11results about How to "Optimizing Model Parameters" patented technology

The invention discloses an audiovisual authenticity identification method, system and device and a storage medium, and relates to the technical field of digital security, the method comprises the following steps: constructing a first annotation data set and a second annotation data set according to multi-modal data, using the first annotation data set to pre-train an audiovisual authenticity identification model, and using the second annotation data set to pre-train the audiovisual authenticity identification model; the audio-visual authenticity identification model comprises an image modal sub-model and an audio modal sub-model; applying the pre-trained audio-visual authenticity identification model to an audio-visual authenticity identification task, performing fine adjustment on the audio-visual authenticity identification model by using a second annotation data set, and optimizing the audio-visual authenticity identification model after fine adjustment; and inputting to-be-identified multi-modal data into the optimized audio-visual authenticity identification model, and performing authenticity identification. By using image data, video data, audio data and other data in various modes, the model can comprehensively judge the authenticity of the digital content from multiple angles, so that the identification accuracy is improved.

This application provides a digital key positioning method, device, vehicle, and storage medium. The method includes: acquiring multi-anchor point ranging values ​​corresponding to the digital key; wherein the multi-anchor point ranging values ​​include ranging values ​​from the digital key to different ranging modules on the vehicle; calling a first model to identify the multi-anchor point ranging values ​​and determining a first identification result; if the first identification result indicates that the digital key is located inside the vehicle, calling a second model and a third model to identify the multi-anchor point ranging values; wherein the second model is trained based on a first training set, and the third model is trained based on a second training set, the second training set including training data that resulted in identification errors during the training of the second model; and determining the position of the digital key based on the identification results output by the second model and the third model. This method can more accurately determine the position of the digital key in complex environments, effectively improving positioning accuracy and robustness.

This invention provides a non-contact, highly anti-interference, and highly accurate wind turbine blade monitoring system and method based on lidar and multi-dimensional data fusion. The method includes the following steps: S1, data acquisition; S2, data preprocessing; S3, feature extraction; S4, data fusion; S5, equilibrium state diagnosis; and S6, dual verification. The system includes a lidar acquisition module (1), a data synchronization module (2), a signal processing module (3), a data fusion module (4), an intelligent diagnosis module (5), and a dual verification module (6). This invention is applied in the field of wind power equipment condition monitoring technology.

The invention discloses a range hood spare part multistage inventory intelligent distribution system and method, and belongs to the technical field of range hood spare part inventory management and control, and the system comprises range hood equipment, a plurality of self-service spare part shelves, an integrated spare part warehouse, a spare part management system access terminal, a spare part management system server, and a supplier spare part warehouse. With the spare part management system server as a connection center, the range hood equipment, the self-service spare part goods shelf, the spare part management system access terminal and the supplier spare part warehouse are connected together through a computer communication network. According to the invention, interconnection and intercommunication of internal and external digital systems of an enterprise are realized, data of multi-stage spare part inventory positions are integrated, the spare part demand of the next period of the current time is predicted based on the optimized random forest predictor, and high efficiency, intelligence and refinement of spare part management are realized.

This invention discloses a pre-trained online Gaussian process device and method for predicting hydrogen fuel cell voltage, applicable to marine propulsion systems. Addressing the problems of traditional models relying on real-ship data, weak generalization ability, and inability to adapt to battery aging, the device includes a sensor module, a memory, and a processor. The sensor module collects pressure, current, temperature, and voltage data; the memory stores the model and program; and the processor performs data processing and prediction calculations. The method pre-trains a Gaussian process regression model using experimental data, updates it online according to preset rules using real-time ship data, and outputs the prediction results after data preprocessing, model construction, updating, and prediction. This invention achieves high-precision prediction in zero-sample / few-sample stages, dynamically adapts to complex operating conditions and battery aging, operates efficiently without redundancy, and ensures reliable operation of marine hydrogen fuel cell propulsion systems.

The invention discloses a meter box topology identification method based on a high-speed broadband power line carrier technology. According to the method, a master station issues a ranging task to a concentrator, a carrier communication link is established with a plurality of terminal nodes through the concentrator, ranging data of the terminal nodes are collected, and signal strength data and signal-to-noise ratio data are received. According to background noise, normal noise, periodic impulse noise and sudden impulse noise existing in the low-voltage power line, noise analysis, frequency domain processing and filtering suppression are carried out on collected signals, so that effective data capable of reflecting communication link characteristics can be obtained. And after abnormal value detection and standardization processing are carried out on multiple rounds of measurement results, weights of different characteristics are determined based on correlation between communication characteristics and an actual topological relation, and a weighted linear regression model is constructed and used for predicting meter box affiliation of a terminal node or a relative position of the terminal node in a topology. And judging a topological structure of the terminal node according to a model output result, and verifying an identification effect through a model evaluation index.

The invention discloses a diesel generating set fault prediction method and model, and a storage medium, and the method comprises the steps: obtaining operation data, and processing the operation data to obtain a data sample set; mapping the data sample set to obtain an extended data sample set; performing calculation according to the extended data sample set to obtain an Euclidean distance; extracting according to the Euclidean distance and the correlation function to obtain local features; obtaining a first category and a first possibility according to the local feature output; obtaining a second fault category and a second fault occurrence possibility according to conversion between the first category and the first possibility; according to the second fault category and the second fault occurrence possibility output, obtaining a prediction value; analyzing according to the predicted value and the expected value to obtain a first parameter; performing calculation optimization according to the first parameter to obtain a second parameter; and optimizing the prediction model according to the second fault parameter, and outputting to obtain a fault prediction result. According to the method, the optimal parameters can be obtained through model iteration updating, and the accuracy of fault prediction is improved.

The invention provides a lightweight multi-scale detection method and system for unmanned aerial vehicle small target aerial photography, and the method comprises the steps: obtaining visible light or infrared image data of unmanned aerial vehicle aerial photography, carrying out the preprocessing of an image, and obtaining a preprocessed image; constructing a detection model based on a YOLOv11 model, extracting multi-scale features, and performing weighted fusion processing on the multi-scale features to obtain a cross-layer information aggregation feature map; constructing an unmanned aerial vehicle aerial photography small target data set, obtaining a training set sample, inputting a detection model, optimizing model parameters, and obtaining a lightweight multi-scale detection model; performing parallel detection on the cross-layer information aggregation feature map based on a lightweight multi-scale detection model, and outputting a detection result; the detection result is transmitted to the terminal in real time according to a set mode; the feature extraction capability is enhanced by constructing a detection model, effective complementation of shallow spatial information and deep semantic information is realized, a multi-scale feature map with rich semantic information is generated, and the small target detection precision is improved.

The invention relates to the technical field of medicine and deep learning, in particular to a tumor radiotherapy reaction prediction model and method based on a capsule graph structure neural network. The method comprises the following steps: acquiring a tumor data set, and extracting tumor data features from the tumor data set; the tumor data features are classified according to the risk levels, graph structure data sets are constructed respectively, and whether the tumor reacts or not is marked; constructing a deep learning model based on a capsule graph structure neural network; pre-training the deep learning model, and performing performance evaluation on the deep learning model to obtain a trained deep learning model; and inputting a to-be-predicted tumor data set into the trained deep learning model, and predicting whether the whole tumor has a radiotherapy reaction or not. According to the application, the image data before tumor radiotherapy and the radiotherapy dose can be utilized to predict whether the regional radiotherapy reactions of different severity degrees of the tumor are effective, so that a doctor is assisted to reasonably formulate the radiotherapy dose, precise radiotherapy is realized, and the radiotherapy side effect is reduced.

The application provides a training method, device and storage medium of a collaborative retrieval enhancement generation system. The method is applied to the field of artificial intelligence technology. The method comprises the following steps: performing a retrieval operation by a retriever based on input query content, and outputting a candidate document set associated with the query content; wherein the candidate document set comprises multiple candidate documents; performing sorting processing on the multiple candidate documents in the candidate document set by a rearranger to obtain a target document subset; generating an actual response result for the retrieval operation based on the query content and the target document subset by a generator; calculating a reward value corresponding to the actual response result, and optimizing the rearranger and the generator based on the reward value to obtain a trained collaborative retrieval enhancement generation system. The method can effectively solve the problems of asymmetric dependence, excessively high sorting sensitivity and limited generalization ability of the traditional RAG system.

This invention relates to the field of photoelectric detection and fault diagnosis technology, specifically disclosing an intelligent fault diagnosis system for a photoelectric pod image acquisition device. The system includes a data acquisition and preprocessing module, a cross-domain feature alignment and knowledge distillation module, a few-sample meta-learning diagnosis module, and a diagnosis decision and visualization module. By adapting general fault knowledge to the target domain through cross-domain knowledge transfer and combining it with a few-sample meta-learning mechanism, the system can achieve high-precision diagnosis with very few fault samples and possesses the adaptive capability to quickly adapt to new fault modes.