution, and/or movement of objects corresponding to the frame.

Abstract: The present invention provides a signal fusion system that is a probabilistic system fusing an arbitrary combination of heterogeneous signals for determining the location of electronic devices. The system includes a signal sampling device for detecting one or more signals emitted by an electronic device to be located, including geolocation signals, WiFi signals, Bluetooth signals, 4G communication signals, 5G communication signals, geomagnetism signals, or inertial navigation system signals (INS). A likelihood processor cooperates with the signal sampling device to receive information about selected sampled signals, and creates a grid of reference points for an interested area in which the electronic device may be located. The likelihood processor independently computes, for each selected sampled signal, a location likelihood that is a probability of observing the sampled signal given that the electronic device is located at different reference points in the grid.

Abstract: A VSR approach with temporal consistency using generative adversarial networks (VistGAN) that requires only the training HR video sequence to generate the HR/LR video frame pairs, instead of the pre-artificial-synthesized HR/LR video frame pairs, for training. By this unsupervised learning method, the encoder degrades the input HR video frames of a training HR video sequence to their LR counterparts, and the decoder seeks to recover the original HR video frames from the LR video frames. To improve the temporal consistency the unsupervised learning method provides a sliding window that explores the temporal correlation in both HR and LR domains. It keeps the temporal consistent and also fully utilizes high-frequency details from the last-generated reconstructed HR video frame.

Abstract: An architecture is provided for cooperative target tracking and signal propagation learning using mobile sensors. A method can comprise as a function of sensing data representative of a location of a target device at a first defined moment and model data relating to a motion model representing a probability density function, determining, by a system comprising a processor, a group of locations for the target device at a second defined time point, wherein the probability density function facilitates determining, based on the location of the target device at the first defined moment, a current location of the target device at a third defined moment; and as a function of the group of locations, generating, by the system, a data structure representing a matrix of received signal strength values; and identifying, by the system, a location of the group of locations for the target device at the third defined moment based on the data structure.

Abstract: Various embodiments disclosed herein enable a mobile device to accurately locate where the mobile device is located indoors by measuring geomagnetic fields, and tracing relative changes in the magnetic fields as the mobile device is moved. The relative changes can be compared to a fingerprint signal map, and a trace of a portion of the path can be determined by matching the relative changes in the magnetic fields to relative changes in the fingerprint signal map. A collection of the path portions can then be connected by determining a shortest path that connects each of the path portions. In another embodiment, a step counter can also be used to constrain possible locations, and fuse the geomagnetic field and step counter information for joint indoor localization.

Abstract: The present invention provides a signal fusion system that is a probabilistic system fusing an arbitrary combination of heterogeneous signals for determining the location of electronic devices. The system includes a signal sampling device for detecting one or more signals emitted by an electronic device to be located, including geolocation signals, WiFi signals, Bluetooth signals, 4G communication signals, 5G communication signals, geomagnetism signals, or inertial navigation system signals (INS). A likelihood processor cooperates with the signal sampling device to receive information about selected sampled signals, and creates a grid of reference points for an interested area in which the electronic device may be located. The likelihood processor independently computes, for each selected sampled signal, a location likelihood that is a probability of observing the sampled signal given that the electronic device is located at different reference points in the grid.

Abstract: A VSR approach with temporal consistency using generative adversarial networks (VistGAN) that requires only the training HR video sequence to generate the HR/LR video frame pairs, instead of the pre-artificial-synthesized HR/LR video frame pairs, for training. By this unsupervised learning method, the encoder degrades the input HR video frames of a training HR video sequence to their LR counterparts, and the decoder seeks to recover the original HR video frames from the LR video frames. To improve the temporal consistency the unsupervised learning method provides a sliding window that explores the temporal correlation in both HR and LR domains. It keeps the temporal consistent and also fully utilizes high-frequency details from the last-generated reconstructed HR video frame.

Abstract: Constructing and/or recovering paths of mobile devices is enabled. For instance, a method comprises: receiving, by wireless fidelity (Wi-Fi) sensors, respective probes from mobile devices, grouping probes having a same media access control (MAC) address into segments, identifying fingerprints of information elements the mobile devices, grouping segments according to an identified fingerprint, determining an increment of a sequence number corresponding to consecutive segments of a segment group, determining a time gap between the sequence number corresponding to a first probe and an incremented sequence number corresponding to a second probe having a timestamp that is later in time than the first probe, and comparing a growth rate of the sequence number corresponding to the consecutive segments to determine a forward segment of the consecutive segments, resulting in a constructed path of the mobile devices, and storing the constructed path in a database.

Abstract: Various embodiments disclosed herein enable a mobile device to accurately locate where the mobile device is located indoors by measuring geomagnetic fields, and tracing relative changes in the magnetic fields as the mobile device is moved. The relative changes can be compared to a fingerprint signal map, and a trace of a portion of the path can be determined by matching the relative changes in the magnetic fields to relative changes in the fingerprint signal map. A collection of the path portions can then be connected by determining a shortest path that connects each of the path portions. In another embodiment, a step counter can also be used to constrain possible locations, and fuse the geomagnetic field and step counter information for joint indoor localization.

Abstract: An architecture is provided for cooperative target tracking and signal propagation learning using mobile sensors. A method can comprise as a function of sensing data representative of a location of a target device at a first defined moment and model data relating to a motion model representing a probability density function, determining, by a system comprising a processor, a group of locations for the target device at a second defined time point, wherein the probability density function facilitates determining, based on the location of the target device at the first defined moment, a current location of the target device at a third defined moment; and as a function of the group of locations, generating, by the system, a data structure representing a matrix of received signal strength values; and identifying, by the system, a location of the group of locations for the target device at the third defined moment based on the data structure.

Abstract: Constructing and/or recovering paths of mobile devices is enabled. For instance, a method comprises: receiving, by wireless fidelity (Wi-Fi) sensors, respective probes from mobile devices, grouping probes having a same media access control (MAC) address into segments, identifying fingerprints of information elements the mobile devices, grouping segments according to an identified fingerprint, determining an increment of a sequence number corresponding to consecutive segments of a segment group, determining a time gap between the sequence number corresponding to a first probe and an incremented sequence number corresponding to a second probe having a timestamp that is later in time than the first probe, and comparing a growth rate of the sequence number corresponding to the consecutive segments to determine a forward segment of the consecutive segments, resulting in a constructed path of the mobile devices, and storing the constructed path in a database.

Abstract: When access point signals are altered, conventional fingerprint-based indoor localization techniques are note accuracy. Localization with altered access point and fingerprint updating can achieves accurate indoor localization and automatically update a fingerprint database with altered access points. Using subset sampling, the system detect the altered access points, filter them out by a received signal strength vector and find the location of a client. Given the received signal strength vectors received and the estimated location, the system can update a fingerprint database with the signal changes by applying a non-parametric Gaussian process regression method.

Abstract: When access point signals are altered, conventional fingerprint-based indoor localization techniques are note accuracy. Localization with altered access point and fingerprint updating can achieves accurate indoor localization and automatically update a fingerprint database with altered access points. Using subset sampling, the system detect the altered access points, filter them out by a received signal strength vector and find the location of a client. Given the received signal strength vectors received and the estimated location, the system can update a fingerprint database with the signal changes by applying a non-parametric Gaussian process regression method.

Abstract: Locating mobile devices based on a junction of signal tiles is disclosed herein. Mobile devices are mapped to a convex hull within which the mobile device is likely to be located in based on a received signal strength indicator (RSSI) associated with an access point (AP). Using the RSSI from a set of APs, the mobile device can be located within a junction of a set of convex hulls, or tiles using a linear programming function. To mitigate signal noise and improve the accuracy of the location determination, the localization system can use the RSSIs of the APs that have signals with a predetermined signal dynamic range that leads to smaller tiles. Additionally, the localization system can reduce the search complexity of matching fingerprints to APs by partitioning the space into clusters with reference points that have similar signal vectors.

Abstract: Video on demand is orchestrated between systems by linear program (LP) based multimedia storage and retrieval. An LP based storage and retrieval system can partition multimedia data into an integral number of segments. Further, the LP based storage and retrieval system can store a first amount of the integral number of segments in a storage device of storage devices of a storage network in response to a determination that a storage size of the storage device satisfies a first condition with respect to the first amount of the integral number of segments. Furthermore, such system can store a segment of a second amount of the integral number of segments in a remote storage device of the remote storage devices in response to a determination that the segment satisfies a second condition with respect to a defined number of remote storage devices of the storage devices that have stored the segment.

Abstract: Distributive content delivery techniques are applied in a content delivery system. A content delivery overlay is generated as a function of delay times, bandwidth requirements, and throughput of a network. Helpers are added to the content delivery network as a function of delay times, bandwidth requirements, and throughput. Further, content can be transmitted without exchanging buffermaps or waiting for whole packets to be transmitted.

Abstract: Locating mobile devices based on a junction of signal tiles is disclosed herein. Mobile devices are mapped to a convex hull within which the mobile device is likely to be located in based on a received signal strength indicator (RSSI) associated with an access point (AP). Using the RSSI from a set of APs, the mobile device can be located within a junction of a set of convex hulls, or tiles using a linear programming function. To mitigate signal noise and improve the accuracy of the location determination, the localization system can use the RSSIs of the APs that have signals with a predetermined signal dynamic range that leads to smaller tiles. Additionally, the localization system can reduce the search complexity of matching fingerprints to APs by partitioning the space into clusters with reference points that have similar signal vectors.

Abstract: Systems, methods and apparatus facilitating joint channel and routing assignment are provided. In some embodiments, a system includes nodes in a network configured to perform distributed joint channel and routing assignment of traffic through exchange of information with one another. The assignments are iteratively updated based on whether the update will result in an increase in a value of a function associated with throughput for traffic in the network. In some embodiments, the assignments are determined based on random decisions that are adopted if the function is improved utilizing the routing and channel information in the assignment. In some embodiments, the assignments are determined based on traffic prioritization in which destination nodes having a high level of traffic are assigned routes and channels having characteristics amenable to reduction of interference in the network.

Abstract: Embodiments of the present invention disclose a cooperative caching method and apparatus, relating to the field of network technologies, to improve the local hit ratio without increasing the local server costs. The technical solution provided in the present invention includes: obtaining, according to cache information, end-to-end delay between a local server and a neighbor server, and popularity in a cache list, a consolidated gain value of a cached video segment and a consolidated gain value of a candidate video segment in the local server; and replacing the cached video segment with the candidate video segment when the consolidated gain value of the cached video segment and the consolidated gain value of the candidate video segment in the local server meet a replacement condition.