16results about How to "Improve annotation accuracy" patented technology

The invention discloses an object labeling method and system based on laser point cloud, particularly relates to the technical field of laser point cloud processing, and is used for solving the problem of inaccurate labeling result caused by a complex reflection phenomenon caused by a non-Lambertian surface in the prior art. The method comprises the steps of obtaining original point cloud data collected by a laser radar and performing preprocessing; identifying a non-Lambertian reflection region based on the reflection intensity information and the geometric topology feature, and carrying out filtering correction; correcting false mirror image points through multi-layer iteration correction, and reconstructing boundaries to generate corrected point cloud data; grouping and classifying target objects in the corrected point cloud data based on a clustering algorithm combining global features and local features; finely dividing the boundary of the preliminary segmentation result by adopting an attention mechanism network, and evaluating the uncertainty of a boundary region through a local mutual information entropy; and adjusting a boundary division strategy according to an uncertainty evaluation result, and generating an optimized segmentation result.

The invention relates to the technical field of three-dimensional computer vision and point cloud processing, and provides a point cloud rapid labeling method based on a three-dimensional CAD model and RGB values thereof. The method aims at solving the problems that in the deep learning data set construction process, three-dimensional point cloud semantic information labeling efficiency is low, the labor cost is high, and the automatic labeling quality is poor. The color information of the three-dimensional point cloud is effectively converted into the semantic label, so that quick and automatic labeling of different three-dimensional and profile surfaces in the point cloud is realized. The method comprises the steps of establishing a category-RGB value-label mapping table, carrying out CAD entity modeling and assigning a standard color to an entity surface geometrical shape, setting entity model gridding parameters and exporting the entity model gridding parameters as a PLY format, assigning color RGB values of mesh model patches to vertexes, uniformly sampling expansion points of the mesh model patches, traversing the points and reading the RGB values, and finally obtaining a model. And adding labels according to the category-RGB value-label mapping table, and the like. The method can be applied to the fields of three-dimensional modeling, product quality detection, industrial design and the like.

The invention provides an image key point labeling method and device, electronic equipment and a computer readable storage medium. The image key point labeling method comprises the steps that gesture key point labeling information of an initial hand frame is acquired, the initial hand frame is an image containing a calibration plate, and a plurality of reference points are arranged on the calibration plate; obtaining a relative pose between a camera and a calibration plate when the initial hand frame is collected, and taking the relative pose as a first camera pose; acquiring a relative pose between the camera and the calibration plate when the current hand frame is acquired, and taking the relative pose as a second camera pose; and based on the first camera pose, the second camera pose and the gesture key point labeling information of the initial hand frame, performing key point labeling on the current hand frame to obtain the gesture key point labeling information of the current hand frame. According to the method, the image labeling efficiency and the image labeling precision can be improved under the condition that the fitting degree of the labeled image and the real image is ensured.

The invention relates to the technical field of image annotation processing, and provides an image annotation processing method and device, electronic equipment and a medium. According to the method, feature extraction processing can be performed on a to-be-recognized image to obtain a feature vector of the to-be-recognized image, and feature extraction processing is performed on a prompt image to obtain a feature vector of the prompt image; performing feature extraction processing on the position information of the labeled object in the prompt object to obtain a feature vector of the position information; performing feature extraction processing on the prompt text to obtain a feature vector of the prompt text; performing fusion processing on the feature vector of the prompt image, the feature vector of the position information and the feature vector of the prompt text to obtain a prompt vector of the annotation object; and according to the feature vector of the to-be-identified image and the prompt vector of the annotation object, annotating the to-be-identified image to obtain annotation information of the annotation object in the to-be-identified object. Through the feature extraction and fusion processing method, the image annotation effect and efficiency are remarkably improved.

The invention discloses a code abstract generation method and system based on hierarchical context awareness, and belongs to the field of natural language processing and software engineering. The method comprises the following steps: generating hierarchical annotations for a code warehouse based on a function call graph, firstly extracting a call relationship in the graph and eliminating a ring structure, and then classifying leaf nodes and generating annotations; a topological sequence from bottom to top is adopted in the core step, the importance score of a child node is calculated for each father node, child node annotations, class function descriptions and code bodies of the child nodes are fused, and a large language model is input to generate annotations till the whole graph is covered. The method is suitable for an automatic document generation scene of a complex software project, and through a technical path combining function call graph analysis, child node importance measurement and class function description, the accuracy and semantic consistency of code annotation are improved; the method can be widely applied to the fields of software maintenance, code understanding, intelligent development tools, automatic document generation of enterprise-level code warehouses and the like.

The invention relates to the technical field of engineering drawing recognition, and discloses a CAD drawing primitive-level semantic annotation method and device and a medium, and the method comprises the steps: obtaining a drawing dwg file, preprocessing the drawing, and forming a preprocessed drawing; all component instances in the preprocessed drawing are grouped in the form of blocks, and a grouped drawing is obtained; utilizing the layer information and the semantic information to establish a layer semantic correspondence table; forming an annotation file according to the corresponding relationship between the grouped drawings and the layer semantic corresponding table; reading the annotation file, screening all instances, carrying out semantic search for each instance to obtain the semantic category of the instance, carrying out primitive reverse annotation according to the semantic category of the instance, and generating a semantic category annotation result of the primitive, so that the primitive-level semantic annotation of the CAD drawing can be realized, and then a CAD vector drawing annotation data set is generated. Compared with a traditional image-based labeling method, the method has the advantage that the labeling precision of the vector diagram is obviously improved.

This invention discloses a semi-automatic annotation and improved deep learning-based method for detecting foreign objects on railway tracks. The method includes: semi-automatic annotation of a dataset based on thermal infrared human body segmentation results; improvement of the ResNet 50 neural network for all-weather outdoor railway track environments to identify different environmental scenarios; improvement of the YOLOv5 neural network based on scene recognition results to identify foreign objects on the tracks; extraction of railway track contours from images, combined with the location of the foreign object, to design an alarm strategy for foreign object intrusion on the tracks, and to perform tiered alarm judgments. By introducing deep learning scene classification combined with an improved target recognition network, this invention improves the effectiveness of foreign object identification on railway tracks in different environmental scenarios, enhances the accuracy of alarms for foreign object intrusion on railway tracks, and, combined with a scientific alarm strategy, reduces the possibility of train accidents caused by foreign object intrusion on railway tracks.

The application relates to the technical field of intelligent driving, in particular to a sparse-labeled obstacle label enhancement method and device, equipment and a storage medium, which comprises the following steps: based on the 3D box parameter of a sparse-labeled key frame target, a 3D box parameter change curve of the target is fitted and obtained, and interpolation is carried out based on the 3D box parameter change curve to obtain a 3D interpolation box of the target; a true value 2D box of a sparse-labeled adjacent key frame target is inversely projected to a 3D space to obtain a key frame 3D inverse projection box, interpolation is carried out to obtain a 3D inverse projection interpolation box, and a 2D optimized interpolation box is obtained by projection with the width-height ratio of the 3D inverse projection interpolation box as a constraint; the 3D interpolation box is corrected based on the 2D optimized interpolation box to obtain a 3D optimized interpolation box. The technical problem that the prior art completely depends on a 3D point cloud detection model, and the detection and interpolation precision sharply decreases due to insufficient point clouds in a point cloud sparse scene such as a long-distance, low obstacle and serious occlusion can be solved.

The invention relates to a road topological structure generation method based on a road end camera picture, and relates to the field of automatic driving. The method comprises the following steps: acquiring image data and vehicle data of a target intersection in a target time period; inputting the image data into a semantic segmentation model, and outputting an object preliminary identification result corresponding to the image data; inputting the vehicle data and the image data into a multi-target tracking network, and outputting target tracking result information corresponding to the target intersection; and generating road topological structure optimization data in combination with the object preliminary identification result and the target tracking result information. The road end camera is used for collecting data, the road topological structure at the crossroad is constructed through a computer vision method, vehicles can be helped to more accurately understand the traffic information of the road end, higher marking precision is achieved, and meanwhile manpower resources needed by manual marking are saved.

This invention relates to the field of spatial transcriptomics technology and provides a few-sample spatial transcriptomics cell annotation system and method based on graph cue learning, comprising: a preprocessing module; a graph construction module, which constructs spatial adjacency relationships based on cell spatial coordinates and constrains or weights graph connectivity relationships by combining inter-cell transcriptomics expression similarity; a graph representation learning module, which learns discriminative cell embedding representations from the cell graph structure under unsupervised or weakly supervised conditions; a cue fine-tuning module, which injects the category information contained in a small number of labeled samples into a pre-trained graph representation model in the downstream cell type annotation task; and a cell type prediction module, which completes cell type annotation based on a small number of labeled samples. This invention can fully integrate gene expression information and multi-source spatial relationships under conditions of very few labeled samples to construct highly discriminative cell representations, achieving high-precision automatic annotation of large-scale spatial transcriptomics data.

The present invention relates to the field of neural network technology, and in particular to a robot posture estimation method and system based on convolutional neural network. The method comprises acquiring industrial robot posture information data; automatically labeling the acquired industrial robot posture information data to obtain a data set; preprocessing the data set; training a convolutional neural network model using the preprocessed data set; and estimating the posture of the robot using the trained convolutional neural network model. The present invention generates a high-precision industrial robot arm posture data set by combining automatic labeling with manual calibration, thereby overcoming the time-consuming and inaccurate manual labeling problems in traditional methods.

The invention provides a target annotation method and device based on inter-frame association, and belongs to the technical field of image processing, and the method comprises the steps: constructing a target detection module based on spatial pyramid pooling and multi-dimensional feature fusion, inputting a to-be-annotated image set to the target detection module, obtaining an annotation set of each frame of image, and outputting the annotation set of each frame of image; dividing the annotation set of each frame of image into a high-confidence-coefficient set and a low-confidence-coefficient set; initializing the annotation set of the first frame of image into an associated track, predicting a target annotation box of the associated track in the next frame of image, and obtaining a prediction set; and based on the prediction set, associating the high-confidence-coefficient set of the next frame of image with the associated trajectory, performing trajectory updating by using the associated high-confidence-coefficient labeling box, associating the unassociated trajectory with the low-confidence-coefficient set of the next frame of image, performing trajectory updating, and continuously completing association of each frame of unassociated image. According to the processing scheme, the labeling precision is improved, and the labeling missing rate is reduced.

A point cloud annotation method, device, computer equipment and storage medium, comprising: obtaining a point cloud sequence to be annotated (S202); pre-annotating each frame of point cloud in the point cloud sequence to be annotated, obtaining pre-annotation information of each frame of point cloud, the pre-annotation information including: identification information and first pose information of an annotation frame for marking each target object in the corresponding frame of point cloud (S204); determining the annotation frame sequence corresponding to each target object according to the identification information, and determining the first point cloud in each annotation frame in the annotation frame sequence according to the first pose information of each annotation frame in the annotation frame sequence (S206); obtaining a point cloud registration result between corresponding annotation frames according to the first point cloud in each annotation frame in the annotation frame sequence, and obtaining the second pose information of each annotation frame in the annotation frame sequence based on the point cloud registration result (S028); determining the annotation result of the point cloud sequence to be annotated according to the second pose information of each annotation frame in each annotation frame sequence (S210). The method can improve the annotation efficiency and the annotation accuracy.

The invention provides a point cloud labeling method and device, and relates to the field of data processing. According to the invention, a plurality of nodes are created in the point cloud to mark the lane lines. When the nodes are marked, the plane positions of the nodes are preliminarily determined through the first marking operation of marking the visual angle, then the coordinates of the nodes in the height direction are determined through the second marking operation of adjusting the visual angle, and therefore all the nodes capable of reflecting the lane line are determined in the three-dimensional space. Therefore, the marking of the lane lines can be realized through additionally created nodes, and the marking difficulty is reduced. In addition, the nodes are presented through multiple visual angles, the distribution conditions of the point clouds in different coordinate systems can be presented more visually, and therefore the node labeling precision is improved.

The application discloses a cell type annotation method, device, equipment and medium, relates to the technical field of bioinformatics and medical image processing, and comprises the following steps: constructing a whole image feature matrix and a spatial neighborhood graph according to cell self characteristics and cell spatial positions of single cell samples of historical multiple tissue images; inputting the whole image feature matrix and the spatial neighborhood graph into a cell annotation model comprising a feature encoder, a spatial aggregator, a classification decoder and a prototype consistency constraint module; mapping the whole image feature matrix into an initial hidden vector by the encoder, determining attention coefficients of effective neighboring cells on the single cell samples based on the spatial neighborhood graph, weighting and summing the initial hidden vector according to the attention coefficients by the aggregator, outputting a predicted type probability distribution according to spatial perception characteristics by the decoder to generate a reconstructed feature; obtaining a prototype consistency loss value by the constraint module; updating and iterating the model based on the prototype consistency loss value to obtain a target cell annotation model; and performing cell type annotation by using the model.

Disclosed are a method and system for an autonomous driving cleaning robot to collect and produce high-precision maps, the method comprising: S1, synchronizing a laser radar sensor and a camera with a global satellite time system in real time through a controller, and collecting camera data and laser radar point cloud data; S2, extracting semantic information from the laser radar point cloud data and the camera data to obtain dynamic point information and road semantic information; S3, performing motion compensation and global pose optimization on the laser radar point cloud data, and converting the laser radar point cloud data into a point cloud set in a global coordinate system; S4, filtering out dynamic points in the point cloud set based on the dynamic point information, extracting statistical features of the point cloud set based on the road semantic information, and constructing a high-precision map based on the statistical features.