30results about How to "Guaranteed to study normally" patented technology

The invention discloses a depth image super-resolution reconstruction method based on deep learning, and the method comprises the steps: training the whole network when an upsampling factor r is equalto 2, and comprises the steps: selecting a certain number of depth images from different depth image public data sets; and data enhancement: designing a deep convolutional neural network structure: training the whole network by using the processed network input data and the data label, inputting the low-resolution depth image into the trained network model after the training is completed, and outputting the depth image of which the super-resolution is completed on an output layer. According to the method, the high-dimensional feature map is generated through simultaneous training of multiplechannels of the convolutional neural network, the accurate pixel value of the original low-resolution image is reserved, and the training and convergence speed of the whole network is increased.

The invention provides a small-target human body detection method based on balanced sampling and nonlinear feature fusion, and the method comprises the steps: carrying out the fusion and enhancement of the features of all scales in a training process, and enabling the scale where a small-target human body is located to obtain sufficient human body semantic information; using an instance balance sampling strategy for positive samples during subsequent loss calculation, it is guaranteed that human bodies of different scales can be trained in the same way, an intersection-parallel ratio balance sampling strategy is used for negative samples, it is guaranteed that enough samples are difficult to participate in training, and finally the classification capacity of the network is improved; and aknowledge distillation training method is adopted in a training strategy, so that the model size is compressed while the precision is ensured, and the reasoning speed is increased.

The invention discloses a method for diagnosing leading screw faults, belongs to the technical field of machinery fault diagnosis, and especially relates to ball screw fault diagnosis. The problem that a conventional intelligent leading screw fault diagnosis system is difficult for manual extraction of characteristics and limited in applying shallow-layer network non-linear expression capability. A diluted self-coding deep neural network structure is employed, and a network identification model employs a Softmax regression classifier to determine the hidden layer quantity of a network structure. The quantity of the input terminals and the quantity of the output terminals of a fault diagnosis model are determined, a training sample set, pre-training, fine tuned training and a fault diagnosis model test sample set are prepared, and the fault diagnosis performance of the fault diagnosis model is tested. Data sections in the test sample are integrally input continuously and sequentially, and the output quantity of the model is recorded to obtain an actual output list of the model. The ideal output list of the model is compared with the designed output list to obtain the fault performance test and evaluation result of the fault diagnosis model.

The invention discloses a scene-based learning content recommendation method and learning device. The method comprises the steps of controlling a camera device in the learning equipment to shoot a current scene image of a user; Identifying current scene information by analyzing the current scene image; Obtaining a scene label corresponding to the current scene information; Searching a target knowledge point type label which is matched with the scene label and has the highest user evaluation level from a plurality of preset knowledge point type labels, enabling one knowledge point type label tocorrespond to one user evaluation level, and enabling the user evaluation level to be in direct proportion to the evaluation accuracy of the user on the learning content corresponding to the knowledge point type label; Extracting target learning content matched with the target knowledge point type label from a preset database; Recommending the target learning content to the user; By implementingthe embodiment of the invention, appropriate learning contents can be recommended to the user in combination with the current environment of the user and the knowledge point type, so that the learningefficiency is improved, and a better learning effect is achieved.

The invention discloses a cloth defect detection method and system based on a deep neural network, and belongs to the technical field of pattern recognition. The method comprises the steps that a defect cloth image training set is used for training a deep neural network model, labels are defect types and real frame position information, the deep neural network model is composed of a backbone network and a detection network, and the backbone network is used for extracting three feature maps with different scales from defect cloth images; the detection network includesthree detection sub-networks and the detection result fusion module, wherein the three detection sub-networks are the same in structure. Each detection sub-network is used for detecting a defect type and prediction frame position information from the feature map, and consists of three dense connecting blocks, and the feature channel connection between the dense blocks is used for enhancing feature transfer, and the detection result fusion module is used for performing non-maximum suppression on the prediction result to obtain a final prediction frame and a defect type, and inputting to-be-detected cloth into the traineddeep neural network model to obtain a detection result, so that the type and the position of the defect in the cloth can be detected more quickly and accurately.

The invention provides a pushing method and a pushing system for a management course. The pushing method comprises the following steps that: obtaining time information and user liveness information from a user terminal, sending the time information and the user liveness information to a cloud server, and generating an active time period by the cloud server on the basis of the time information andthe liveness information; sending a standardized questionnaire to the user terminal, and sending an answer, which aims at the standardized questionnaire, of the user to the cloud server; judging the answer by the cloud server, analyzing a judgment result, and obtaining an analysis result, wherein the cloud server is based on the analysis result; when the length of the active time period is smallerthan a first threshold, pushing a first category of courses to the user terminal; when the length of the active time period is greater than the first threshold and small than a second threshold, pushing a second category of courses to the user terminal; and when the length of the active time period is greater than the second threshold and small than a third threshold, pushing a third category ofcourses to the user terminal. By use of the method disclosed by the invention, course pushing is targeted and humanized.

The invention discloses a saliency detection method for a color-thermal infrared image, and belongs to the field of computer vision image detection. The method comprises the following steps of: combining a thermal infrared image serving as an image channel with a color image, and performing super-pixel segmentation on the thermal infrared image by using a super-pixel segmentation method; extracting multiple types of feature vectors from each superpixel, establishing a graph model, and calculating an original graph adjacency matrix; stacking the multi-modal multi-feature graph adjacency matrixinto the tensor, and performing graph learning of a low-rank tensor; and finally, performing adaptive learning on the plurality of image adjacency matrixes by using an adaptive collaborative sorting algorithm to obtain an optimal image adjacency matrix, and performing superpixel saliency sorting by using background and foreground seed points respectively to obtain a final saliency map. According to the invention, image information of a color mode and a thermal infrared mode can be adaptively combined, and the robustness and accuracy of saliency detection are greatly improved.

The invention discloses a method, a device and a system for detecting a path MTU (Maximum Transmission Unit), which belong to the technical field of network. The method comprises the following steps: sending at least one detection message according to the address of a destination switch and a default path MTU, and acquiring at least one response message corresponding to the at least one detection message; determining a first path MTU according to the at least one response message; and determining a second path MTU between a source virtual machine connected to a source switch and a destination virtual machine connected to the destination switch according to the first path MTU and a preset package length. By sending at least one detection message to conduct learning, the first path MTU between the source switch and the destination switch is detected, and the second path MTU between the source virtual machine and the destination virtual machine is determined. The length of messages sent by the source virtual machine is controlled according to the second path MTU, and therefore, message sending failure is avoided, and the success rate of message sending is improved.

The invention discloses a control method and device for MAC address learning. The control method comprises the steps that a massage is received; the massage is analyzed, and the source MAC address of the message and the corresponding target MAC address of the message are obtained; matching between the target MAC address of the message and address entries in a MAC list is conducted, if matching succeeds, the corresponding source MAC address is allowed to be learned, and if matching fails, the corresponding source MAC address is forbidden to be learned. By the adoption of the control method and device for MAC address learning, according to a matching result between the target MAC address and the address entries in the MAC list, whether the corresponding source MAC address learning is conducted or not is judged, so that the impact of an unknown message on the MAC list is prevented, and it can be guaranteed that the source MAC address of known messages can be learned.

The invention discloses an image retrieval method and device. The method comprises the steps: carrying out first feature extraction processing of a to-be-retrieved image and a database image through employing a trained feature extraction network, carrying out the regularization global attention sampling processing, carrying out second feature extraction processing through employing the trained feature extraction network, carrying out the feature dimension reduction processing to obtain a to-be-retrieved dimensionality reduction feature map and a database dimensionality reduction feature map, and performing similarity calculation on the two to obtain a retrieval result. The feature extraction network training process is as follows: performing scaling processing on the training image, and performing feature extraction on the scaled training image by using a feature extraction network to obtain a feature matrix; performing regularization attention sampling processing on the feature matrix to obtain a global structure sampling graph and a local structure sampling graph; and performing knowledge distillation processing based on the global structure sampling graph and the local structure sampling graph to obtain a trained feature extraction network. According to the method, scene local features can be deeply learned, and the retrieval accuracy is improved.

The invention discloses an intelligent auxiliary driving guidance method, which comprises the steps of providing guidance content of an intelligent auxiliary driving function for a driver when the driver is judged to be a brand-new driver through face recognition; judging whether the vehicle is in a static state; and when the vehicle is in the static state, continuously providing the guidance content for the driver, so that the driver learns the guidance content. The invention further discloses a device and a computer readable storage medium. The problem that in the prior art, intelligent auxiliary driving operation instruction guiding content cannot be intelligently pushed to a driver is solved.

The invention provides a suspension cable-stayed classroom exterior wall sound insulation device, which comprises an adjustable up-stay cable whose one end is hinged above the classroom window facing the street of the classroom exterior wall, and the other end of the adjustable up-stay cable is hinged on the sound-proof The upper end of the sound insulation board, the lower end of the sound insulation board is hinged on the lower bracket of the sound insulation board below the classroom window facing the street on the outer wall of the classroom, the lower bracket of the sound insulation board and the inner side of the sound insulation board are provided with rainwater collection and diversion grooves, The invention has the beneficial effects of simple structure and convenient use, and can effectively reduce the influence of road noise on classrooms facing the street, thereby ensuring that teachers and students can study in a relatively quiet environment.

The invention discloses a plant pest and disease management system. The system includes pest and disease files, videos on demand, pest and disease reporting, self-diagnosis, online experts, pest and disease prediction, pest and disease prevention and control, statistical analysis and pest and disease control and prevention, the videos on demand include video uploading, video review and video release, the online experts include expert diagnosis, expert answering about questions and teleconsultation, the pest and disease prediction includes disaster prediction report generation, disaster prediction review and disaster prediction report release, and the pest and disease prevention and control includes prevention and control program formulation, prevention and control program evaluation and prevention and control program release. The system provides convenience for public to inquire about reports, videos and treatment programs of pests and diseases, provides convenience for a user to quickly find a solution to the damage to plants caused by the pests and diseases to ensure the yield of the plants, provides an online expert consultation and diagnosis module by means of network technologies, better serve the majority of users, and ensures that the public solves the damage to the plants caused by the pests and diseases in time to ensure the yield of the plants.

The invention provides a neural network model training method, an image feature extraction method and a related device, which are used for solving the problems of high hardware requirements and low training speed of self-supervised learning in related technologies. According to the comparison learning method provided by the embodiment of the invention, the concept of a difficult negative sample pair is provided, a neural network model can learn the features between the positive samples by constructing a positive sample pair and the difficult negative sample pair, and the negative samples with small differences can be accurately distinguished. Therefore, learning of difficult negative samples is ensured, that is, learning of negative samples with large differences is ensured, so that the neural network model can accurately extract image features.

The invention discloses a transformer damping and denoising device and an installation method thereof, and belongs to the technical field of denoising equipment. The transformer damping and denoisingdevice comprises a cover body, a temperature sensor, a single chip microcomputer and multiple damping systems; the cover body is sequentially provided with an air flowing cavity and an accommodating cavity from the outside to inside; the air flowing cavity comprises an intake cavity and an exhaust cavity; the intake cavity is communicated with the accommodating cavity via an air guide hole; a fanis arranged between the accommodating cavity and the exhaust cavity; the fan drives the air in the accommodating cavity to flow to the exhaust cavity; a transformer is arranged in the accommodating cavity; a sound insulation reinforcing plate is arranged at the outer part of the transformer; damping material is filled into the sound insulation reinforcing plate; the damping system is arranged above the cover body; the damping system comprises a cross arm and two liquid dampers; the temperature sensor is arranged in the accommodating cavity; and the single chip microcomputer is separately connected with the temperature sensor and the fan. The transformer damping and denoising device and the installation method thereof are simple in structure, easy to mount and dismount, low in manufacturingcost and good in denoising effect.

The invention relates to a multimedia demonstration device, and discloses a device which can prevent a lecturer from shielding multimedia projection during walking, ensure normal watching and learning of students below, clean a projection screen and avoid dust generated by long-time use of the projection screen. The device is characterized in that two fixing seats are fixedly arranged on the top of the protective cover through nuts, each fixing seat is composed of an L-shaped plate and a triangular reinforcing rib, one end of each L-shaped plate is fixedly arranged on the edge of the protective cover through two rows of nuts, the other end of each L-shaped plate extends out of the protective cover from the edge of the protective cover and then is bent upwards by 90 degrees, two through holes are formed in the other end of each L-shaped plate, the other end of each L-shaped plate is close to one side of the protective plate, the triangular reinforcing rib is fixedly arranged in the middle of each L-shaped plate, reinforcing bases are arranged at the two ends of the protective cover and are L-shaped structures, and one end of each L-shaped structure is fixedly arranged on the edge of the end of the protective cover through the nut.

The invention discloses a cross-category fault diagnosis method based on small sample learning. The method comprises the following steps: establishing a fault diagnosis model in combination with labeled historical working data of a part A; establishing a support set by combining the labeled historical working data of the part B; obtaining test data of the part B, pairing the test data with the labeled sample data in the support set to form a test sample, judging whether two pieces of data in the test sample belong to the same signal category or not through a fault diagnosis model, and obtaining the labeled sample data which are in the support set and have the same signal category as the test data; and collecting a signal category associated with the labeled sample data as a signal category of the test data. According to the method, the historical working data of the part A is adopted to train the fault diagnosis model, it is guaranteed that sufficient training data are learned in fault diagnosis model training, the part B provides the support set containing a small amount of labeled sample data, and the feasibility of signal category diagnosis on the part B through the fault diagnosis model is guaranteed.