20results about How to "Improve model quality" patented technology

The invention relates to a gradient aggregation federated learning method based on combination of zero knowledge proof and a block chain technology, and belongs to the field of distributed machine learning. Through combination of zero knowledge proof (zk-SNARK) and an extensible block chain technology, limitations of a traditional method in aspects of data privacy protection and model verification are broken through. According to the method, the correctness of a global model is verified by using zero-knowledge proof, the transparency and data privacy of an aggregation process are ensured, the Chinese remainder theorem is introduced to compress the proof, and the storage burden of block chain nodes is reduced. And the hash value updated by the encryption model is recorded through the block chain, verification is carried out based on a block chain consensus node, and the data integrity and the system transparency are further improved. Besides, a dynamic client selection mechanism is introduced in the method, participants are optimized based on training performance scores, and the training efficiency and the model quality are improved. As a whole, the invention provides a safe, efficient and transparent distributed learning solution.

The invention provides a limit test method, system and device for an artificial intelligence model and a medium, and the method specifically comprises the steps: constructing a plurality of scene templates for different types of games, and generating a test case set by combining and transforming the scene templates according to a boundary condition input by a user; based on the test case set, according to a preset optimization objective function, adopting a heuristic search algorithm to determine an optimal test parameter combination meeting the artificial intelligence model; performing automatic testing on the artificial intelligence model according to the optimal test parameter combination, and generating a model test result; and analyzing the model test result, determining the performance of the artificial intelligence model under the boundary condition, generating a model test report, and carrying out optimization and re-test on the artificial intelligence model through the model test report. According to the method, full-process automation and intellectualization of game AI model testing are achieved, and the testing efficiency and the model quality are remarkably improved.

According to the BIM-based parameterized spatial steel truss structure modeling method, electronic drawing spatial steel truss data are extracted through simple operation, an Excel table is processed and generated, spatial steel truss node three-dimensional coordinate data are extracted, precision errors caused by human factors are eliminated, meanwhile, a Revit template file is created, and a BIM-based parameterized spatial steel truss structure modeling method is provided. According to the method, a self-adaptive rod piece and bolt-sphere joint parameterization family is established, the modeling efficiency of the space steel truss in Revit is effectively improved, at least 90% of modeling time is saved, a running program written by Dynamo is applied, table data and parameterization family data are automatically called, accurate space steel truss and bolt-sphere joints are automatically generated, and the modeling efficiency and the model quality are greatly improved.

The invention discloses a fruit maturity comprehensive evaluation and shelf life prediction method based on a near infrared spectrum technology, relates to the technical field of agricultural product detection, and solves the problems that an existing nondestructive detection method is single, low in accuracy, incapable of meeting requirements and inaccurate in shelf life prediction. A quality and near infrared spectrum change model of fruits in the whole biological process in the actual production process is established, the optimal harvesting time is accurately determined through a fruit maturity comprehensive evaluation model, the fruit quality is guaranteed, and loss is reduced; in addition, by establishing a fruit shelf life prediction model, the shelf life of the fruits is judged, and proper circulation channels and selling time are selected, so that it is guaranteed that the consumers receive the fruits with the best quality. The method has the effects of improving the accuracy of timely harvesting of the fruits, realizing judgment of the optimal harvesting time of the fruits and field grading and sorting of the fruits, and accurately predicting the shelf life after harvesting.

The utility model relates to the field of 3D printing, in particular to a movement device and a 3D printer, and the movement device comprises a first transmission belt extending in the first direction; a second transmission belt extending in the first direction; the third transmission belt extends in the second direction perpendicular to the first direction, the first end of the third transmission belt is in transmission connection with the first transmission belt, and the second end of the third transmission belt is in transmission connection with the second transmission belt; the output assembly is fixed on the third transmission belt; the third transmission belt is configured to drive the output assembly to move in the second direction under the condition that the first transmission belt and the second transmission belt rotate in the same direction; under the condition that the first transmission belt and the second transmission belt rotate reversely, the third transmission belt drives the output assembly to move in the first direction. The problem that the transmission belt is loosened after being used for a long time can be solved, the movement precision of the output assembly is improved, and therefore the printing precision and the model quality are improved.

The invention discloses a battery life prediction method, system and device based on transfer learning and a medium, and relates to the technical field of battery monitoring. The method comprises the following steps: acquiring first battery operation data, vehicle driving data and environment data of a first energy storage battery; determining target domain data according to the first battery operation data, the vehicle driving data and the environment data, and constructing a corresponding initial battery life prediction model; determining a plurality of second energy storage batteries different from the first energy storage battery in battery type, and acquiring second battery operation data and battery life of each second energy storage battery; the second battery operation data and the battery life serve as source domain data, model parameters of the initial battery life prediction model are adjusted through a transfer learning algorithm and the source domain data, and a target battery life prediction model is obtained; and predicting the target battery life of the first energy storage battery based on the target battery life prediction model. By implementing the technical scheme provided by the invention, the effect of improving the battery life prediction efficiency is achieved.

The utility model relates to a transmission device and a 3D printer, and the transmission device comprises a first transmission belt which extends in the first direction and is provided with a first end and a second end which are located in the first direction, and the relative positions of the first end and the second end of the first transmission belt are fixed; the first tensioning assembly comprises a fixing seat and a pressing block connected to the fixing seat, the fixing seat is fixed to the first transmission belt, and the pressing block is configured to abut against the first transmission belt in the direction forming an included angle with the first direction so as to apply tensioning force to the first transmission belt. The first transmission belt can be effectively tensioned, and the printing precision and the model quality are improved.

The invention discloses a large model three-dimensional modeling system and method based on artificial intelligence, particularly relates to the technical field of three-dimensional modeling, and comprises a large model three-dimensional modeling data acquisition module, a large model three-dimensional modeling data analysis module, a large model three-dimensional modeling data evaluation module and a large model three-dimensional modeling comprehensive analysis module. And a large model three-dimensional modeling man-machine interaction module. The large-model three-dimensional modeling data acquisition module is used for acquiring performance data for analyzing large-model three-dimensional modeling based on artificial intelligence to obtain various parameters for evaluating performance indexes of large-model three-dimensional modeling, and outputting the various parameters to the large-model three-dimensional modeling data analysis module; according to the method, the modeling efficiency parameters are analyzed, an index is obtained to measure the efficiency of the three-dimensional modeling process, various indexes in the modeling process are monitored in real time, an administrator can adjust a modeling strategy and optimize the modeling process according to the indexes, the modeling speed is increased, and resources are optimized.

The invention belongs to the technical field of data processing and parameterized modeling, and particularly relates to a containment prestressed steel beam parameterized modeling method based on B-spline interpolation, which comprises the following steps of: determining position information and shape parameters of each steel beam on an expanded drawing according to a design drawing of a containment prestressed steel beam; nodes of complex prestressed steel beams near an equipment gate and a hole opening are calculated; calculating a B spline primary function based on the nodes of the complex prestressed steel beam; coordinate interpolation of a vertical curve section part in the complex prestressed steel beam is achieved through a B-spline primary function, space coordinates are calculated, and a three-dimensional space coordinate point set of the prestressed steel beam is generated; based on a secondary development function of the ABAQUS, a Python script is compiled to realize parametric modeling and automatic numbering of the prestressed steel beam, so that a three-dimensional model of the prestressed steel beam is established. According to the method, coordinate interpolation of the prestressed steel beam is achieved through a B-spline interpolation function, and secondary development of ABAQUS and parameterized modeling of the prestressed steel beam are achieved based on Python.

The invention provides a clothes image acquisition method and device, a storage medium and a clothes image acquisition device, and relates to the technical field of image processing and intelligent devices. The clothes image acquisition method comprises the following steps: acquiring a target light-supplementing lamp position according to a brightness difference between the brightness of clothes in an initial clothes picture sent by clothes image acquisition equipment and the final brightness, and acquiring the brightness of the target light-supplementing lamp according to the color of the clothes in the initial clothes picture and the final brightness, and sending the target light supplementing lamp position and the target light supplementing lamp brightness to the clothes image acquisition equipment, and obtaining a target clothes picture of which the clothes brightness is the final brightness according to the returned picture. Through the mode, the image information acquired by the clothes image acquisition equipment is uniform in illumination, the modeling quality of the three-dimensional model acquired according to the image is improved, and meanwhile, the user experience is improved.

The invention provides a time coding method based on time continuity and periodic characteristics in the technical field of knowledge maps, and the method comprises the steps: S1, a server obtains a current moment and an input query moment of a time knowledge map, and calculates a time difference based on the current moment and the query moment; s2, parameterizing the time difference so as to map the time difference into a multi-dimensional vector; s3, based on a trigonometric function with continuity and periodicity characteristics, performing explicit modeling on the multi-dimensional vector to obtain a continuous time characteristic and a periodical time characteristic; s4, performing weighted fusion on the continuous time features and the periodic time features to obtain explicit time embedding representation; and S5, obtaining entity embedding of the time knowledge graph, and splicing the entity embedding and the time embedding representation to obtain a dynamic entity embedding representation so as to complete time coding. The method has the advantages that the comprehensiveness of time coding of the time knowledge graph is greatly improved, and then the entity modeling quality is greatly improved.

The invention relates to the technical field of airborne system design, and discloses an airborne system model modeling rule making method and device, equipment and a medium, which can deeply fuse airborne system design multi-source heterogeneous knowledge, automatically generate executable modeling rules, improve airborne system design efficiency and model quality, and improve airborne system design efficiency and model quality. The problem that modeling rules are inconsistent due to the fact that traditional modeling rules depend on artificial experience is solved, and standardization, consistency and quality control efficiency of airborne system design are improved.

The invention relates to the technical field of three-dimensional model optimization, and discloses a three-dimensional model optimization method based on an indoor transformer substation monitoring video, which comprises the following steps of: establishing a transformer substation model library in which a model is not subjected to mapping processing; arranging a three-dimensional model; acquiring an equipment model in the three-dimensional model, and acquiring a live-action picture of equipment of the same model through a monitoring video; performing mapping processing on a model part of the three-dimensional model by adopting a live-action picture; and completing the construction. Compared with the prior art, the method has the advantages that the modeling speed is higher, local details of all inspection point positions can be covered basically, the inspection point positions can be consistent with real scenes basically, the modeling quality is high, and the modeling efficiency is high. Moreover, the model is constructed by adopting the original monitoring equipment, and a panoramic camera and a laser radar are not needed, so that the construction cost is low.

The application provides a kind of adaptive 3D grid reconstruction and error compensation method for die frame square iron reverse engineering, including using laser scanner to carry out 360 ° scanning on die frame square iron to obtain original point cloud data, and the original point cloud data is registered to the same coordinate system, the original point cloud data is processed by multi-scale bilateral filtering, while constructing point cloud quality evaluation field, identifying data missing area and edge aliasing area, generating point cloud confidence weight map, establishing octree space index and calculating structure tensor to obtain principal curvature value, based on principal curvature and confidence weight, the square iron surface is divided into large plane area, transition feature area and boundary feature area, and the geometric topology skeleton line is extracted.The application decomposes the deviation into systematic error and random error and carries out reverse correction, which not only eliminates the overall deformation error of the workpiece, but also retains the real machining texture, realizes high-quality reverse reconstruction.

This invention provides a fully automated laser scanning detection system and method for tunnels and chambers, belonging to the field of underground space exploration technology. The system includes: a drone, a multi-sensor fusion positioning module, a 3D laser scanner, an adaptive environmental perception and obstacle avoidance module, and a ground control station. The multi-sensor fusion positioning module provides high-precision pose information in environments without GPS signals; the 3D laser scanner collects laser point cloud data of the tunnels and chambers; and the adaptive environmental perception and obstacle avoidance module senses changes in dust and illumination in real time and plans a safe flight path. This invention effectively overcomes the challenges of complex environments such as GPS rejection, uneven illumination, and dust interference within tunnels through multi-source data fusion and adaptive algorithms. It achieves fully automated, high-precision, and high-efficiency 3D topographic detection in areas inaccessible to personnel, providing a reliable technical means for monitoring surrounding rock deformation and assessing risks in goaf areas.

The embodiment of the invention discloses an external ear model edge smoothing method and device based on pipeline cutting and a medium, and relates to the technical field of three-dimensional modeling. The method comprises the steps of obtaining an original external ear model; resampling an edge to be smoothed in the original external ear model according to the average vertex distance of the original external ear model; by taking a new edge formed by connecting the resampling vertexes as a center line, constructing a closed pipeline which extends along the new edge and has the cross section being an equal-radius circular ring; cutting the original external ear model by utilizing the closed pipeline, and reserving a part outside the closed pipeline; and filling the cut surface of the reserved part into a smooth curved surface to obtain a final external ear model. According to the embodiment, the edge smoothing effect of the external ear model can be improved.

The invention discloses a federal forest trusted participant screening and auditing method and system. The method comprises the steps that a model consumer puts forward a model service demand to a cloud platform; the cloud platform issues tasks; when a certain participant Pi participates in the task for the first time, registration is completed at the LGPA; the cloud platform performs identity authentication on the Pi according to the received task request message of the participant Pi; after the participant Pi passes the verification, the cloud platform determines whether the request of the participant is passed or not; the cloud platform and a plurality of participants authorized through the task requests cooperate together to complete construction of a random forest tree model; the model consumer submits rewards to the cloud platform, and the cloud platform gives a result and feeds back the result; the cloud platform performs reputation evaluation on the participant Pi through a subjective logic model to update the reputation level; the cloud platform allocates remuneration according to the latest reputation level of the participant Pi; compared with a traditional method, the reliability of participants in the federated learning system can be improved.

The invention relates to a TIN (Triangulated Irregular Network) adjacent image registration method based on sparse feature points under a high-likelihood image background. According to the method, feature points extracted from images are utilized, a few significant feature points are selected from the feature points to construct a TIN network, and then registration and superposition of adjacent images are completed through congruent analysis and judgment of the TIN network. The method is small in calculation amount, high in precision and high in registration speed, the registration precision and the registration efficiency of the image can be greatly improved, the aerial triangulation calculation and modeling speed of later three-dimensional modeling can be accelerated, and the three-dimensional live-action modeling efficiency and the model quality of a construction project construction site are improved.

The invention relates to a speaking face video generation method based on neural radiation field spatial position optimization. The method comprises the following steps: extracting voice information and a face image area from a reference video; sampling the face image area of the first video frame, and tracking the position of a sampling point on a subsequent video frame; determining N sampling points as two-dimensional key points; defining N three-dimensional space key points in the three-dimensional space, and mapping the three-dimensional space key points to an image plane where the sampling points are located by combining the camera external parameters, the camera internal parameters and the rotation translation parameters of the space points to generate N two-dimensional plane points; calculating Euclidean distances between the N two-dimensional key points and the N two-dimensional plane points, and adjusting related parameters and three-dimensional space key point positions; obtaining a three-dimensional face reconstruction model of the target face; and generating a speaking face video of the target face under the target view angle and the target voice information based on the target view angle, the target voice information and the three-dimensional face reconstruction model of the target face.

The application discloses a question recommendation method based on a schema in a question and answer community, and the steps include: 1, collecting data and constructing a question and answer community network; 2, using a schema mining algorithm to mine the schema in the user network, and constructing a schema network based on the schema; 3, using a network embedding learning algorithm to learn the node embedding of the network, and learning the preferences of the answerer for the questioner and the question; 4, fusing the preferences of the answerer in two aspects, using a scoring function to predict the matching degree of the answerer for a new question, arranging the answerer in descending order according to the matching score, and recommending the first N users to answer the question, so as to complete the question recommendation task. The application combines the schema mining algorithm and the embedding learning algorithm of the schema network, fully captures the historical interaction information of the users in the schema network by using the graph neural network, and fully fuses the personal feature information, network structure information and text semantic information of the users, so that more accurate recommendation effect is realized.