Abstract: In one embodiment, a service receives machine learning-based generative models from a plurality of distributed sites. Each generative model is trained locally at a site using unlabeled data observed at that site to generate synthetic unlabeled data that mimics the unlabeled data used to train the generative model. The service receives, from each of the distributed sites, a subset of labeled data observed at that site. The service uses the generative models to generate synthetic unlabeled data. The service trains a global machine learning-based model using the received subsets of labeled data received from the distributed sites and the synthetic unlabeled data generated by the generative models.

Abstract: Presented herein are techniques for training a central/global machine learning model in a distributed machine learning system. In the data sampling techniques, a subset of the data obtained at the local sites is intelligently selected for transfer to the central site for use in training the central machine learning model. In the model merging techniques, distributed local training occurs in each local site and copies of the local machine learning models are sent to the central site for aggregation of learning by merging of the models. As a result, in accordance with the examples presented herein, a central machine learning model can be trained based on various representations/transformations of data seen at the local machine learning models, including sampled selections of data-label pairs, intermediate representation of training errors, or synthetic data-label pairs generated by models trained at various local sites.

Abstract: Presented herein are techniques for automatically generating object segmentation training data. In particular, a segmentation data generation system is configured to obtain training images derived from a scene captured by one or more image capture devices. Each training image is a still image that includes a foreground object and a background. The segmentation data generation system automatically generates a mask of the training image to delineate the object from the background and, based on the mask automatically generates a masked image. The masked image includes only the object present in the training image. The segmentation data generation system composites the masked image with an image of an environmental scene to generate a composite image that includes the masked image and the environmental scene.

Abstract: In one embodiment, a device determines locations of a plurality of transmitters relative to a particular wireless access point in a wireless network. One of the transmitters comprises a target client to which the particular wireless access point is to communicate. The device compares a plurality of beamforming patterns associated with the particular wireless access point to the determined locations. The device selects, based on the comparison, one of the beamforming patterns for use by the particular wireless access point to communicate with the target client. The device controls the particular wireless access point to use the selected beamforming pattern to communicate with the target client.

Abstract: In one embodiment, a service receives machine learning-based generative models from a plurality of distributed sites. Each generative model is trained locally at a site using unlabeled data observed at that site to generate synthetic unlabeled data that mimics the unlabeled data used to train the generative model. The service receives, from each of the distributed sites, a subset of labeled data observed at that site. The service uses the generative models to generate synthetic unlabeled data. The service trains a global machine learning-based model using the received subsets of labeled data received from the distributed sites and the synthetic unlabeled data generated by the generative models.

Abstract: In one embodiment, a device determines locations of a plurality of transmitters relative to a particular wireless access point in a wireless network. One of the transmitters comprises a target client to which the particular wireless access point is to communicate. The device compares a plurality of beamforming patterns associated with the particular wireless access point to the determined locations. The device selects, based on the comparison, one of the beamforming patterns for use by the particular wireless access point to communicate with the target client. The device controls the particular wireless access point to use the selected beamforming pattern to communicate with the target client.

Abstract: In one embodiment, a device determines locations of a plurality of transmitters relative to a particular wireless access point in a wireless network. One of the transmitters comprises a target client to which the particular wireless access point is to communicate. The device compares a plurality of beamforming patterns associated with the particular wireless access point to the determined locations. The device selects, based on the comparison, one of the beamforming patterns for use by the particular wireless access point to communicate with the target client. The device controls the particular wireless access point to use the selected beamforming pattern to communicate with the target client.

Abstract: Presented herein are techniques for automatically generating object segmentation training data. In particular, a segmentation data generation system is configured to obtain training images derived from a scene captured by one or more image capture devices. Each training image is a still image that includes a foreground object and a background. The segmentation data generation system automatically generates a mask of the training image to delineate the object from the background and, based on the mask automatically generates a masked image. The masked image includes only the object present in the training image. The segmentation data generation system composites the masked image with an image of an environmental scene to generate a composite image that includes the masked image and the environmental scene.

Abstract: Presented herein are techniques for assignment of an identity to a group of captured images. A plurality of captured images that each include an image of at least one person are obtained. For each of the plurality of captured images, relational metrics indicating a relationship between the image of the person in a respective captured image and the images of the persons in each of the remaining plurality of captured images is calculated. Based on the relational metrics, a clustering process is performed to generate one or more clusters from the plurality of captured images. Each of the one or more clusters are associated with an identity of an identity database. The one or more clusters may each be associated with an existing identity of the identity database or an additional identity that is not yet present in the identity database.

Abstract: A methodology includes determining coarse location coordinates for a mobile device, anchoring the coarse location coordinates to a map, receiving inertial measurement unit data supplied by the mobile device, wherein the inertial measurement unit data is indicative of relative location coordinates of the mobile device, generating an unanchored path of the mobile device based on the relative location coordinates, and anchoring the unanchored path of the mobile device to the map in a position that optimizes a match between the coarse location coordinates and the relative location coordinates of the mobile device.

Abstract: Presented herein are techniques for automatically generating object segmentation training data. In particular, a segmentation data generation system is configured to obtain training images derived from a scene captured by one or more image capture devices. Each training image is a still image that includes a foreground object and a background. The segmentation data generation system automatically generates a mask of the training image to delineate the object from the background and, based on the mask automatically generates a masked image. The masked image includes only the object present in the training image. The segmentation data generation system composites the masked image with an image of an environmental scene to generate a composite image that includes the masked image and the environmental scene.

Abstract: A methodology includes determining coarse location coordinates for a mobile device, anchoring the coarse location coordinates to a map, receiving inertial measurement unit data supplied by the mobile device, wherein the inertial measurement unit data is indicative of relative location coordinates of the mobile device, generating an unanchored path of the mobile device based on the relative location coordinates, and anchoring the unanchored path of the mobile device to the map in a position that optimizes a match between the coarse location coordinates and the relative location coordinates of the mobile device.

Abstract: A methodology includes receiving from a first mobile device a first estimated location of the first mobile device and a first estimated error associated with the first estimated location, the first estimated location being based on first coarse data from a first wireless access point location determination system fused with inertial measurement unit (IMU) data from the first mobile device, receiving from a second mobile device a second estimated location of the second mobile device and a second estimated error associated with the second estimated location, the second estimated location being based on second coarse data from the first wireless access point location determination system fused with IMU data from the second mobile device, and based on the first estimated error and the second estimated error, determining a recommended change to a deployment of a wireless access point associated with the first wireless access point location determination system.

Abstract: Presented herein are techniques for training a central/global machine learning model in a distributed machine learning system. In the data sampling techniques, a subset of the data obtained at the local sites is intelligently selected for transfer to the central site for use in training the central machine learning model. In the model merging techniques, distributed local training occurs in each local site and copies of the local machine learning models are sent to the central site for aggregation of learning by merging of the models. As a result, in accordance with the examples presented herein, a central machine learning model can be trained based on various representations/transformations of data seen at the local machine learning models, including sampled selections of data-label pairs, intermediate representation of training errors, or synthetic data-label pairs generated by models trained at various local sites.

Abstract: Presented herein are techniques for assignment of an identity to a group of captured images. A plurality of captured images that each include an image of at least one person are obtained. For each of the plurality of captured images, relational metrics indicating a relationship between the image of the person in a respective captured image and the images of the persons in each of the remaining plurality of captured images is calculated. Based on the relational metrics, a clustering process is performed to generate one or more clusters from the plurality of captured images. Each of the one or more clusters are associated with an identity of an identity database. The one or more clusters may each be associated with an existing identity of the identity database or an additional identity that is not yet present in the identity database.

Abstract: Presented herein are techniques for automatically generating object segmentation training data. In particular, a segmentation data generation system is configured to obtain training images derived from a scene captured by one or more image capture devices. Each training image is a still image that includes a foreground object and a background. The segmentation data generation system automatically generates a mask of the training image to delineate the object from the background and, based on the mask automatically generates a masked image. The masked image includes only the object present in the training image. The segmentation data generation system composites the masked image with an image of an environmental scene to generate a composite image that includes the masked image and the environmental scene.

Abstract: A network device includes Power-over-Ethernet PoE ports to communicate with a communication network and Universal Serial Bus (USB) devices. The network device establishes an Internet Protocol (IP) related connection with an application program, and receives downlink IP packets carrying USB transaction requests destined for the USB device from the application program over the IP connection. The network device converts the received downlink IP packets carrying the USB transaction requests to downlink bus-level USB transactions that are understandable to the USB device. The network device sends the downlink bus-level USB transactions to the USB adaptor device over a PoE connection for forwarding by the USB adaptor device to the USB device.

Abstract: A Network-Universal Serial Bus (NUSB) adaptor exchanges Power-over-Ethernet (PoE) packets with, and receives power from, a Power Source Equipment (PSE) over a PoE connection with the PSE, and exchanges Universal Serial Bus (USB) messages with, and provides power to, a USB device over a USB connection with the USB device. The NUSB adaptor converts between a USB power negotiation protocol implemented between the USB device and the NUSB adaptor and a PoE power negotiation protocol implemented between the NUSB adapter and the network device.

Abstract: A network device includes Power-over-Ethernet PoE ports to communicate with a communication network and Universal Serial Bus (USB) devices. The network device establishes an Internet Protocol (IP) related connection with an application program, and receives downlink IP packets carrying USB transaction requests destined for the USB device from the application program over the IP connection. The network device converts the received downlink IP packets carrying the USB transaction requests to downlink bus-level USB transactions that are understandable to the USB device. The network device sends the downlink bus-level USB transactions to the USB adaptor device over a PoE connection for forwarding by the USB adaptor device to the USB device.

Abstract: A Network-Universal Serial Bus (NUSB) adaptor exchanges Power-over-Ethernet (PoE) packets with, and receives power from, a Power Source Equipment (PSE) over a PoE connection with the PSE, and exchanges Universal Serial Bus (USB) messages with, and provides power to, a USB device over a USB connection with the USB device. The NUSB adaptor converts between a USB power negotiation protocol implemented between the USB device and the NUSB adaptor and a PoE power negotiation protocol implemented between the NUSB adapter and the network device.