36results about How to "Reduce the number of labels" patented technology

The invention provides a cloud storage data auditing method for preventing a secret key from being revealed. The method comprises the first step of system parameter generating, the second step of secret key updating and the third step of file uploading and auditing. According to the method, a physically safe safety device is introduced to help a user to periodically update the secret key, and therefore data auditing in other time periods is still safe even when an attacker attacks the user at one time period and obtains the user secret key at the time period.

The invention provides a scheduling method to prevent conflicts in intelligent RFID read system. The method utilize the mechanism that reader controls multiple antennas, which is: all the antennas are controlled by the reader in order; the reader polls the antennas connected to the reader, and control the interactive between the single antenna and label; each time only one antenna can be activated to respond instructions from the reader; every two antenna effective covered regions are connected to reduce the overlap. The invention uses multiple antennas, at any time only one antenna takes effect, and then labels are divided into groups to recognize. The number of labels in the coverage covered by one antenna is reduced, thus the probability of conflicts is effectively reduced.

The invention provides a deep station caption detection method of weak supervision. The deep station caption detection method comprises the steps of preprocessing mass online video data files, and obtaining a large data set only marking a station caption type and a small data set only marking station caption position; inputting the small data set into a station caption positioning network to be trained, and obtaining a station caption positioning network capable of predicting a station caption area; inputting the large data set into the trained station caption positioning network to obtain a plurality of prediction station caption areas of each picture in the large data set, inputting the prediction station caption areas of each picture into a station caption classification network to be trained, and obtaining a station caption classification network capable of classifying station captions; conducting the same partial preprocessing on videos to be detected, inputting the preprocessed pictures into the trained station caption positioning network, and obtaining the prediction station caption areas of the pictures; inputting the prediction station caption areas of the pictures into the trained station caption classification network, and obtaining station caption positions and types of the pictures.

The embodiment of the invention provides a model training method and device based on active learning, and a server, and the method and device can automatically call a training sample of each application scene indicated by a started training task, and do not need a developer to manually export data for model training. After an initial deep learning network model matched with each application scenein advance is scheduled to perform sample labeling on training samples, an active learning service is scheduled to perform active screening on the labeled training samples, developers do not need to participate in labeling, the sample labeling number is effectively reduced, the active screening sample is calibrated according to a calibration instruction of a user, a training service associated with each application scene is flexibly scheduled to perform model training based on the calibrated active screening sample, and the model is published to a software application program corresponding toeach corresponding application scene; therefore, training services of different application scenarios can be flexibly docked, and automatic labeling, training and service updating of the training process are achieved.

The invention discloses a method and system for preventing collision of RFID tags. The method comprises the following steps that a reader uses the total number of digits of an assigned numerical value in the ID of a tag as a group number and groups all the tags according to the group numbers; and the reader successively recognizes the tags according to the group numbers. The method and system reduce the number of tags recognized by the reader at a time, reduces the collision probability of the tags, and improves the recognition efficiency of the reader. The method and system for preventing collision of RFID tags have high recognition efficiency, small system traffic and high system throughput, and can be widely applied in the technical field of RFID.

The invention discloses an optimal label selection method for an RFID equipment-free human body tracking system, which comprises the following steps: S1, region division: dividing a monitoring region into N positions; s2, feature extraction and calculation: extracting a mean value and a variance of T RSSIs sampled at each position within a period of time, and establishing a mapping relationship between the distribution of the RSSIs in the period of time and the positions of the RSSIs; s3, constructing a deep learning model, and analyzing a corresponding position sequence according to the RSSI sequence of the T, namely a real movement track of a human body; and S4, label layout mode selection: preferentially selecting labels according to the classification accuracy of the deep learning model for the positions. Through the deep learning model, while the human body tracking precision is maintained, the number of labels is reduced, the flexibility of the model is improved, precision reduction is avoided, the performance of processing a long-path sequence is improved, and the gap between training and inference of a position sequence prediction task is filled up.

The embodiment of the invention provides a message transmission method and device, and the method comprises the steps of obtaining a first message, and determining a forwarding entry corresponding tothe first message, wherein the forwarding table entries comprise a main path forwarding table entry and a standby path forwarding table entry, and an outbound interface included in the standby path forwarding table entry is a tunnel identifier of the TE tunnel; if the main path corresponding to the main path forwarding table entry fails and the second node is not the destination node correspondingto the destination address included in the first message, obtaining a segment routing label, pointing to the destination node, on the second node and a tunnel label corresponding to the TE tunnel from the standby path forwarding table entry; adding the segment routing label and the tunnel label in the first message to obtain a second message; and sending a second message to a second node throughthe TE tunnel corresponding to the tunnel identifier, so that the second node sends the second message to the destination node by using the segment routing label. Through the technical scheme providedby the embodiment of the invention, the message can be correctly forwarded through the standby path.

The invention relates to the field of artificial intelligence, and discloses an entity recognition model training method, which comprises the steps of obtaining an incompletely labeled specified training sample; inputting the specified training sample into a probability prediction model to obtain label probabilities corresponding to all unlabeled characters in the specified training sample; according to the label probabilities corresponding to all the unlabeled characters in the specified training sample, obtaining a label sequence with the highest probability through calculation by means of a Viterbi algorithm; according to the label sequence with the highest probability, determining covering labels respectively corresponding to all unlabeled characters in the specified training sample; obtaining a label sequence set corresponding to the specified training sample according to the covering label; obtaining label sequence sets corresponding to all the training samples in the incomplete annotation data set according to the obtaining mode of the label sequence sets corresponding to the specified training samples; and under the constraint of a preset loss function, training an entity recognition model through the label sequence sets corresponding to all the training samples. And a real tag sequence can be identified more easily.

The invention relates to a preference prediction method and device and computer equipment. The method comprises the steps of obtaining a network graph comprising combination nodes and edges, wherein the combination node is used for representing the combination of a user and an object, a part of combination nodes in the combination nodes carry the preference degree of the user in the represented combination to the object in the combination; dividing the network graph according to the combination similarity between the combination nodes of the building edges in the network graph to obtain at least one sub-network, wherein the same sub-network comprises combined nodes clustered into one class; and selecting a target preference degree corresponding to the combined node in the sub-network fromthe preference degrees carried by the combined node in the same sub-network. According to the scheme, the cost can be saved.

The invention discloses a refrigerator. The refrigerator comprises a refrigerator body internally provided with at least one storage chamber; and a refrigerator door arranged at the opening of the storage chamber, wherein an antenna and a display screen are arranged on the refrigerator door. A controller is configured to respond to label identification operation of the antenna, generate a corresponding food material control after obtaining food material information of all food materials corresponding to a newly-added label according to a preset label coding rule, and then update a food material management interface in the display screen according to the food material control. The invention further discloses a food material management method. By adopting the refrigerator and the food material management method, the food material information of various food materials is recorded in one RFID tag according to the preset tag coding rule, so that the excessive tag number in the refrigerator caused by one-to-one correspondence of the food materials and the tags is reduced, the probability that the tags are not identified due to shielding is reduced, and the accuracy of food material management is improved.

The invention relates to an artificial intelligence technology, and provides a word vector generation method and device, computer equipment and a storage medium, and the method comprises the steps: training a word in a corpus through a preset model, and obtaining an initialization vector of the word in the corpus; the candidate words of the to-be-constructed word vector words generated based on the initialization vectors of the words are sorted through the preset sorting model to obtain the positive correlation set and the negative correlation set of the to-be-constructed word vector words, and the candidate words are sorted, so that the number of annotations of the words is reduced, the training efficiency of the model is improved, and the training efficiency of the model is improved. The positive correlation set and negative correlation set samples are obtained through the preset sorting model, the quality of the samples is improved, the high-quality samples are sent into the contrast learning model for training, and the generation quality of the word vector of the word vector word to be constructed is improved.