24results about How to "Less attention" patented technology

The invention discloses a multi-granularity instruction generation method and device oriented to visual language navigation. The method comprises the following steps: extracting panoramic image features and action features of coarse granularity and regional features and grid features of fine granularity; coding and fusing the global features by adopting Bi-LSTM and an attention mechanism to obtain coarse-grained trajectory representation; space self-attention and time self-attention based on time constraint and cross attention based on time and local constraint are designed, local features are coded and fused, and fine-grained trajectory representation is obtained; the characterization of the coarse granularity and the characterization of the fine granularity are fused, and multi-granularity track characterization is generated; decoding is carried out through the LSTM-RNN, and a navigation instruction is synthesized; and the synthesized instruction and the sampled trajectory form enhanced data which is used for training a navigation agent. From the perspective of data enhancement, a high-quality natural language navigation instruction is synthesized, the generalization ability of an intelligent agent is improved, and an innovative and practical solution is provided.

The invention relates to the technical field of equipment fault prediction, in particular to an airport light station key equipment fault prediction method based on digital twinning, and the method comprises the steps: collecting the real-time operation data and real-time physical data of power distribution equipment, and collecting the real-time environment data corresponding to the power distribution equipment; judging an operation fault of the power distribution equipment based on the plurality of real-time operation data, and judging a physical fault of the power distribution equipment based on the plurality of real-time physical data; the physical fault judgment result is corrected based on the multiple pieces of real-time environment data, and then the actual fault degree of the power distribution equipment is judged according to the operation fault of the power distribution equipment and the corrected physical fault result; determining a predicted fault degree based on the real-time operation data, the real-time physical data and a power distribution fault digital twinborn model; and comparing the actual fault degree with the predicted fault degree to correct the power distribution fault digital twinborn model. According to the invention, the accuracy of the fault prediction result of the digital twin model is improved.

The invention provides a Transform-based multi-factor load prediction method, and belongs to the technical field of load prediction. The method comprises the following steps: acquiring source data to be subjected to multi-factor load prediction, preprocessing the source data, and extracting timestamp information, load data and external factor data from the preprocessed source data; and carrying out position coding on the timestamp information, training a Transform prediction model based on the load data, the external factor data and the timestamp information subjected to position coding, and carrying out multi-factor load prediction by utilizing the trained Transform prediction model. The load data refers to data representing power consumption conditions, the external factor data refers to environment data and electricity price data, the environment data comprises humidity and temperature, and the timestamp information refers to sampling time points of the load data and the external factor data. The method is used for improving the time sequence modeling capability of the model and improving the prediction precision.

The invention relates to a fuel cell system operation scheme prediction method and device, computer equipment, a computer readable storage medium and a computer program product. The method comprises the following steps: acquiring a preset operation parameter item of a to-be-predicted fuel cell system, and acquiring a preset value range and preset value precision of the preset operation parameter item; randomly generating a plurality of groups of operation parameter values of the preset operation parameter item according to a preset value range and preset value precision; respectively converting each operation parameter value into gene information of individuals in a genetic algorithm to obtain an initial population; and obtaining preset constraint condition information and a plurality of preset objective functions, and obtaining a predicted operation scheme of the to-be-predicted fuel cell system according to a preset constraint condition, each preset objective function and the initial population. By adopting the method, the running scheme prediction efficiency of the fuel cell system can be improved.

The invention provides a hyperspectral reconstruction method based on a single RGB image, and the method is characterized in that a hyperspectral reconstruction model comprises a conical multi-scale feature extraction module which is used for carrying out the feature extraction of the feature mapping of the single RGB image after the slicing and partitioning processing, and obtaining the aggregation features of a shallow region; a plurality of multi-scale adaptive residual attention modules connected in sequence and used for processing the shallow region aggregation features to obtain deep features; the combination processing module is used for carrying out convolution and activation function processing on the deep features to obtain a hyperspectral image, and the multi-scale adaptive residual attention module comprises a conical multi-scale feature extraction module, an optimal non-local module, a pixel self-attention module, a LayerNorm module and a multi-layer perceptron module which are connected in sequence. In a word, the method can improve the precision of the reconstructed hyperspectral image.

The invention discloses a fire smoke early recognition system and method based on deep learning, and relates to the technical field of fire smoke early recognition, and the method comprises the steps: obtaining image or video data in real time through a front-end data collection module, and transmitting the image or video data to a rear-end processing unit; the method comprises the following steps: marking various fire smoke samples through a deep learning training model module, training a deep learning model, learning basic morphological characteristics of smoke, and constructing a fire smoke detection model; the real-time smoke recognition module integrates a fire smoke detection model, analyzes image or video data acquired in real time, quickly recognizes early fire smoke, and gives an alarm through the alarm feedback module to guide related personnel to take actions. Therefore, by adopting the fire smoke early recognition method based on deep learning, the fire smoke can be monitored in real time, the recognition accuracy of the model is improved, the adaptability of the model to different environmental conditions is enhanced, and the recognition of the fire smoke under a complex background is met.

The invention discloses a belt running state detection method and device, electronic equipment and a storage medium. The method comprises the following steps: acquiring video frame sequence data containing a to-be-detected belt; inputting the video frame sequence data containing the to-be-detected belt into a pre-trained belt running state detection model to obtain a belt running state detection result; wherein the belt running state detection model comprises at least one convolution module, a convolution block attention module, at least one full connection layer and a classifier, and the convolution block attention module obtains a channel attention feature map and a space attention feature map based on the feature map of each video frame in the video frame sequence data. According to the technical scheme, the channel attention feature map and the space attention feature map are obtained through the convolution block attention module, so that the model pays more attention to the belt area, the attention to irrelevant background areas is reduced, and the accuracy of the belt running state detection result is improved.

The invention discloses a multi-label image classification method for multi-layer prompt information. The method comprises the following steps: acquiring a test image and a label set corresponding to a training set sample; using the topic model to obtain topic distribution of all samples in the training set; learning a group of prompt blocks representing theme label distribution information in the model, injecting the prompt blocks Prompt Token into the intermediate features, and inputting the prompt blocks Prompt Token into a selected prompt layer Prompt Block; and finally, judging which category the sample belongs to according to an output result of the Vision Transform. Compared with the prior art, the method has the advantages that the accuracy of a multi-label classification task is improved, an attention mechanism is helped to pay attention to a smaller area which is more critical for distinguishing which category an object belongs to, label missing and errors can be made up to a certain degree, the robustness of a model is improved, and meanwhile the labor cost is reduced.

The utility model provides a multifunctional fermenting chamber which comprises a fermenting chamber main body, a humidifying device arranged on the left side of the fermenting chamber main body, a heating device arranged on the right side of the fermenting chamber main body, a plurality of fermenting frames arranged in the fermenting chamber main body and an external control center, the humidifying device is provided with a high-pressure micro-mist humidifier, a humidifying pipeline is arranged above the high-pressure micro-mist humidifier, and the other end of the humidifying pipeline leads to the inside of the fermentation chamber main body; the heating device is provided with an electric heating pipe, a fan is arranged below the electric heating pipe, an air inlet pipeline is connected above the electric heating pipe, and the other end of the air inlet pipeline is connected with the interior of the fermentation chamber main body; a humidity sensor and a temperature sensor are arranged in the fermentation chamber main body; and the control center is correspondingly and electrically connected with the electrical devices. The pastry fermenting chamber solves the problem that the fermenting chamber can only ferment one kind of pastry food, and can be suitable for producing and fermenting different kinds of pastry.

The present invention discloses a network fault detection method, device, equipment and medium based on semi-supervised learning, the method comprising: collecting a network status data set from a network environment; preprocessing the network status data set, wherein the data set containing a label is recorded as a labeled data set, and the remaining data is recorded as an unlabeled data set; using the Tri-training algorithm to train the three pre-trained base classifiers to obtain a trained network fault model; assigning weights to the three base classifiers in the network fault model respectively; inputting the network status data to be detected into the trained network fault model to obtain a detection result of the network status. The present invention can use a small amount of labeled data to detect whether the network status is faulty, thereby improving the accuracy of detecting network faults, being applicable to a wider range of scenarios, and reducing the cost of annotating data.

The present invention belongs to the field of vehicle-mounted terminal systems, and specifically relates to a method, device, equipment and storage medium for multimedia data calling, the method comprising the following steps: receiving a media data calling request; determining a media service component that can execute the media data calling request according to the media data calling request; sending the media data calling request to the media service component; obtaining a source SDK corresponding to the media service component; parsing the media data calling request into a data calling instruction that can be recognized by the source through the source SDK, and establishing a link with the source; receiving media data sent by the source in response to the data calling instruction; and throwing out a media data display page. The multimedia data calling method of the present application can integrate multiple audio decoders, so that the product can focus more on providing users with an interface and interactive experience, greatly improving development efficiency and product iteration efficiency.

The invention relates to a face mask wearing automatic monitoring method suitable for a complex scene, and belongs to the technical field of artificial intelligence. Comprising functions of detecting whether a pedestrian wears a mask and automatically alarming after detecting that the pedestrian does not wear the mask. And the content of the detection part is realized by using a crawler and a camera to obtain video streams to collect picture data, clean the picture data, label a data set, train the data and improve a model. The model is improved as follows. Replacing a Focus module in the trunk with a convolution with a convolution kernel size of 6 * 6, and replacing a C3 module with a Ghostv2-C3 module provided by the invention; an original PAN feature fusion network is replaced by combining a BiFPN feature fusion structure, and standard convolution is replaced by GSConv; and finally, replacing the bounding box CIoU loss function with a Wise-IoU loss function. Therefore, the problems of false detection and missing detection of an existing model under a complex background are solved, and the memory occupied by hardware is reduced. The method reduces the loss of manpower and material resources, and promotes the high-quality development of enterprises.

The invention relates to the technical field of data transmission, in particular to a data transmission method and system based on edge computing, comprising: acquiring motion data of a vehicle and channel state information of a communication channel, the motion data comprising acceleration, and the channel state information comprising a signal-to-noise ratio; and based on a plurality of signal-to-noise ratios and accelerations within a set sliding window length. According to the method, the channel-motion coupling fluctuation index is constructed by fusing the vehicle acceleration and the channel signal-to-noise ratio, the time delay sensitivity weighted backlog coefficient reflecting the real business urgency degree is calculated according to the channel-motion coupling fluctuation index, and the self-adaptive parameters of the transmission control algorithm are dynamically adjusted. Therefore, the system can sense the potential channel deterioration risk caused by strenuous movement of the vehicle, the attention on energy saving is automatically reduced when the environment becomes worse or data burst occurs, and the real-time performance of data transmission is preferentially guaranteed, so that the transmission backlog and time delay are remarkably reduced in a high-dynamic edge computing scene.

The utility model provides a portable seal storage assembly which comprises a main body, and a storage mechanism used for storing a seal is installed on the outer wall of the main body. The utility model relates to the technical field of accounting supplies, compared with the prior art, the seal can be easily placed in a pocket through a storage mechanism, a user can carry the seal anytime and anywhere, the problem that the size is too large or the weight is too heavy does not need to be worried about, and the seal is not easy to perceive in daily carrying due to the hidden design; meanwhile, the seal is very convenient to take and use, only the storage assembly needs to be taken out from the pocket, in addition, the seal can be effectively protected from being easily obtained or embezzled by other people through hidden carrying, the safety of the seal is improved, and the seal is convenient to use. The method is particularly important for users who need to frequently carry seals and have certain requirements on safety.

The present invention relates to the technical field of equipment failure prediction, and in particular to a method for predicting key equipment failures of airport lighting stations based on digital twins, the method comprising: collecting real-time operation data and real-time physical data of power distribution equipment, and collecting real-time environmental data corresponding to the power distribution equipment; judging the operation failure of the power distribution equipment based on a number of real-time operation data, and judging the physical failure of the power distribution equipment based on a number of real-time physical data; correcting the physical failure judgment result based on a number of real-time environmental data, and then judging the actual failure degree of the power distribution equipment according to the operation failure of the power distribution equipment and the corrected physical failure result; determining the predicted failure degree based on real-time operation data, real-time physical data and a digital twin model of power distribution failure; respectively comparing the actual failure degree with the predicted failure degree to correct the digital twin model of power distribution failure. The present invention improves the accuracy of the fault prediction results of the digital twin model.

This invention discloses a multispectral and hyperspectral image fusion method based on a cross-scale Octave convolutional network, comprising the following steps: High spatial resolution hyperspectral images used for training are spatially downsampled and spectrally downsampled to simulate multispectral images (HR-MSI) and hyperspectral images (LR-HSI) captured by a sensor in the same region at the same time; a certain proportion of the resulting simulated multispectral and hyperspectral images are selected as the training set. For the training set, tensors corresponding to each training pair are generated, and each pair of training tensors is input into the convolutional network to obtain the optimal model parameters for multispectral and hyperspectral image fusion; the trained model is used to fuse low spatial resolution hyperspectral images and high spatial resolution multispectral images captured by the sensor to obtain a hyperspectral image with both high spatial and high spectral resolution.

The invention relates to the field of image generation material mapping, and discloses a single image material recovery method based on CBAM and an adversarial network, and the method comprises the following steps: S1, a network structure, S2, a CBAM attention module, and S3, a joint loss function. According to the method, a new joint loss function is provided, pixel-level loss and perception loss are combined by introducing the new joint loss function, so that high-level semantic features and pixel-level information are learned, loss based on human perception is introduced, the estimation quality is further improved, and the estimation accuracy is improved. An attention mechanism based on a spatial domain and a channel domain is introduced, the attention mechanism helps the network to improve attention on important features, reduce attention on less important features and enhance the ability of the network to capture related image details, and an end-to-end framework based on a CGAN is established; and the network can generate a more accurate SVBRDF material map which accords with expectation.

The application discloses an agricultural drought grade prediction method and device, electronic equipment and a storage medium, the method comprises the following steps: extracting time series data of a land surface from a CRU data center; extracting features of the time series data on a time scale, and performing feature fusion and global average pooling to obtain a time feature matrix; extracting features of the time series data on a variable scale, and performing global average pooling to obtain a variable feature matrix; extracting features of the time series data in space, and integrating information of adjacent grid points into the grid point information to obtain a spatial feature matrix; and splicing the time feature matrix, the variable feature matrix and the spatial feature matrix together, and outputting a final drought grade classification result through a Softmax layer. The application improves the prediction accuracy of agricultural drought grades and further reduces agricultural losses caused by drought.

The invention belongs to the technical field of campus safety, and provides a concealed alarm device and method, and the device comprises an alarm part which comprises a housing, a sealing cover connected with the housing in a clamping manner, and an alarm assembly, a trigger assembly and an adjustment assembly which are disposed in the housing; the mounting component comprises a mounting box, a mounting cover clamped with the mounting box, a mounting assembly connected with the mounting box and an operation assembly, and the shell is clamped with the mounting cover; the decoration part comprises a schoolbag strap and a decoration which is arranged on the alarm part and the mounting part in a sleeving manner; the alarm component is camouflaged through the ornament, so that the alarm component is not easily found by manufacturers under special conditions, the attention of the manufacturers under the special conditions can be reduced, and a user can conveniently use the alarm component to ask for help under the condition that the attention of the other side is not aroused; and therefore, a hidden alarm means can be provided for a user in an emergency.

The present invention belongs to the technical field of natural language processing, and particularly relates to a method and system for joint extraction of entity relations based on relation orientation, encoding sentences in a target text, obtaining a sentence vector representation in the target text; for the sentence vector representation, extracting the relationship type contained in the target text using a relation extraction module; fusing the extracted relationship type with the word vector representation in the target text sentence as prior knowledge, and using an entity recognition module to recognize entities in the target text corresponding to the extracted relationship type. The present invention can reduce attention to irrelevant entities, avoid extracting redundant entities, and then respectively identify the entity pairs corresponding to the identified multiple relationship types, solve the entity overlap problem, and finally extract all entity relationship triples contained in the sentence, improve the accuracy of entity relationship recognition, and facilitate application in actual scenarios.

The present invention discloses a method and device for detecting urine formed components based on deep learning, using a single-stage target detection algorithm with faster detection speed and a network model with fewer parameters, and proposes a local context attention module and a global background suppression module to improve the detection effect in view of the global sparseness and local density of urine sediment images, so that the model focuses on the area containing the target, enhances the ability to distinguish targets in dense areas, and reduces attention to the background. The technical solution of the present invention can effectively detect urine formed components in urine sediment images, and greatly improves the detection of local dense distribution.

The invention discloses a foldable forklift cargo blocking frame, relates to the technical field of forklift accessories, and aims to solve the problem that the performance requirements of an existing foldable cargo blocking frame and a fork material are high. The technical scheme is characterized in that the foldable forklift cargo blocking frame comprises a frame and two sets of extension frames, and the two side edges of the frame are each provided with a first rotating frame in a sliding mode; a first rotating frame is fixedly arranged on the frame, a second rotating frame is fixedly arranged on the extending frame, the first rotating frame and the second rotating frame on the same side are rotationally connected with each other, a plurality of driving parts are arranged on the frame, and the driving parts drive the first rotating frame. According to the foldable forklift cargo blocking frame, in the ordinary carrying process, the extending frame and the frame are only kept in the attached state, the pallet fork and the frame are kept in the normal embedded state, the performance of the pallet fork and the performance of the frame are the same as those of an ordinary pallet fork and an ordinary cargo blocking frame, the pallet forks of the same type can be matched with one another, and the maintenance difficulty is lowered.

This article discloses a finger vein recognition method based on a lightweight fusion global and local feature network. 1. Construct a data set and divide it into a training set and a test set; 2. Based on a lightweight residual unit, design a lightweight fusion global and local feature network FGL‑MobileNet; 3. Design an FGL‑Net network based on the fusion of global and local features, including a backbone network based on an improved residual network and a global feature and local feature extraction module; 4. Design a network loss function; 5. Train the entire FGL‑MobileNet network model until the entire training set is iterated for several times; 6. Input the test set image into the trained FGL‑MobileNet network model to extract the finger vein features and perform recognition and comparison. The present invention builds FGL‑MobileNet by stacking the unit, so that the network can quickly and effectively expand the receptive field of the finger vein features to obtain richer finger vein detail features, making the extracted finger vein features more discriminative, and the number of network parameters is also greatly reduced.