37 results about "Loss network" patented technology

An automation network includes automation network devices connected to the network and a system layer interface that interfaces with a transport layer and an application layer of the home automation network. The system layer interface includes command libraries configured to upgrade a remote network device. The automation network queries the network devices to determine if there are lost network devices or newly added network devices. The automation network may update a new remote device with scene information related to any lost network devices.

A method of authenticating a user access to network stations is disclosed. Users of the new authentication system do not need to input passwords to gain access to the network stations for on-line transactions, as the authentication job is handled by the authentication server and the net entry apparatus through a host computer. A token is generated dynamically and sent to the application server to which the user intends to gain access, and the verification process is then activated between the authentication server and the application server, which then retrieves a symmetrical copy of the token to compare with the token passed from the application server. If both tokens match up, the user ID has passed the security check. Users are freed from having to memorize different user IDs and passwords to operate many network accounts, with no risk of losing network account numbers and passwords.

A method of an efficient adaptive mode selection for H.264 / AVC-coded video delivery in burst-packet-loss networks to generate the image data of the missing macroblocks in the decoded current frame by using the information related to the spatial redundancy in the same frame and the temporal redundancy in the inter frames is disclosed. The method first employs the Intra High-Speed Spatial Error Concealment (SEC) method for the initial frame. For the succeeding inter frames, the Temporal Error Concealment (TEC) method is used when the Adjacent External Boundary Matching Error (AEBME) of the surrounding macroblocks is not more than the dynamic threshold (DTα). The Intra High-Speed SEC is used when AEBME is more than (DTα) and the surrounding macroblocks are all coded intra-method, otherwise the task proceeds by employing Normal SEC.

The type of audio stored in the payload of a data packet transmitted over a data network is identified as speech audio or non-speech audio through the use of a non-speech identifier included in a header in the data packet. Upon detection of data packet containing non-speech audio, the receiver of the data packet may modify jitter buffer latency while the non-speech audio is being received. Modifying the jitter buffer latency while non-speech audio is being received minimizes the loss of spoken words during jitter buffer latency modification.

The invention relates to a network search method and relative mobile terminal, wherein said method comprises: setting the network search period of mobile terminal; when processing network search at the network loss process, said network search period can increase with the increase of network search times; said network search period has one initial value; when the network is in lost state, the mobile terminal based on the network search times in the network loss process, fixes present network search period, and based on present network search period, operates the network search. In said invention, the mobile terminal can process network search based on the increase network search period that increases with the search times, to make the mobile terminal quickly find network at the instant that losing network, and reduce the network search frequency of mobile terminal when the network quality is worse, effectively reduce the working load and reduce the electricity consumption of mobile terminal.

A method for enabling communication when network connectivity is reduced or lost during a conversation and for resuming the same conversation when connectivity improves.

The invention proposes a three-dimensional style transfer model based on a double-path stylized network, and the main contents of the method comprise the double-path stylized network and a loss network. The process of the method is that a three-dimensional style transfer model consists of two parts: the double-path stylized network and the loss network; the double-path stylized network employs a stereo pair, and each view is processed in a single path; a feature fusion block is embedded into the stylized network, thereby achieving the effective sharing of the feature-level information betweenthe two paths; the loss network calculates the sensing loss and the multilayer view loss, so as to coordinate the training of the two paths of the stylized network, and generate a stylization result with the consistent views. The invention proposes the novel feature fusion block for transmitting the information to the other path from one path, and the method can generate the stylization result with the better view consistency. Moreover, the quality of an image cannot be affected.

The invention discloses an intrinsic image decomposition method and device based on deep learning. The method comprises that a 3D model is selected from multiple 3D models and loaded with a physics based renderer, illumination is set randomly, an image is rendered at a random visual angle, the renderer obtains corresponding reflection and illumination components, and above operations are repeatedto generate a large batch of data sets with marked intrinsic image decomposition; a full convolutional neural network is trained into an intrinsic image decomposition network via the generated data sets; and the intrinsic image decomposition network is used, and a decomposition target of expected output is obtained from a decomposition result of predicted output. Via the intrinsic image decomposition method, a large batch of marked data sets is obtained in an image rendering manner, the deep neural network is trained to obtain a decomposition model of high robustness, and the loss network is used to further improve the generalization performance and avoid difficulty in design of the loss function.

The embodiment of the invention discloses a training method and apparatus of network models, equipment, and a storage medium. The training method of network models includes the steps: when a first network model achieves the preset update stopping condition, according to the update result of the first network model, determining the first target network model, and inserting a relay loss network layer in the preset pooling layer in the first target network model so as to determine the second network model; according to the second network model and the relay loss network layer, determining the relay loss function corresponding to the relay loss network layer; according to the second network model and the relay loss function, determining the global loss function of the second network model; andapplying the relay loss function and the global loss function to update the parameters of the second network model so as to obtain the updated second network model. The training method and apparatusof network models can solve the problem that during the network model training process, the high-layer characteristics are over fitting while the low layer characteristics are poor fitting, so that the network model training can become more thoroughly and has higher accuracy.

The invention discloses a rapid image style migration method based on group normalization, and the method comprises the steps: inputting an input image x into a generator network, generating an imageyk, and outputting the image yk; inputting the output image yk and the original content image yc into a loss network, and calculating to obtain a content loss function; inputting the output image yk and the original style image ys into a loss network, and calculating to obtain a style loss function; linearly combining the content loss function and the style loss function into a new loss function as a total loss function; using an MS-COCO data set as a content image for training; and obtaining a result image after style migration. According to the method, batch normalization is replaced by group normalization, the performance of the deep neural network is remarkably improved, the convergence speed of the model is increased in image style migration work, a high-quality image is obtained, anda framework combining a generator network and a loss network is adopted when the rapid image style migration model is constructed, so that the purpose of real-time performance is achieved.

The invention discloses a FTF-based complex product quality loss network analysis method, which comprises the following steps: selecting one or a plurality of fault modes corresponding to severity grades as the top events of a fault tree by utilizing the severity grades in a product FMECA (Failure Mode Effects and Criticality Analysis) result by virtue of a forward FMECA-FTA (Fault Tree Analysis) method, establishing the fault tree diagram of a system, and carrying out qualitative and quantitative analysis by utilizing bottom event fault rate data and fault loss data obtained by a perniciousness analysis result in the FEMCA; then, analyzing different network structures by utilizing an established fault distribution diagram; drawing an equivalent fault tree containing virtual top events by utilizing a minimum cutset principle, and carrying out loss measurement on the basis of a fault mode visual angle; constructing a conversion rule of FTA and GERT (Graphical Evaluation and Review Technique), converting a multi-FTA network into a GERT network to carry out the steady-state analysis and the significance and sensitivity analysis of the network loss so as to obtain a key node of a loss network.

A method for suspending and resuming a connection for desktop virtualization between two computing devices. In response to a client computing device shutting down, suspending, hibernating, or losing network connectivity during virtualization, the server computing device may itself shut down, suspend, or hibernate, or may pause or suspend the operation of one or more applications currently hosted by the server computing device. The server may detect that connectivity has been restored and resume operation of hosted applications. Alternatively, the client may transmit a command to the server indicating that the client is ready to resume virtualization. The client may also be configured to transmit a command that may cause the server to resume a powered-on state after the server was shut down or in a state of hibernation.

The invention relates to a bamboo cane surface defect detection method based on a triple loss network. The method comprises the following steps: S1, collecting bamboo cane surface defect data througha camera mounted on a bamboo cane sorting robot to form a bamboo cane surface defect data set; S2, performing triple loss network training on the images in the bamboo cane surface defect data set to obtain an anchor point prediction matrix, a size prediction matrix and a thermodynamic diagram prediction matrix; S3, calculating anchor point prediction matrix loss, size prediction matrix loss and thermodynamic diagram prediction matrix loss to obtain total loss of triple loss network updating; S4, continuously updating and optimizing to obtain an optimal convolution weight parameter and an optimal offset parameter based on the updated total loss of the triple loss network calculated in the step S3; and S5, obtaining an anchor point prediction matrix and a size prediction matrix by updating the triple loss network of the convolution weight parameter and the offset parameter of the test image in the bamboo cane surface defect data set, and then obtaining the category and the size of the detected target defect.

A system and method for protecting from the loss of data in an optical data network includes receiving the data over a service optical fiber line, delaying reception of the data over a protection optical fiber line by a first delay amount with respect to the reception of the data over the service optical fiber line, and detecting a fault condition in the service optical fiber line. In response to the detection of the fault condition, the transmission of data over the protection optical fiber line is received. The first delay amount corresponds to at least the amount of time to switch to the reception of the data over the protection optical fiber line from the reception of the data over the service optical fiber line after the detection of the fault condition.

A method of an efficient adaptive mode selection for H.264 / AVC-coded video delivery in burst-packet-loss networks to generate the image data of the missing macroblocks in the decoded current frame by using the information related to the spatial redundancy in the same frame and the temporal redundancy in the inter frames is disclosed. The method first employs the Intra High-Speed Spatial Error Concealment (SEC) method for the initial frame. For the succeeding inter frames, the Temporal Error Concealment (TEC) method is used when the Adjacent External Boundary Matching Error (AEBME) of the surrounding macroblocks is not more than the dynamic threshold (DTα). The Intra High-Speed SEC is used when AEBME is more than (DTα) and the surrounding macroblocks are all coded intra-method, otherwise the task proceeds by employing Normal SEC.

The invention relates to a multichannel heat exchanger flow distribution prediction method based on computational fluid dynamics (CFD), an artificial neural network and a hydraulic loss network model.Aiming at any multi-channel heat exchanger, the multi-channel heat exchanger is divided into an inlet part, a heat exchange core body and an outlet part, a hydraulic loss network is established, andhydraulic loss of each channel comprises inlet local loss, on-way resistance loss and outlet local loss. The internal on-way loss of the channel is predicted through an empirical correlation; obtaining import and export local loss coefficient sample data through a CFD method. An import and export local loss coefficient prediction model is established through an artificial neural network. The initial flow distribution is given. The total hydraulic loss prediction values of all the channels are consistent through iterative computation, and then the flow distribution can be determined. Accordingto the prediction method, the flow distribution of the multi-channel heat exchanger under different working conditions can be accurately and efficiently predicted, the flow non-uniformity is effectively evaluated, and reference is provided for engineering application.

The invention provides a congestion detection method and system for a loss-free network, and the method comprises the steps: obtaining a port state switching moment difference value according to a current timestamp and a port closing state ending timestamp; and obtaining a congestion detection result of the loss-free network according to the port state switching moment difference value and the switch port state at the current moment. By detecting the duration of the switch port in the opening period and the evolution characteristics of the data packet queue length, switching among three states of the switch port in the loss-free network can be accurately detected and identified, and the data packet is correctly marked, so that the congestion control algorithm can more efficiently complete speed regulation decision, and the flow completion time and throughput are improved.

The invention discloses an article anti-loss method and system and electronic tag. The anti-loss method includes the following steps: an intelligent mobile terminal establishes a circular linked list according to the total number of electronic tags, the circular linked list includes the serial number and Bluetooth ID of each electronic tag and the article name corresponding to each electronic tag, all the electronic tags establish a Bluetooth connection according to the circular linked list to form an anti-loss network, the electronic tag with the smallest serial number detects a situation of Bluetooth connection with the electronic tags with other serial numbers in real time, and if a certain electronic tag cannot be connected, an article corresponding to the electronic tag is judged to be lost and an alarm is given. The intelligent mobile terminal is no longer an indispensable element in the anti-loss network, the practical range of the article anti-loss system is expanded, and the flexibility of the anti-loss network is improved. The electronic tags can also detect the state of the intelligent mobile terminal, and if the intelligent mobile terminal is lost, the electronic tag with the smallest serial number gives an alarm and prompt, thereby realizing mutual detection of the electronic tags and the intelligent mobile terminal and an alarm and prompt are given.

The invention provides a method of network information acquisition which comprises the following steps: receiving voice information sent by network users when they lost network information, recognizing voice features of the voice information, judging whether the voice features meet a first preset condition, and sending the network information which is prestored by the network users and corresponding to the voice features to the network users when the voice information is judged to meet the first preset condition. With the adoption of the method of network information acquisition, when the network users lost the network information, the lost network information can be retrieved in time by a mode of sending the voice information to acquire the prestored network information by the network users.

The invention relates to a low-loss network type optical receiver and a control method thereof. The low-loss network type optical receiver successively comprises a photoelectric conversion circuit, a first amplification circuit, an automatic control circuit, a second amplification circuit, an attenuation control circuit, an equalization control circuit, a third amplification circuit, a temperature compensation circuit, a fourth amplification circuit and a level demodulation circuit, and further comprises a power supply module, a display circuit, a button circuit, an ARM chip network management controller, an Ethernet network and a computer, wherein the photoelectric conversion circuit, the level demodulation circuit and the button circuit input signals to the ARM chip network management controller, the ARM chip network management controller outputs control signals to the automatic control circuit, the attenuation control circuit and the equalization control circuit, the circuits employ low-loss chips, the amplification circuits realize multistage amplification, the heating capacity of a product is reduced, and the service life of the product is prolonged. The optical power receiving scope is large, the output level is controllable, the construction debugging is convenient, the operation state of equipment can be monitored in real time, and remote management can be realized.

The invention discloses a network threat attack feature unified quantification method based on style migration, and is used for carrying out unified processing on native feature maps of different dimensions. The method comprises the following steps: 1) data acquisition: acquiring network flow data in real time; 2) selecting basic characteristics of the network flow data; 3) converting the basic characteristics of the selected network flow into native characteristic graphs of different dimensions by using Minkowski Distance; 4) establishing a generative network by adopting a CNN, and inputtingthe "native feature map" in the data set into the "generative network" to generate a "result map"; 5) extracting high-level features of the native feature map through a loss network and making loss calculation of generated results with target "style graph" and "native feature graph" respectively; and 6) characteristic unified quantification: adjusting the weight of the generative network accordingto the loss value calculated in the step 5) so as to automatically extract the high-level characteristics of the multi-dimensional network threat data and also simultaneously realize unified quantification processing on multiple threat sources.

The invention provides a congestion judgment method in a loss-free network. The method comprises the following steps of pausing sending of a data packet when a PAUSE frame of a next-hop device arrives; when a RESUME frame of the next-hop device arrives, recording the number of the data packets in a current sending queue as the reference number, and recovering the sending of the data packet; and after the number of the sent data packets exceeds the reference number, acquiring a first congestion detection index when each data packet is sent continuously, and if the first congestion detection index exceeds a first preset threshold, carrying out congestion marking on the sent data packet. According to the embodiments of the invention, the defects that the accuracy is low and the flow independent of the congestion is easily influenced due to the fact that whether a congestion mark is added to the data packet is judged only according to the length of the queue are overcome, the defects thatPFC is easy to generate a congestion tree and is rapidly expanded are overcome, the method has an advantage of being more accurate in adding the congestion mark, and a foundation is provided for further judgment and congestion processing of the congestion flow.

The present invention provides a brute force face feature point recognition method based on a large-scale classifier. The main content of the method comprises the steps of: regression analysis throughadoption of large-scale classification, a solution strategy of classification problems, a preprocessing process and a post-processing process. The method concretely comprises the steps of: decomposing all the possible prediction conditions into a plurality of discrete categories, wherein the step consists of clustering and probability reasoning; employing a multi-tag frame to train a large-scalemulti-class network, and employing an Softmax network layer to replace a loss network layer to cope with a decision boundary problem; and finally, employing a preprocessing process to learn a linear bounding box regressor variable, and adding a post-processing regression step to improve the efficiency. Based on the large-scale classifier, the brute force face feature point recognition method is provided to avoid the condition that the regression problem has many solutions, and is higher in robustness and more accurate in face feature point recognition rate.

The invention discloses a lane line detection method based on an improved SCNN network, and the method comprises the steps: selecting a CULane lane line data set as a training data set for the training of the improved SCNN network; constructing an improved SCNN network, classifying different lane line features, and outputting lane line candidate points; classifying the lane lines by adopting a lane line seed point classification method; and based on the classified lane line candidate points obtained by the lane line seed point classification method, using a weighted least square method to carry out fitting on the lane line containers to obtain a lane line model of each lane line container so as to realize accurate positioning of the lane lines. According to the method, the processing speedof the network structure is improved on the premise that the accuracy of the SCNN network framework is not lost, and the method has good robustness and real-time performance, can be popularized and applied to systems such as a driving assistance system and has wide practicability.

The invention discloses a lane line detection method based on the improved SCNN network, by selecting the CULane lane line data set as the training data set for the improved SCNN network training; constructing the improved SCNN network, for different lane line features Classify and output the lane line candidate points; use the lane line seed point classification method to classify the lane lines; based on the lane line seed point classification method, the classified lane line candidate points are obtained using the weighted least squares method The lane line container is divided into fittings, and the lane line model of each lane line container is obtained, so as to realize the precise positioning of the lane line. The present invention improves the processing speed of the network structure without losing the accuracy of the SCNN network framework, has good robustness and real-time performance, can be popularized and applied in systems such as driving assistance systems, and has wide practicability.

The invention discloses a low-network-loss distribution ring network open-loop operation site selection method. The method comprises the steps that: obtaining an equivalent circuit of a distribution ring network, introducing a flexible interconnection switch, and dividing the switch into two systems; obtaining an initial line loss model of the power distribution network; obtaining a power distribution network line loss model; obtaining an optimal current active component and an optimal current reactive component output by the flexible interconnection switch under minimum line loss according tothe line loss mathematical expression of the power distribution network; obtaining a resistance distance, an active current moment and a reactive current moment of each node in the two divided systems; calculating the combined active current moment and the combined reactive current moment of the two systems, and obtaining the relationship between the combined active current moment and the reactive current moment and the optimal current active and reactive components; and determining an optimal opening point of the open-loop operation mode of the power distribution network with minimum networkloss as an objective according to the optimal current active component. According to the method of the invention, the relation between the loss of the power distribution network and load distributionimbalance degree can be obtained through the combined current moments, so that the optimal open-loop operation mode with the minimum loss network of the power distribution network is obtained.

Methods and systems are described for downloading media assets efficiently and quickly before loss of network connectivity. The method may allow a user to download portions of a media asset in different versions to download a maximum portion of the media asset before loss of network connectivity. Certain versions of media assets may be downloaded more quickly than other versions of media assets. The method may also allow a user to view partially downloaded media assets without network connectivity. The method may also replace portions of media assets downloaded in a first version with portionsof the media assets in a second version once network connectivity is restored.

The invention discloses a model training method and device, a gesture recognition method and device, equipment and a medium. In the embodiment of the invention, when a gesture recognition model is trained, loss function values corresponding to loss networks are determined respectively based on at least two loss networks in the gesture recognition model; and determining a joint loss function value according to the loss function values corresponding to the at least two loss networks, and training the gesture recognition model based on the joint loss function value. On the premise that the number of layers of the network is not increased, at least two loss networks are combined to supervise the training network, the precision of the gesture recognition model can be remarkably improved, and then the precision of gesture recognition is improved under limited computing power.

The invention provides a low-loss network information transmission device. The low-loss network information transmission device comprises an enclosure, a circuit board and several pins; a network information transmission circuit is arranged on the circuit board; the network information transmission circuit comprises a first transformer, a second transformer and a signal control chip; two input ends of the signal control chip are separately connected with two input ends of the first transformer; the first transformer comprises a first magnetic core, a first primary coil wound on the first magnetic core, and a first secondary coil wound on the first magnetic core; the first primary coil is a stranded wire made of multiple strands of wires; and the first secondary coil is a stranded wire madeof multiple strands of wires. According to the low-loss network information transmission device provided by the invention, signals can be enhanced through the first transformer and the second transformer; the transmission distance is increased; therefore, the chip side is isolated from outside; the anti-interference capability is improved; the signal control chip is protected; the stranded wiresmade of multiple strands of wires are used as the first primary coil and the first secondary coil; insertion loss can be reduced; and thus, loss can be reduced.