17results about How to "Guaranteed recall" patented technology

The invention discloses a method and equipment for evaluating throwing motion and a computer storage medium. The method comprises the following steps: acquiring a first video and a second video of throwing a target object by a target person; detecting feet of the target person in each frame of image of the first video to obtain foot positions of the target person in each frame of image; detecting a target object in each frame of image of the second video to obtain a throwing position of the target object in each frame of image, and determining a landing position of the target object according to the throwing position; determining whether the target person has an illegal behavior or not according to the foot position and a preset throwing circle; and when the target person does not have the illegal behavior, determining the throwing distance of the target object by using perspective transformation based on the landing position and the preset calibration area, and obtaining the throwing score of the target person. According to the scheme disclosed by the invention, the evaluation process is simplified, so that the evaluation efficiency can be improved.

The invention provides an industrial controller screen display anomaly detection method, which comprises the following steps of: judging an acquired screen image through an anomaly detection algorithm comprising three models, namely, a Wafer model, a patchcore model and a Fastflow model, and a bagging module, judging that the display is abnormal when the two models judge that the image is abnormal at the same time, and judging that the display is normal when the two models judge that the image is normal at the same time; and when one model judges to be normal and the other model judges to be abnormal, the normalized image enters a trained Fastflow unsupervised deep learning model for judgment, when Fastflow is judged to be abnormal, screen display is judged to be abnormal by an algorithm, and otherwise, display is judged to be normal. According to the industrial controller screen display anomaly detection method, the requirement for the deployment environment is relatively low, and the hardware cost can be saved. According to the technical scheme, the recall rate can be guaranteed, the accuracy can be improved, the overall detection effect is improved, and meanwhile, the dependence on negative samples is reduced by adopting an unsupervised deep learning model.

The application provides a sparse-to-dense visual positioning method and system based on feature Gaussian splash, and the method comprises the following steps: initializing a color-decoupled feature Gaussian field based on a training image set, optimizing the color-decoupled feature Gaussian field based on a query feature map set in combination with feature rendering and feature alignment loss cyclic optimization, and outputting a compact feature Gaussian scene model; screening a Gaussian landmark set in the compact feature Gaussian scene model by using a matching-oriented sampling strategy; training a scene-specific detector; extracting sparse local features of a query landmark heat map corresponding to a query image and performing sparse feature matching with the Gaussian landmark set to obtain an initial pose of a query perspective camera; based on 3D Gaussian splash, rendering a dense feature map and a depth map of the query perspective in the compact feature Gaussian scene model by using the initial pose of the query perspective camera, performing cluster-based proxy matching-based sparse-to-dense accelerated pose optimization, and obtaining accurate positioning of the query perspective camera.

The application relates to the technical field of data processing, in particular to an image clustering method, a computer device and a storage medium. The image clustering method provided by the application comprises the following steps: acquiring a plurality of images, and separating a first-class image with high image quality and a second-class image with low image quality from the images; acquiring a real-time library archive from a real-time library, performing real-time clustering on the first-class image based on the real-time library archive, and integrating the first-class image that has passed the real-time clustering into the real-time library archive, and adding the first-class image that has failed in the real-time clustering as a class cluster; acquiring the second-class image, the class cluster and the first-class image in the real-time library archive, and performing offline clustering on the second-class image and the first-class image; and adding the second-class image that has passed the offline clustering into the class cluster corresponding to the first-class image, or updating the second-class image that has passed the offline clustering into the real-time library archive corresponding to the first-class image. Through the above method, high-quality images can be quickly clustered, and low-quality images can be effectively recycled, so that the efficiency, real-time performance and integrity of the clustering are ensured.

The invention provides a model training method and device, an identification method and device, electronic equipment and a storage medium, and relates to the technical field of machine learning. The model training method comprises the steps of training a first model based on an initial training sample, and generating a basic classification model; obtaining a classification result of the initial training sample by the basic classification model; determining a predicted positive sample in the classification result as an enhanced training sample, determining classification features of the basic classification model as training features, separating a real case and a false positive case in the predicted positive sample, determining the separated real case and false positive case as training targets, training the second model, generating an enhanced classification model, and generating a screening rule by the enhanced classification model, the screening rule is used for screening out a plurality of target subsets from the enhanced training sample, and the accuracy rate of the plurality of target subsets is greater than the target accuracy rate. By means of the technical scheme, the model recall rate can be guaranteed, and meanwhile the model recognition accuracy can be improved.

The application discloses a kind of generation method and device of pharmaceutical information extraction template, it is related to natural language processing technical field.The specific embodiment of the method includes: obtaining at least one initial template for information extraction;Wherein, each initial template contains multiple components belonging to preset type, the component is at least one in matching symbol and fixed word, the preset type includes at least one in entity type and relationship type;Determine multiple template units according to the component in any initial template;With multiple preset sequences, form multiple new templates corresponding to the initial template for the multiple template units determined.This embodiment can automatically generate multiple new templates according to existing initial template to execute information extraction, so as to improve information extraction recall rate in the case of ensuring information extraction accuracy and not increasing artificial workload.

The present application provides a risk control strategy generation method, device, storage medium and program product, the method comprising: dividing N cheating users into M cheating user groups, N and M are both integers greater than 1; determining P characteristic indicators of the first cheating user group in the M cheating user groups arranged from high to low in importance, P is an integer greater than 1; generating a target risk control strategy corresponding to the first cheating user group according to the P characteristic indicators. Thereby improving the efficiency and accuracy of generating the risk control strategy, and the risk control strategy is targeted at the cheating user group.

The embodiment of the application discloses a kind of multi-modal drawing model compliance detection method, equipment and medium, belong to drawing model examination technical field, solve the review accuracy of industrial machinery field drawing model is lower, and the reliability of review result is lower.The problem includes, multi-modal drawing model is identified to multi-modal drawing model by multi-modal big model, and design element and geometric attribute are output;Wherein, multi-modal drawing model at least includes two-dimensional engineering drawing and three-dimensional geometric model;According to design element, generate drawing specification retrieval key, based on drawing specification retrieval key in preposition hierarchical vector knowledge base, execute semantic vector retrieval and structured field matching, determine target clause and target tolerance parameter;Geometric attribute, target clause and target tolerance parameter are as context, input large language model, and numerical comparison and cooperation determination are carried out to context by large language model, and compliance reasoning result is obtained;Based on compliance reasoning result, generate compliance detection report.

The invention relates to a training method of a false propaganda detection model, a false propaganda detection method and related equipment. The training method comprises the following steps: acquiring original propaganda data; inputting each piece of original propaganda data into the large language model to obtain structured propaganda data; inputting the propaganda text content into a false propaganda detection model to obtain a plurality of false propaganda prediction results; calculating a value of a loss function based on the plurality of false publicity prediction results and corresponding tags indicating whether false publicity is involved; and adjusting parameters of the false propaganda detection model according to the value of the loss function so as to train the false propaganda detection model. Compared with a mode of obtaining the training data by manually marking the original propaganda data, the large language model provided by the invention can significantly improve the efficiency of obtaining the structured propaganda data. The false propaganda detection model is trained by using the structured propaganda data, that is, more standard and more normative training data, so that the training effect, accuracy and recall ratio of the model can be ensured.

The application discloses a kind of lightweight parking stall detection methods, belong to computer vision technical field.The algorithm of a kind of lightweight MobileNet+FPN+Aware Headers is designed, and depth separable convolution is used in convolution layer to reduce parameter quantity most;MobileNet is used to compress feature;Different resolution features are fused using FPN, and finally, Aware Headers is used to complete the prediction of parking stall key points.Aware Headers includes 3 prediction heads in total, respectively for predicting the position of key points, the direction of key points and the center point of parking stall entrance.Labels generated by Gaussian kernel with parking stall direction are used to make the direction and position of parking stall have better correlation.In post-processing, according to the center point of parking stall, the corresponding two parking stall key points are matched to determine a complete parking stall.The detection method disclosed by the application has lower calculation and parameter quantity while the detection result is reliable.

This application discloses an information acquisition method, apparatus, device, and storage medium, belonging to the field of artificial intelligence technology. The method includes: acquiring the intent classification result of a natural language question, whereby the intent classification result indicates the intent type corresponding to the natural language question; determining a target retrieval method from multiple retrieval methods based on the intent classification result; wherein the multiple retrieval methods are used to acquire knowledge at different levels of expertise; and retrieving the answer to the natural language question from the knowledge base corresponding to the target retrieval method using the target retrieval method. This application ensures the professionalism of the answers by retrieving answers at different levels of expertise for natural language questions with different intent types, thereby improving the accuracy and flexibility of information acquisition. Furthermore, since this application is not limited to obtaining answers from a single knowledge base, it ensures the recall of information. This application can be applied to intelligent question answering in the medical field (such as medication and medical consultation).

The application provides a kind of aggregate three-dimensional reconstruction method and system based on multi-view vision, it is related to computer vision and three-dimensional reconstruction technical field, the method comprises: obtaining the multi-view image of aggregate;Feature matching processing is carried out to multi-view image, including metric reconstruction to original feature descriptor, dynamically adjusting matching threshold based on local texture statistics, performing spatial balancing processing of griding adaptive quota allocation, and removing outliers based on residual evaluation of geometric model, to obtain geometric consistency in-point matching set;Based on in-point matching set, multi-view geometry reconstruction is carried out to obtain dense point cloud;Dense point cloud is mapped to regular grid to carry out high-order interpolation and low-pass filtering smoothing processing, to obtain the continuous three-dimensional surface model of aggregate.The application improves the robustness of feature matching and the integrity of reconstruction model in aggregate scene by feature metric reconstruction and spatial balancing distribution strategy, and provides high-precision three-dimensional geometric basis for aggregate granularity analysis.

The invention discloses a target detection method and device, a storage medium and an automatic driving vehicle, and belongs to the field of target detection. The method comprises the following steps: acquiring first sensing information, acquiring a first confidence coefficient based on the first sensing information, comparing the first confidence coefficient with a first threshold value, taking the first confidence coefficient greater than the first threshold value as a candidate type, and calculating a second confidence coefficient according to second sensing information in tracking information, and determining whether the candidate type is a target type according to the second confidence coefficient and a second threshold value. The recall rate and the accuracy rate of target detection can be improved.

This invention relates to the field of intelligent manufacturing and industrial quality management technology, and in particular to a management system for die-cast housings of new energy vehicles. The system includes a data acquisition module for periodically collecting production data, quality data, and production equipment data of the die-cast housings; a data analysis module for determining the defect type of the die-cast housings within a single cycle based on the average defect density and defect clustering index of the die-cast housings within that cycle; a traceability module for determining different traceability methods based on the defect type of the die-cast housings within a single cycle; and an adjustment module for adjusting the judgment parameters for the defect type of the die-cast housings within a single cycle based on the defect traceability accuracy. This invention, through automatic and accurate defect classification and differentiated traceability, quickly locates the root cause of problems, greatly shortens the problem response time, and reduces the generation of defective products.

The invention discloses a digital information processing method and system based on mixed retrieval, and aims to solve the problems of low retrieval efficiency and poor precision of a knowledge base in a professional field. The method comprises the following steps: firstly, constructing two indexes for a document set in parallel, wherein one index is a vector index for recording semantic vectors of document blocks, and the other index is an inverted index based on a preset regular expression; after user query is received, the system starts two retrieval paths in parallel: one is that global semantic retrieval is executed through vector indexes to obtain a preliminary candidate set; and 2, analyzing a specific mode in query, and if the specific mode is matched with the regular expression, performing rapid and accurate matching by using an inverted index to obtain a candidate subset with a smaller range. And the system preferentially performs fine sorting in the subset in combination with semantic similarity and outputs a result. According to the method, by introducing regular rapid preliminary screening, the follow-up processing range is effectively narrowed, and the retrieval efficiency and the retrieval precision of information such as specific codes and identifiers are remarkably improved.

The application relates to the fields of database management and natural language processing, and specifically discloses a txt-to-sql generation method based on database sub-pattern division, which comprises the following steps: S1, constructing a database graph; S2, performing sub-pattern division on a weighted undirected graph G based on graph cutting to obtain a plurality of subgraphs, each of which corresponds to a database sub-pattern; S3, generating a sub-pattern abstract; S4, adopting a two-stage matching mechanism to select the optimal sub-pattern most relevant to a user input natural language query from a sub-pattern set; and S5, constructing a prompt word based on the optimal sub-pattern, the abstract corresponding to the optimal sub-pattern and the user natural language query, and inputting the prompt word into a large model to generate an SQL query statement. The technical scheme provided by the application fuses a database static structure and dynamic query behavior, automatically divides high correlation sub-patterns, accurately locates relevant sub-patterns through a two-stage matching, greatly reduces the input complexity of the large model, improves the accuracy and efficiency of SQL generation, and is suitable for large-scale database scenes.

This invention belongs to the field of computer vision and artificial intelligence technology, and relates to a multi-target tracking and segmentation method and system based on feature fusion and spatiotemporal memory. The multi-target tracking and segmentation method includes: acquiring the original state features and predicted bounding boxes of each target in each frame of the image, and constructing a spatiotemporal memory database containing a short-term buffer and a long-term storage area; for the first detected target in the multi-target video sequence, instantiating an independent long-term trajectory prototype for it, and establishing a corresponding index in the spatiotemporal memory database; performing hierarchical matching of the original state features and predicted bounding boxes of each target in the frame of the image based on the spatiotemporal memory database; updating the Kalman filter parameters and the spatiotemporal memory database based on the matching results; and continuing the tracking and segmentation of the next frame of the image using the updated Kalman filter and spatiotemporal memory database until each frame of the image has been traversed to obtain the multi-target tracking and segmentation result. This invention can effectively solve the problem of re-identification after long-term target occlusion.