215results about How to "Training accurately" patented technology

The invention discloses a large-data-flow-based network traffic abnormality detection system and method. The large-data-flow-based network traffic abnormality detection method is characterized by comprising the following steps: acquiring network packet information in network equipment in real time in a distributed manner; transmitting the network packet information to a distributed flow processing platform in real time for network data analysis, feature matching and access counting; storing the analyzed and detected network data into a large data platform according to an abnormal status so as to facilitate clustering analysis and classified training of the network data and dynamically update a network data protocol characteristic library. Through the system and the method, real-time detection is achieved through a distributed flow-type processing mechanism; through distributed storage of the large data platform and through the data calculating and analyzing capability, the distributed storage of the network data is achieved and the network data protocol characteristic library can be trained more accurately.

The invention relates to a spinal orthotic device configured from one or more elements of a modular system, comprising the following elements:

• • a lower abdominal corset (40, 120),
  • an upper abdominal corset (17, 130) that can be attached cranially to the lower abdominal corset (40, 120),
  • a corset supporting element (41) that can be secured posteriorly in the lower abdominal corset (40, 120) and is arranged along the lumbar spine, supporting the spine while restricting sagittal mobility,
  • a thoracic spinal corset (10, 200) that can be attached cranially to the lower abdominal corset (40, 120),
  • at least one curved supporting clasp (47) that can be inserted posteriorly optionally into a bandage of a lower abdominal corset (40, 120) and an upper abdominal corset (17, 130) or into an bandage of a lower abdominal corset (40, 120) and a thoracic spinal corset (10, 200), said curved supporting clasp being attached to a corset supporting element (41) for correction of lordosis and for restriction of sagittal and frontal mobility in the area of the lumbar spine,
  • at least one supporting element (23, 160) which can optionally be secured cranially in the thoracic spinal corset (10, 200) and caudally to the corset supporting element (41, 150) and extends laterally along the spine to align and relieve the spine in the sagittal plane,
  • and an abdominal truss pad (190) that can be attached ventrally to a lower abdominal corset (40, 120) for correction of lordosis of the lumbar spine and increasing the intra-abdominal pressure.

The invention discloses a graph convolutional neural network model and a vehicle trajectory prediction method using the same. The model is composed of an encoder module, a spatial information extraction layer module and a decoder module. The method comprises the following steps: firstly, sampling a predicted vehicle and surrounding vehicles in a traffic scene at a frequency of 5Hz, and collectingposition coordinates and kinetic parameters of each vehicle sampling point, including horizontal and longitudinal coordinates, horizontal and longitudinal vehicle speeds and accelerations; calculatingcollision time TTC between the predicted vehicle and surrounding vehicles according to the coordinates and speeds of the predicted vehicle and the surrounding vehicles, and judging vehicle behaviors;inputting each historical track of the vehicle containing the information into the model, encoding time sequence interaction features in the track, extracting spatial features, summarizing the features into context vectors, and inputting the context vectors into an LSTM decoder to generate future track coordinates of the vehicle. According to the method, the problem that feature information generated by vehicle interaction cannot be obtained by using a traditional recurrent neural network is solved, and the prediction precision of the vehicle trajectory is greatly improved.

A system, devices and methods for training of a health care worker, real time monitoring of a health care worker and providing real time corrective guidance for procedures related to hospital infection control. Furthermore, the devices, systems and methods of the present invention are configured to provide monitoring the performance and the quality of a hand hygiene procedure and providing real time, in accordance with the World Health Organization's “5 Moments Hand Hygiene” events protocol, corrective guidance to the health care worker procedures and practice related to hospital infection control.

The invention relates to an OCT (Optical Coherence Tomography) image layer segmentation method based on a neural network and a constraint graph search algorithm, and is designed for accurately segmenting a retina layer and new vessels. The OCT image layer segmentation method based on the neural network and the constraint graph search algorithm comprises the following steps that: obtaining OCT image features to train a neural network classifier; obtaining a final SF1 through a multi-resolution map search algorithm SF1; extracting 24 features of an OCT image, and using the neural network classifier to find initial surfaces S1, S2,...S8; according to initial boundaries S2 to S8, using the constraint graph search algorithm to find accurate SF2 to SF8 in sequence; and segmenting the new vessels and hydrops between SF7 and SF8. The OCT image layer segmentation method based on the neural network and the constraint graph search algorithm has the advantages of being simple in operation and accurate in detection results. The existing problems of low identification rate, poor segmentation effect and the like of a lesion OCT image segmentation algorithm.

The invention discloses a personal-credit evaluation method of intelligent combination and a system. The method is suitable for use in the system. The system includes a sample data acquisition module,a sample data processing module, a data feature training module, a scoring model construction module and a scoring model test module. The method includes the steps of: S1, acquiring sample data of model training by the sample data acquisition module; S2, dividing samples into a training set I, a training set II and a test set; S3, carrying out feature grouping on the sample data of the training set I, the training set II and the test set; S4, carrying out training on each set of data features, and generating a corresponding trained sub-model by each set of data features; S5, carrying out prediction on the trained sub-models; and S6, testing a finally constructed credit scoring model on the test set by the scoring model test module to evaluate model test effect.

The invention provides a multi-style traditional Chinese painting peanut forming method based on a neural network, and the method comprises the steps: mapping from white painting to crafts or Chineseink painting through employing an attention architecture, carrying out training through employing a non-paired data set to acquire crafts and Chinese ink paintings with certain style characteristics.According to the invention, a new loss MSSSIM loss function is introduced into a GAN-based attention architecture; the distance between the reconstructed image and the source image is compared in a multi-scale manner from three aspects of brightness, contrast and structure, and the image style and content are guided to be accurately generated together with a cyclic consistency loss L1 function, sothat the problems of unclear original network picture, artifacts, fuzzy ink marks, noisy points and the like are solved. According to the method, the problem that high-quality pictures with clear pictures cannot be generated due to abstraction and complexity of traditional Chinese painting styles in research of a traditional method is solved.

The invention discloses a lesion monitoring method and device, a computer device and a storage medium. The lesion monitoring method comprises the following steps: inputting the sample data of an original CT image into a preset segmentation model for segmentation calculation, and outputting segmented liver image data; inputting the sample data of the original CT image and the segmented liver imagedata into a preset identification model for calculation, and outputting an identification result, wherein the segmentation model and the recognition model are respectively trained through a first convolution neural network and a second convolution neural network, and the first convolution neural network and the second convolution neural network are in cascade arrangement. According to the method,the neural network is used for learning liver features and lesion features in the original CT image, and the relation between the CT slices and the tags is established through the two cascaded full-convolution neural networks to distribute the task training model, so that the optimal network parameters can be found as soon as possible, and the disease treatment efficiency and the accuracy of disease analysis are improved.

The invention provides a vegetable identification method and device. The method includes the steps of establishing the first convolution neural network model, collecting the images of each vegetable in different scenes, building the corresponding input dataset, iteratively training the first convolution neural network model according to a learning rate adaptively updated gradient descent algorithmand the input data set, obtaining the second convolution neural network model, carrying out the vegetable recognition according to the second convolution neural network model, and improving the efficiency of the vegetable recognition is improved. The first convolution neural network model is trained by a stochastic gradient descent algorithm, so that the second convolution neural network model obtained by the training can accurately identify the vegetable species under the complex background.

The invention relates to the field of artificial intelligence. The embodiment of the invention provides a method and device for detecting a target object in an image. The method comprises the following steps: executing the following prediction operations by using a pre-trained neural network: detecting a target object in a two-dimensional image, and determining a two-dimensional bounding box of the target object; determining a relative position constraint relationship between the two-dimensional bounding box of the target object and a three-dimensional projection bounding box obtained by projecting the three-dimensional bounding box of the target object into the two-dimensional image; and determining a three-dimensional projection bounding box of the target object according to the two-dimensional bounding box of the target object and the relative position constraint relationship between the two-dimensional bounding box and the three-dimensional projection bounding box of the target object. According to the method, the accuracy of target object position detection is improved.

The invention discloses a deformable convolution kernel method based on a WGAN (Wasserstein-Generative Adversarial Network) model, and belongs to the field of the deep learning neural network. The method comprises the following steps that: S1: constructing an original generative adversarial network model; S2: constructing a Wasserstein distance as the judgment index of the adversarial network model; S3: initializing random noise, and inputting the random noise into a generator; S4: in the WGAN model, utilizing a deformable convolution kernel to carry out convolution on an image; and S5: inputting a loss function obtained by a deformable convolution operation into the generator for subsequent training. By use of the deformable convolution kernel method, which is constructed by the invention, based on the WGAN model, a convolution way generated after a discriminator and the generator receive a picture is changed, the discriminator and the generator can automatically change the size of the convolution kernel according to a training situation, so that the characteristics of a dataset image can be adaptively learnt, and the robustness of whole network training is improved.

The invention discloses a touch-screen multifunctional non-human primate animal cognitive-function testing cage and belongs to the technical field of animal experiment devices. The touch-screen multifunctional non-human primate animal cognitive-function testing cage comprises a testing cage and a transfer cage. An infrared-induction touch screen connected with a computer is fastened on the rear vertical face of the testing cage capable of moving back and forth, an observation window and a movable camera are arranged on the top face, a water supply nozzle is located at the rear locking-fixing position capable of moving vertically in front of the touch face of the touch screen, a feces receiving plate which can be pushed into or pulled out of the lateral vertical face of the testing cage is arranged under a grid-shaped base, and a testing cage door which can be pushed into or pulled out of the side face of the testing cage is arranged on a grid-shaped front vertical face. The transfer cage is provided with truckles and an observation window, and the front vertical face is a push-pull type transfer cage door which can be in butt joint with the testing cage door after being opened. The touch-screen multifunctional non-human primate animal cognitive-function testing cage can be used for flexibly, conveniently, intelligently, efficiently and accurately performing non-human primate animal training and testing based on a modern computer technology, does not need an additionally-equipped corresponding cage, has multiple practical functions and is low in purchase cost and usage cost.

The invention discloses a rumor detection method combining a self-attention mechanism and a generative adversarial network. The rumor detection method comprises the steps of collecting rumor text datato form a rumor data set; based on a self-attention mechanism, constructing a generative adversarial network generator comprising a self-attention layer; constructing a discriminator network, and respectively carrying out rumor detection and classification on the original rumor text and the text decoded by the generator; training the generative adversarial network, and adjusting model parametersof a generator and model parameters of a discriminator; and extracting a discriminator network of the generative adversarial network, and performing rumor detection on the to-be-detected text. Compared with an existing rumor detection method, the rumor detection method is higher in detection precision and better in robustness; a self-attention layer is adopted in the generator, key features are constructed through semantic learning of rumor samples, text examples rich in expression features are generated to simulate information loss and confusion in the rumor propagation process, and the semantic feature recognition capacity of the discriminator is enhanced through adversarial training.

The embodiment of the invention provides a machine translation model training method and device and electronic equipment. The machine translation model training method comprises the steps: employing afirst parallel corpus, and carrying out preliminary training on a translation model from a target end to a source end; utilizing the trained translation model from the target end to the source end totranslate each sentence in a definite monolingual corpus to obtain a synthetic corpus, and splicing the synthetic corpus with the definite monolingual corpus to obtain a second parallel corpus; evaluating the credibility of a translation result obtained by using the trained translation model from the target end to the source end by using a Monte-Carlo random inactivation algorithm; and based on the credibility, training a translation model from a source end to a target end by utilizing an integral corpus formed by the first parallel corpus and the second parallel corpus. According to the embodiment of the invention, the translation model can still be accurately trained in the presence of noise interference, and the accuracy of the translation model is ensured.

The invention relates to the technical field of user portraying, and provides a federation model training method and device, a customer portraying method and device, equipment and a medium, and the method comprises the steps: obtaining a participant list and an initial customer portraying federation model, and screening out qualified participants from the participant list according to a preset screening scheme; sending the initial customer portrait federation model to each qualified participant; receiving returned model parameters; performing abnormal feature extraction through a malicious parameter detection model by applying an MPI parallel method, and outputting an identification result of each model parameter according to the extracted abnormal features; filtering malicious parameters to obtain final normal parameters; and performing updating and federation learning to obtain a global customer portrait federation model. According to the method, the MPI parallel method is applied, abnormal feature extraction and malicious parameter filtering processing are carried out through the malicious parameter detection model, malicious parameters provided by malicious participants are automatically removed, and the efficiency and precision of federal learning modeling are improved.