Abstract: A convolutional neural network framework is described that uses reverse connection and obviousness priors for object detection. A method includes performing a plurality of layers of convolutions and reverse connections on a received image to generate a plurality of feature maps, determining an objectness confidence for candidate bounding boxes based on outputs of an objectness prior, determining a joint loss function for each candidate bounding box by combining an objectness loss, a bounding box regression loss and a classification loss, calculating network gradients over positive boxes and negative boxes, updating network parameters within candidate bounding boxes using the joint loss function, repeating performing the convolutions through to updating network parameters until the training converges, and outputting network parameters for object detection based on the training images.

Abstract: An example apparatus for mining multi-scale hard examples includes a convolutional neural network to receive a mini-batch of sample candidates and generate basic feature maps. The apparatus also includes a feature extractor and combiner to generate concatenated feature maps based on the basic feature maps and extract the concatenated feature maps for each of a plurality of received candidate boxes. The apparatus further includes a sample scorer and miner to score the candidate samples with multi-task loss scores and select candidate samples with multi-task loss scores exceeding a threshold score.

Abstract: Method and apparatus for load balancing in a Cloud-radio access network (C-RAN) are disclosed. A method includes receiving control plane (CP) data associated with a user from a core network or a remote access point; and dispatching the CP data to a first user equipment (UE) virtualized network function component (VNFC) based on a first route.

Abstract: An example apparatus for mining multi-scale hard examples includes a convolutional neural network to receive a mini-batch of sample candidates and generate basic feature maps. The apparatus also includes a feature extractor and combiner to generate concatenated feature maps based on the basic feature maps and extract the concatenated feature maps for each of a plurality of received candidate boxes. The apparatus further includes a sample scorer and miner to score the candidate samples with multi-task loss scores and select candidate samples with multi-task loss scores exceeding a threshold score.

Abstract: A system and method determines allergens which may present a risk of a respiratory attack to a user of the system. Historical data which precedes a respiratory attack is statistically analyzed. The data relates to the locations where the user has been present and includes environmental factors and optionally also human or animal contact factors. A location at which a respiratory attack is most likely to have been triggered is derived as well 5 as a set of most likely plant types (or animal types, mold or dust mites) to have caused the respiratory attack.

Abstract: The invention provides a system for determining a risk level of a pollen-induced allergy of a user. The system comprises an input for receiving information relating to a pollen level in a location, an input for receiving information relating to a particulate matter level in the location and an input for receiving information relating to a sensitivity of a user to pollen allergen. The system further comprises a processor which is adapted to determine the risk level by taking account of the information relating to a pollen level, the information relating to a particulate matter level and the information relating to a sensitivity of the user to pollen allergen. A user interlace communicates the risk level in the location to the user.

Abstract: A system is for estimating a remaining lifetime of an aldehyde filter. The system comprises aa aldehyde filter (14) through which at least a portion of the gas is to be passed for removing aldehyde (10) from the gas (12) and a detecting medium (16) through which at least a portion of the gas is to be passed comprising photoluminescent carbon-based dots (18), a light source for emitting an excitation light (E) through the detecting medium for exciting the carbon-based dots which thereby emit luminescent light (L), a detector (28) for detecting the luminescent light (L), the luminescent light (L) having a luminescence property; and a controller (30) for determining information relating to intensity of a red, a green or a blue component of the luminescence property and estimating a remaining lifetime of the aldehyde filter from the determined information.

Abstract: A convolutional neural network framework is described that uses reverse connection and obviousness priors for object detection. A method includes performing a plurality of layers of convolutions and reverse connections on a received image to generate a plurality of feature maps, determining an objectness confidence for candidate bounding boxes based on outputs of an objectness prior, determining a joint loss function for each candidate bounding box by combining an objectness loss, a bounding box regression loss and a classification loss, calculating network gradients over positive boxes and negative boxes, updating network parameters within candidate bounding boxes using the joint loss function, repeating performing the convolutions through to updating network parameters until the training converges, and outputting network parameters for object detection based on the training images.

Abstract: Region proposal is described for image regions that include objects of interest. Feature maps from multiple layers of a convolutional neural network model are used. In one example a digital image is received and buffered. Layers of convolution are performed on the image to generate feature maps. The feature maps are reshaped to a single size. The reshaped feature maps are grouped by sequential concatenation to form a combined feature map. Region proposals are generated using the combined feature map by scoring bounding box regions of the image. Objects are detected and classified objects in the proposed regions using the feature maps.

Abstract: Embodiments of the present invention disclose a method and an apparatus for service offloading between different networks. By means of capability negotiation between an ANDSF unit and user equipment, a network side controls the user equipment to send an ISRP request to the ANDSF unit; a non-3GPP network is accessed by using an instruction, generated by the ANDSF unit according to the ISRP request, for instructing to access the non-3GPP network; and after the UE is simultaneously connected to a 3GPP network and the non-3GPP network, services of the user equipment that camp on the 3GPP network and services of the user equipment that camp on the non-3GPP network are indicated according to an ISRP offloading policy that is delivered by the ANDSF unit according to the ISRP request.

Abstract: Embodiments of the present disclosure provide a network access processing method and apparatus. The method includes: receiving traffic information and network load information that are reported by multiple access networks; determining a target access network in the multiple access networks according to the traffic information and the network load information; and when a current access network of a user terminal is different from the target access network: if the user terminal supports an inter-system mobility policy (ISMP), handing over the user terminal to the target access network; or if the user terminal supports an inter-system routing policy (ISRP) and it is determined that at least one application on the user terminal can be handed over to the target access network, handing over the at least one application to the target access network. In this way, properness of selecting an access network for a terminal or a terminal application can be improved.

Abstract: Embodiments of the present invention disclose a method and an apparatus for service offloading between different networks. By means of capability negotiation between an ANDSF unit and user equipment, a network side controls the user equipment to send an ISRP request to the ANDSF unit; a non-3GPP network is accessed by using an instruction, generated by the ANDSF unit according to the ISRP request, for instructing to access the non-3GPP network; and after the UE is simultaneously connected to a 3GPP network and the non-3GPP network, services of the user equipment that camp on the 3GPP network and services of the user equipment that camp on the non-3GPP network are indicated according to an ISRP offloading policy that is delivered by the ANDSF unit according to the ISRP request.

Abstract: Embodiments of the present disclosure provide a network access processing method and apparatus. The method includes: receiving traffic information and network load information that are reported by multiple access networks; determining a target access network in the multiple access networks according to the traffic information and the network load information; and when a current access network of a user terminal is different from the target access network: if the user terminal supports an ISMP, handing over the user terminal to the target access network; or if the user terminal supports an ISRP and it is determined that at least one application on the user terminal can be handed over to the target access network, handing over the at least one application to the target access network. In this way, properness of selecting an access network for a terminal or a terminal application can be improved.

Abstract: Embodiments of the present invention provide an access processing method, an apparatus, and a system, where the method includes: generating, by a capability enabling gateway, a RAT selection policy according to context information of a UE, and sending a RAT selection request that includes the RAT selection policy to an inter-RAT coordination controller corresponding to a cell in which the UE is located, so that the inter-RAT coordination controller determines an access standard of the UE according to the RAT selection policy. In this way, network utilization is improved, and service quality can be ensured and improved for an end user.

Abstract: Embodiments of the present invention provide a system architecture, a subsystem, and a method for opening a telecommunications network capability, where the system includes: an API open subsystem, a telecommunications network subsystem, and a third-party subsystem. A first interface exists between the telecommunications network subsystem and the API open subsystem, and a second interface exists between the API open subsystem and the third-party subsystem; the telecommunications network subsystem is configured to provide a telecommunications network capability for the API open subsystem; the API open subsystem is configured to operate and manage opening of the telecommunications network capability and provide an API of the telecommunications network capability for the third-party subsystem; and the third-party subsystem is configured to invoke, on the API open subsystem, the API of the telecommunications network capability.

Abstract: A method and a system for distributing key of media stream are provided. The method comprises: determining, by a security management server, whether a domain to which the calling terminal belongs and a domain to which a called terminal belongs subscribe a key distribution protocol; generating a key based on encryption capability information obtained in a calling process, and distributing the generated key to the calling terminal and the called terminal, if the protocol has been subscribed; generating a key based on encryption capability information obtained in a calling process, and distributing the generated key to the terminal at the same side as the security management server, if the protocol has not been subscribed. By applying the disclosure, the key is generated by the security management server, so that complexity may be reduced for media stream key negotiation to facilitate promotion of media stream encryption service.

Abstract: A method and a system for distributing key of media stream are provided. The method comprises: determining, by a security management server, whether a domain to which the calling terminal belongs and a domain to which a called terminal belongs subscribe a key distribution protocol; generating a key based on encryption capability information obtained in a calling process, and distributing the generated key to the calling terminal and the called terminal, if the protocol has been subscribed; generating a key based on encryption capability information obtained in a calling process, and distributing the generated key to the terminal at the same side as the security management server, if the protocol has not been subscribed. By applying the disclosure, the key is generated by the security management server, so that complexity may be reduced for media stream key negotiation to facilitate promotion of media stream encryption service.