CN117271822A - Layout searching method and system based on multi-modal house type knowledge graph - Google Patents
Layout searching method and system based on multi-modal house type knowledge graph Download PDFInfo
- Publication number
- CN117271822A CN117271822A CN202311261743.XA CN202311261743A CN117271822A CN 117271822 A CN117271822 A CN 117271822A CN 202311261743 A CN202311261743 A CN 202311261743A CN 117271822 A CN117271822 A CN 117271822A
- Authority
- CN
- China
- Prior art keywords
- house type
- information
- type
- knowledge graph
- mode
- Prior art date
- Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
- Granted
Links
- 238000000034 method Methods 0.000 title claims abstract description 38
- 230000004927 fusion Effects 0.000 claims abstract description 17
- 238000013528 artificial neural network Methods 0.000 claims abstract description 16
- 230000006870 function Effects 0.000 claims description 32
- 238000013527 convolutional neural network Methods 0.000 claims description 12
- 230000002776 aggregation Effects 0.000 claims description 8
- 238000004220 aggregation Methods 0.000 claims description 8
- 238000004422 calculation algorithm Methods 0.000 claims description 7
- 238000004364 calculation method Methods 0.000 claims description 6
- 238000007781 pre-processing Methods 0.000 claims description 6
- 238000012545 processing Methods 0.000 claims description 6
- 238000012549 training Methods 0.000 claims description 6
- 238000013135 deep learning Methods 0.000 claims description 5
- 238000001514 detection method Methods 0.000 claims description 5
- 238000013136 deep learning model Methods 0.000 claims description 4
- 230000007246 mechanism Effects 0.000 claims description 4
- 238000005295 random walk Methods 0.000 claims description 4
- 238000012163 sequencing technique Methods 0.000 claims description 4
- 230000004913 activation Effects 0.000 claims description 3
- 238000013507 mapping Methods 0.000 claims description 3
- 238000010586 diagram Methods 0.000 description 9
- 238000012216 screening Methods 0.000 description 4
- 238000012986 modification Methods 0.000 description 3
- 230000004048 modification Effects 0.000 description 3
- 238000009877 rendering Methods 0.000 description 3
- 238000006467 substitution reaction Methods 0.000 description 3
- 230000009286 beneficial effect Effects 0.000 description 2
- 239000003086 colorant Substances 0.000 description 2
- 238000013461 design Methods 0.000 description 2
- 239000000463 material Substances 0.000 description 2
- 230000003287 optical effect Effects 0.000 description 2
- 230000008569 process Effects 0.000 description 2
- 238000007637 random forest analysis Methods 0.000 description 2
- 238000012706 support-vector machine Methods 0.000 description 2
- 238000004458 analytical method Methods 0.000 description 1
- 230000000903 blocking effect Effects 0.000 description 1
- 238000006243 chemical reaction Methods 0.000 description 1
- 238000004891 communication Methods 0.000 description 1
- 238000013499 data model Methods 0.000 description 1
- 238000011161 development Methods 0.000 description 1
- 230000000694 effects Effects 0.000 description 1
- 238000005516 engineering process Methods 0.000 description 1
- 238000002474 experimental method Methods 0.000 description 1
- 238000000605 extraction Methods 0.000 description 1
- 239000013307 optical fiber Substances 0.000 description 1
- 238000003672 processing method Methods 0.000 description 1
- 230000000007 visual effect Effects 0.000 description 1
Classifications
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F16/00—Information retrieval; Database structures therefor; File system structures therefor
- G06F16/50—Information retrieval; Database structures therefor; File system structures therefor of still image data
- G06F16/58—Retrieval characterised by using metadata, e.g. metadata not derived from the content or metadata generated manually
- G06F16/5866—Retrieval characterised by using metadata, e.g. metadata not derived from the content or metadata generated manually using information manually generated, e.g. tags, keywords, comments, manually generated location and time information
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F16/00—Information retrieval; Database structures therefor; File system structures therefor
- G06F16/30—Information retrieval; Database structures therefor; File system structures therefor of unstructured textual data
- G06F16/36—Creation of semantic tools, e.g. ontology or thesauri
- G06F16/367—Ontology
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F16/00—Information retrieval; Database structures therefor; File system structures therefor
- G06F16/50—Information retrieval; Database structures therefor; File system structures therefor of still image data
- G06F16/58—Retrieval characterised by using metadata, e.g. metadata not derived from the content or metadata generated manually
- G06F16/583—Retrieval characterised by using metadata, e.g. metadata not derived from the content or metadata generated manually using metadata automatically derived from the content
- G06F16/5846—Retrieval characterised by using metadata, e.g. metadata not derived from the content or metadata generated manually using metadata automatically derived from the content using extracted text
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
- G06N3/00—Computing arrangements based on biological models
- G06N3/02—Neural networks
- G06N3/04—Architecture, e.g. interconnection topology
- G06N3/045—Combinations of networks
- G06N3/0455—Auto-encoder networks; Encoder-decoder networks
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
- G06N3/00—Computing arrangements based on biological models
- G06N3/02—Neural networks
- G06N3/04—Architecture, e.g. interconnection topology
- G06N3/0464—Convolutional networks [CNN, ConvNet]
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
- G06N5/00—Computing arrangements using knowledge-based models
- G06N5/01—Dynamic search techniques; Heuristics; Dynamic trees; Branch-and-bound
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06V—IMAGE OR VIDEO RECOGNITION OR UNDERSTANDING
- G06V10/00—Arrangements for image or video recognition or understanding
- G06V10/40—Extraction of image or video features
- G06V10/44—Local feature extraction by analysis of parts of the pattern, e.g. by detecting edges, contours, loops, corners, strokes or intersections; Connectivity analysis, e.g. of connected components
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06V—IMAGE OR VIDEO RECOGNITION OR UNDERSTANDING
- G06V10/00—Arrangements for image or video recognition or understanding
- G06V10/40—Extraction of image or video features
- G06V10/44—Local feature extraction by analysis of parts of the pattern, e.g. by detecting edges, contours, loops, corners, strokes or intersections; Connectivity analysis, e.g. of connected components
- G06V10/443—Local feature extraction by analysis of parts of the pattern, e.g. by detecting edges, contours, loops, corners, strokes or intersections; Connectivity analysis, e.g. of connected components by matching or filtering
- G06V10/449—Biologically inspired filters, e.g. difference of Gaussians [DoG] or Gabor filters
- G06V10/451—Biologically inspired filters, e.g. difference of Gaussians [DoG] or Gabor filters with interaction between the filter responses, e.g. cortical complex cells
- G06V10/454—Integrating the filters into a hierarchical structure, e.g. convolutional neural networks [CNN]
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06V—IMAGE OR VIDEO RECOGNITION OR UNDERSTANDING
- G06V10/00—Arrangements for image or video recognition or understanding
- G06V10/70—Arrangements for image or video recognition or understanding using pattern recognition or machine learning
- G06V10/764—Arrangements for image or video recognition or understanding using pattern recognition or machine learning using classification, e.g. of video objects
- G06V10/765—Arrangements for image or video recognition or understanding using pattern recognition or machine learning using classification, e.g. of video objects using rules for classification or partitioning the feature space
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06V—IMAGE OR VIDEO RECOGNITION OR UNDERSTANDING
- G06V10/00—Arrangements for image or video recognition or understanding
- G06V10/70—Arrangements for image or video recognition or understanding using pattern recognition or machine learning
- G06V10/82—Arrangements for image or video recognition or understanding using pattern recognition or machine learning using neural networks
Landscapes
- Engineering & Computer Science (AREA)
- Theoretical Computer Science (AREA)
- Physics & Mathematics (AREA)
- General Physics & Mathematics (AREA)
- Evolutionary Computation (AREA)
- Data Mining & Analysis (AREA)
- General Engineering & Computer Science (AREA)
- Artificial Intelligence (AREA)
- Computing Systems (AREA)
- Databases & Information Systems (AREA)
- General Health & Medical Sciences (AREA)
- Health & Medical Sciences (AREA)
- Software Systems (AREA)
- Computational Linguistics (AREA)
- Computer Vision & Pattern Recognition (AREA)
- Multimedia (AREA)
- Life Sciences & Earth Sciences (AREA)
- Mathematical Physics (AREA)
- Library & Information Science (AREA)
- Biomedical Technology (AREA)
- Molecular Biology (AREA)
- Medical Informatics (AREA)
- Biophysics (AREA)
- Animal Behavior & Ethology (AREA)
- Biodiversity & Conservation Biology (AREA)
- Information Retrieval, Db Structures And Fs Structures Therefor (AREA)
Abstract
The invention discloses a layout searching method and a system based on a multi-modal house type knowledge graph, wherein the method comprises the following steps: acquiring multi-mode house type information of a plurality of house types; reconstructing the house type image into a house type drawing and a room drawing, and encoding to generate encoded data; determining the function type and furniture distribution of each room in the house type according to the house type image; organizing house type data according to the form of a multi-mode house type knowledge graph; respectively carrying out information fusion on the house type data of each house type in the multi-mode house type knowledge graph through a multi-mode neural network to generate first high-dimensional coding information of each house type layout; and carrying out tag information matching or high-dimensional coding information matching in the multi-mode house type knowledge graph according to the house type information to be matched, which is input by the user, so as to obtain a target house type layout set. The invention can perform multi-mode search, is more humanized, combines more house type information to construct a knowledge graph, and has more accurate and reasonable searched results, thereby meeting the requirements of various house type searches.
Description
Technical Field
The invention relates to the technical field of computers, in particular to a layout searching method and system based on a multi-modal house type knowledge graph.
Background
When considering the room layout, a general designer firstly measures the room size, selects proper furniture according to functions, ensures streamline layout, balances symmetrical and asymmetrical elements, selects main furniture positions, keeps the furniture size and the room proportion to be coordinated, creates a visual focus, coordinates colors and styles, fully utilizes the wall space, avoids blocking light rays, and ensures the practicability and aesthetic feeling of the layout through experiments and adjustment.
The house type layout search is used as a method for a user to design furniture placement in a specific house type, and not only relates to various graphic information of house type design drawings, but also relates to various text information of house types such as various labels, descriptions and the like. The general house type identification only focuses on the data covered in the drawing, and ignores additional information such as similar house type styles, positions, orientations and the like, so that the user cannot be completely matched with the user requirements in part of scenes. In the prior art, additional information is generally processed in a tag screening mode, and often due to uneven distribution of collected household type data, a situation that no suitable candidate household type exists in a long tail scene can be encountered.
Disclosure of Invention
The present invention aims to solve at least one of the technical problems in the related art to some extent. Therefore, the layout searching method and system based on the multi-modal house type knowledge graph, which are reasonable in searching result and high in searching accuracy, are provided.
In one aspect, an embodiment of the present invention provides a layout searching method based on a multi-modal user type knowledge graph, including:
acquiring multi-mode house type information of a plurality of house types; the multi-mode house type information comprises tag information, house type images and house type attribute introduction;
reconstructing the house type image into a house type drawing and a room drawing, and encoding to generate encoded data;
determining the function type and furniture distribution of each room in the house type according to the house type image;
organizing house type data according to the form of a multi-mode house type knowledge graph; wherein the household type data comprises the multi-mode household type information, the coded data, the function type and the furniture distribution; the multi-mode house type knowledge graph has a four-layer entity structure of a community, a house type, a room and furniture in sequence;
the method comprises the steps of respectively carrying out information fusion on the house type data of each house type in the multi-mode house type knowledge graph through a multi-mode neural network to generate first high-dimensional coding information of each house type layout;
and according to the user input household type information to be matched, performing tag information matching or high-dimensional coding information matching in the multi-mode household type knowledge graph to obtain a target household type layout set.
Optionally, the reconstructing the house type image into a house type drawing and a room drawing, and encoding, to generate encoded data, includes:
preprocessing the house type image; wherein the preprocessing comprises at least one of image denoising, graying and binarization;
identifying a first contour line segment in all house type images by adopting a feature detection algorithm of deep learning;
determining target contour segments representing doors, windows and walls from the first contour segments;
determining the direction, the starting point and the ending point of the target contour line segment according to the geometric characteristics of the target contour line segment;
taking the geometric feature, the direction, the starting point and the ending point as house type profile features;
inputting the house type profile features into an SVG drawing tool for image drawing and converting the SVG drawing tool into a standard PNG image to obtain a house type drawing;
dividing the house type drawing into a plurality of room drawings according to the house type outline characteristics;
and adopting an encoder of a deep learning model to encode the house drawing and the room drawing to obtain encoded data.
Optionally, the determining the function type and the furniture distribution of each room in the house type according to the house type image includes:
identifying a furniture bounding box in the house type image through a furniture bounding box identification model to obtain an identification result; the furniture bounding box comprises position information and boundary information of furniture; the furniture bounding box recognition model is constructed based on a convolutional neural network;
mapping the furniture bounding box in a house type drawing according to the position information;
determining the quantity and distribution of furniture in the room according to the identification result by using a classifier;
and determining the function type of the room according to the furniture quantity.
Optionally, the organizing the house type data according to the form of the multi-mode house type knowledge graph includes:
based on historical house type experience, constructing a graph database and a vector database of a four-layer entity structure to form a multi-mode house type knowledge graph database; wherein the vector database is associated with the graph database by means of an index;
storing all characters in the multi-mode household type information into the graph database in a graph form;
the encoded data is stored in a vector database.
Optionally, the generating, by using a multi-modal neural network, the first high-dimensional coding information of each household type layout by respectively performing information fusion on household type data of each household type in the multi-modal household type knowledge graph includes:
carrying out random walk processing on the multi-modal house type knowledge graph, and determining attribute sets of all entities in the house type layout;
and based on a self-attention mechanism of the multi-modal neural network, carrying out aggregation processing on the attribute set to obtain the first high-dimensional coding information of the house type.
Optionally, the performing tag information matching or high-dimensional coding information matching in the multi-mode family type knowledge graph according to family type information to be matched input by a user to obtain a target family type layout set includes:
searching entity nodes in the multi-mode household type knowledge graph in an index mode according to the label information in the household type information to be matched;
determining the entity range associated with each type of tag information according to the entity node;
the union sets are taken for all the layout entity ranges, and all first house type layouts which accord with the label range are obtained;
when the house type information to be matched comprises a contour drawing, converting the house type coding information to be matched into high-dimensional coding information to be matched, carrying out high-dimensional coding matching on the high-dimensional coding information and all the first house type layouts, and calculating Euclidean distance according to similarity;
and sequencing the first house type layout according to the Euclidean distance to obtain a target house type layout set.
Optionally, the performing tag information matching or high-dimensional coding information matching in the multi-mode family type knowledge graph according to family type information to be matched input by a user to obtain a target family type layout set includes:
organizing data of the to-be-matched house type information according to a multi-mode house type knowledge graph;
calculating an initial multi-mode code of each entity in the family type information to be matched;
performing iterative fusion on all the initial multi-mode codes through a multi-mode model to obtain high-dimensional coding information to be matched;
calculating Euclidean distance between the high-dimensional coding information to be matched and first high-dimensional coding information of each household type layout in the multi-mode household type knowledge graph database;
and sequencing the house type layouts according to the Euclidean distance to generate a target house type layout set.
Optionally, the calculation formula of the initial multi-mode coding is:
wherein,representing an initial multi-modal encoding; mean represents the mean of solving feature codes; v ip Feature coding of the attribute p of the entity node i; p represents the attribute of the entity node i; n is n i Representing all attribute sets belonging to entity node i.
Optionally, the calculation formula of the high-dimensional coding information is:
wherein,information aggregation coding information representing the present round; sigma is the activation function of the multimodal model; w is a weight obtained by training the multi-modal model; />Information aggregation coding representing a previous round; j represents the i-th current entity node circumferenceCoding the j-th target entity node of the periphery; n is the total number of target entity nodes around the ith current entity node.
On the other hand, the embodiment of the invention also discloses a layout searching system based on the multi-modal house type knowledge graph, which comprises the following steps:
the first module is used for acquiring multi-mode house type information of a plurality of house types; the multi-mode house type information comprises tag information, house type images and house type attribute introduction;
the second module is used for reconstructing the house type image into a house type drawing and a room drawing and encoding to generate encoded data;
the third module is used for determining the function type and furniture distribution of each room in the house type according to the house type image;
a fourth module for organizing house type data according to the multi-mode house type knowledge graph; wherein the household type data comprises the multi-mode household type information, the coded data, the function type and the furniture distribution; the multi-mode house type knowledge graph has a four-layer entity structure of a community, a house type, a room and furniture in sequence;
a fifth module, configured to perform information fusion on each household type data of each household type in the multi-mode household type knowledge graph through a multi-mode neural network, and generate first high-dimensional coding information of each household type layout;
and a sixth module, configured to perform tag information matching or high-dimensional coding information matching in the multi-mode family knowledge graph according to family information to be matched input by a user, so as to obtain a target family layout set.
In another aspect, an embodiment of the present invention further provides an electronic device, including: a processor and a memory; the memory is used for storing programs; the processor executes the program to implement the method as described above.
In another aspect, embodiments of the present invention also provide a computer storage medium in which a processor-executable program is stored, which when executed by a processor is configured to implement the method as described above.
The embodiment of the invention has the following beneficial effects: through the combination of the multi-modal network and the multi-modal house type knowledge graph, multi-modal search can be performed, humanization is achieved, house type search requirements of various requirements are met, in addition, the embodiment of the invention combines more house type information to construct the knowledge graph, layout matching of high-dimensional feature coding is performed based on the multi-modal house type knowledge graph, and the result obtained by matching is accurate and reasonable.
Drawings
The accompanying drawings are included to provide a further understanding of the invention and are incorporated in and constitute a part of this specification, illustrate and do not limit the invention.
Fig. 1 is a step diagram of a layout searching method based on a multi-modal house type knowledge graph provided by an embodiment of the present invention;
fig. 2 is a flow chart of a layout searching method based on a multi-modal house type knowledge graph according to an embodiment of the present invention;
FIG. 3 is a data model structure diagram of a multi-modal house type knowledge graph provided by an embodiment of the present invention;
FIG. 4 is a schematic diagram of a layout search process provided by an embodiment of the present invention;
fig. 5 is a schematic structural diagram of a layout search system based on a multi-modal house type knowledge graph according to an embodiment of the present invention;
fig. 6 is a schematic structural diagram of an electronic device according to an embodiment of the present invention.
Detailed Description
The present invention will be described in further detail with reference to the drawings and examples, in order to make the objects, technical solutions and advantages of the present invention more apparent. It should be understood that the specific embodiments described herein are for purposes of illustration only and are not intended to limit the scope of the invention.
It should be noted that although functional block diagrams are depicted as block diagrams, and logical sequences are shown in the flowchart, in some cases, the steps shown or described may be performed in a different order than the block diagrams in the system. The terms first/S100, second/S200, and the like in the description and in the claims and in the above-described figures, are used for distinguishing between similar objects and not necessarily for describing a particular sequential or chronological order.
Reference herein to "an embodiment" means that a particular feature, structure, or characteristic described in connection with the embodiment may be included in at least one embodiment of the invention. The appearances of such phrases in various places in the specification are not necessarily all referring to the same embodiment, nor are separate or alternative embodiments mutually exclusive of other embodiments. Those of skill in the art will explicitly and implicitly appreciate that the embodiments described herein may be combined with other embodiments.
In order to solve at least one of the technical problems in the related art to a certain extent, the embodiment of the invention provides a layout searching method and system based on a multi-mode house type knowledge graph.
Referring to fig. 1 and 2, the layout searching method based on the multi-modal house type knowledge graph according to the embodiment of the present invention may include, but is not limited to, the following steps S100 to S600.
S100, acquiring multi-mode house type information of a plurality of house types; the multi-mode house type information comprises tag information, house type images and house type introduction.
The method comprises the steps of acquiring house type information of a plurality of house types, wherein the house type information comprises house type images, tag information and house type introduction, and the house type introduction can be the position, year data, lighting orientation and the like of a community, and can also comprise other house type data. The house type image can have a plurality of rooms, and furniture can be arranged in the rooms. Specifically, the house type image may be acquired from a scanner, a camera, or other image acquisition device.
S200, reconstructing the house type image into a house type drawing and a room drawing, and encoding to generate encoded data.
Specifically, step S200 includes the following steps S210 to S280.
S210, preprocessing the house type image; wherein the preprocessing comprises at least one of image denoising, graying and binarization.
Further, operations including image denoising, graying, binarization and the like are performed on the house type image, so that the follow-up contour extraction of the house type image is facilitated.
S220, identifying first contour line segments in all house type images by adopting a deep learning feature detection algorithm.
Detecting all possible contour line segments in the image through a feature detection algorithm of deep learning, and taking the contour line segments as first contour line segments; specifically, the feature detection algorithm may employ a convolutional neural network CNN, such as a residual network resnet, yolo, DETR algorithm, for extracting and learning the picture features of the house type image.
S230, determining target contour line segments representing doors, windows and walls from the first contour line segments.
And screening out the contours possibly representing structures such as doors, windows, walls and the like from the extracted first contour line segments.
S240, determining the direction, the starting point and the ending point of the target contour line segment according to the geometric characteristics of the target contour line segment.
And connecting and analyzing contour line segments for the screened possible structural contours. And combining a plurality of line segments into a complete structural contour by a method of communication analysis and the like, and determining a starting point and an ending point of the contour. The direction of the structural contour is determined based on geometric features of the contour line segments, such as slope and angle information. The direction data of the contour is obtained by calculating the angle between the two points.
Specifically, a contour endpoint is found from the contour, and the coincidence degree phi with the contour is calculated by combining the two lines after the repeated points are combined according to the threshold value.
Wherein L is Contour line Representing the connecting lines found by the contour endpoints of the door, the wall and the window; s is S Contour profile Representing the outline areas of the door, the wall and the window detected by the algorithm; l (L) Contour line ∩S Contour profile The contour line represents the contourLine segments within the region.
When the overlapping degree is larger than the threshold value, the two end points are marked as the starting point (x Starting point ,y Starting point ) And end point (x) Endpoint (endpoint) ,y Endpoint (endpoint) ) The two-point line serves as the contour line of the contour. The angle information θ (direction here means angle value) of the contour is calculated from the contour line.
S250, taking geometric features, directions, starting points and ending points as house type profile features.
S260, inputting the house type outline features into an SVG drawing tool for image drawing and converting the house type outline features into standard PNG images to obtain house type drawings.
In order to avoid that house type images with various styles influence the subsequent house type matching and layout matching effects, a SVG (scalable vector graphics) image processing method is adopted to hold the house type images; SVG is an XML-based graphics format and SVG image processing is used to create scalable, high quality vector graphics and images. According to the embodiment of the invention, the standardized house type drawing is redrawn according to the identified information.
Creating an SVG rendering environment can be based on existing SVG rendering libraries or self-developed rendering tools. In the drawing environment, drawing parameters such as coordinate axes, scales, colors, and the like are set. And drawing the line segments of the wall in the SVG drawing environment according to the data of the starting point and the ending point of the imported wall. From the direction data, the orientation of the wall, and thus the position and length of the line segments, can be determined.
After the SVG image is obtained, the SVG image is converted into a standard PNG image format by using a conversion method from SVG to PNG, and a house type drawing is obtained. In the process, the identification degree of the image can be improved by setting parameters such as resolution, transparency and the like of the PNG image.
S270, dividing the house type drawing into a plurality of room drawings according to the house type outline characteristics.
And further, the whole house type drawing is segmented into drawings of a plurality of rooms according to the closed outline.
S280, coding the house type drawing and the room drawing by adopting a coder of the deep learning model to obtain coding data.
A deep learning model, such as a Convolutional Neural Network (CNN) based encoder, is designed for PNG image encoding. A coded training data set is prepared containing the original PNG image and the corresponding coded data. By training on this data, the encoder is enabled to learn the features and patterns in the PNG image for more efficient encoding.
During storage, house drawings and room drawings can be respectively input into a trained deep learning encoder, and PNG images are converted into compressed encoded data by the encoder. These encoded data occupy less memory space relative to the original PNG image while preserving important information of the image.
And S300, determining the function type and furniture distribution of each room in the house type according to the house type image.
Specifically, step S300 may include the following steps S310 to S340.
S310, recognizing a furniture bounding box in the house type image through a furniture bounding box recognition model to obtain a recognition result; the furniture bounding box comprises position information and boundary information of furniture; furniture bounding box recognition models are constructed based on convolutional neural networks.
A drawing dataset is collected having labeled furniture bounding boxes, where each bounding box specifies a location and boundaries of a piece of furniture. The data set will be used to train and validate Convolutional Neural Network (CNN) models.
And constructing a furniture bounding box recognition model based on the CNN model, and detecting and recognizing the furniture bounding box in the drawing. In the model training stage, the CNN model is trained by using the collected home bounding box drawing data set, so that the CNN model can learn the characteristics and modes of furniture.
Inputting the house type images into a trained home bounding box recognition model, and recognizing furniture bounding boxes in all rooms through forward propagation of the model. The furniture bounding box includes location and boundary information for the furniture.
S320, mapping the furniture bounding box in the house type drawing according to the position information.
S330, determining the number and distribution of furniture in the room according to the identification result by using a classifier.
A dataset is collected having individual room furniture counts and corresponding function labels. Each sample contains furniture quantity statistics and a true function label for the room. A classifier, such as a Support Vector Machine (SVM), random Forest (Random Forest), or neural network, is constructed for predicting the function of the room based on the features. The classifier is trained using the drawing dataset of step S310 so that it can learn the relationship between furniture quantity features and room functions.
S340, determining the function type of the room according to the number of furniture.
And inputting the furniture quantity statistical data of the room to be presumed into a trained classifier, and presuming the function of the room through the output of the classifier. The classifier will assign a predicted functional label to the room based on the furniture quantity characteristics. Here tags such as living room, bedroom, bathroom, kitchen, etc.
S400, organizing house type data according to a multi-mode house type knowledge graph; the house type data comprise multi-mode house type information, coding data, function types and furniture distribution; the multi-mode house type knowledge graph has four layers of entity structures of a community, a house type, a room and furniture in sequence.
In the embodiment of the invention, a knowledge graph of a four-layer entity structure can be constructed based on historical house type knowledge experience, the four-layer entity is respectively a community, a house type, a room and furniture, each entity has the corresponding attribute, and all characters are stored in a graph database in a graph form; the coded data of the related image and other multi-mode data are stored in a vector database and are related with the graph database in an index mode.
Illustratively, referring to fig. 3, a cell entity may have cell location and year attributes; the house type entity can have house type drawing and size attributes; the room entity may have properties of layout drawings, outline codes, and room functions; furniture entities may have attributes of furniture type, furniture style, and furniture brands. It should be noted that the specific attribute is merely an example, and in other embodiments, other attributes may be used.
Specifically, step S400 may include the following steps S410 to S430.
S410, constructing a graph database and a vector database of a four-layer entity structure based on historical house type experience to form a multi-mode house type knowledge graph database; wherein the vector database is associated with the graph database by way of an index.
S420, storing all characters in the multi-mode user type information into a graph database in a graph mode.
S430, storing the encoded data into a vector database.
S500, respectively carrying out information fusion on the house type data of each house type in the multi-mode house type knowledge graph through the multi-mode neural network, and generating first high-dimensional coding information of each house type layout.
The embodiment of the invention uses the multi-modal neural network to fuse the coding information of each entity attribute in the house type data of different house types in the multi-modal house type knowledge graph, and projects all the coding information to the same dimension of a higher level to fuse the coding information. Specifically, by means of random walk on the multi-mode house type knowledge graph, an entity attribute ebedding (coding information) set related to each layout is found, various ebedding of graphics, characters and the like is aggregated through a self-attention mechanism by adopting a unified model architecture, and finally a unified ebedding expression form aiming at the entity is formed and is called as first high-dimensional coding information and stored in a corresponding vector database.
Initial multi-modal encoding of each entity by multi-modal model firstUnified feature code v obtained by multi-modal model for attribute data thereof ip ,p∈n i The average value of (2), namely:
wherein,representing an initial multi-modal encoding; mean represents the mean of solving feature codes; v ip Feature coding for the attribute p of the node i; p represents the attribute of node i; n is n i Representing all the attribute sets belonging to node i.
The current entity node i, in the kth-1 round, the multi-mode coding information of N adjacent cells, house, room and furniture entities (called target entity nodes) are fused to obtain kth high-dimensional coding informationThe fusion mode is as follows:
wherein,information aggregation coding information representing the present round; sigma is the activation function of the model; w is the weight of the multi-modal model training; />Information aggregation coding representing a previous round; j represents the code of the j-th target entity node around the i-th current entity node; n is the total number of target entity nodes around the ith current entity node.
Based on this, step S500 may include the following steps S510 to S520.
S510, carrying out random walk processing on the multi-mode house type knowledge graph, and determining attribute sets of all entities in the house type layout.
S520, based on a self-attention mechanism of the multi-modal neural network, the attribute set is aggregated to obtain the first high-dimensional coding information of the house type.
And S600, performing tag information matching or high-dimensional coding information matching in the multi-mode house type knowledge graph according to the house type information to be matched, which is input by the user, so as to obtain a target house type layout set.
Specifically, a single-hop query method and/or a multi-hop query method can be established for querying information of adjacent or neighboring nodes of the target entity node in the knowledge graph, and secondary development is performed according to the graph database API interface and the vector database API interface, so that quick association query between the object and the attribute is realized.
Further, step S600 may include at least one of the following (1) to (2):
(1) And (5) screening and searching.
Referring to fig. 4, by reading various label information input by a user, the associated entity node is rapidly indexed in the knowledge graph, and the entity node at the upper layer or the lower layer is iteratively searched according to the searched entity node until the target entity range associated with the label information is obtained. And combining target entity ranges corresponding to various label information to obtain a union set, so that all layout information in the label range can be searched. If the user provides the house type or room structure outline drawing, the user converts the house type or room structure outline drawing into codes by the method, matches the codes with layout information in the range of the conforming labels, calculates Euclidean distance according to similarity, and outputs matching results according to the distance order.
Based on this, the step of screening the search may specifically include the following steps S610 to S650.
S610, searching entity nodes in the multi-mode house type knowledge graph in an index mode according to the label information in the house type information to be matched.
S620, determining the entity range associated with various label information according to the entity node.
S630, the union sets are obtained for all the layout entity ranges, and all the first house type layouts which are in the range of the labels are obtained.
And S640, when the house type information to be matched comprises a contour drawing, converting the house type coding information to be matched into high-dimensional coding information to be matched, carrying out high-dimensional coding matching with all the first house type layout, and calculating the Euclidean distance according to the similarity.
S650, sorting the first house type layout according to the Euclidean distance to obtain a target house type layout set.
(2) And (5) merging the code searches.
Referring to fig. 4, after all data uploaded by a user are encoded according to different types, the encoded data are organized by using the structure of the multi-modal user type knowledge graph provided by the embodiment of the invention, so as to obtain a data set, and then the data set is input into a multi-modal neural network to calculate a corresponding layout entity ebedding. And calculating Euclidean distance according to the similarity between the calculated ebedding result and the layout entity ebedding result, and taking the minimum distance value as a matched house type result.
When performing layout search, generating a fusion code of the current environment and performing similarity calculation on layout entities in the knowledge graph, and specifically, according to the Euclidean distance calculation formula, solving a multi-mode fusion code finally obtained by all data uploaded by the current userThe Euclidean distance of fusion codes with other layouts in the database is recommended from small to large.
Based on this, the step of merging the code searches may specifically include the following steps S660 to S6100.
And S660, organizing data of the to-be-matched house type information according to the multi-mode house type knowledge graph.
S670, calculating initial multi-mode codes of each entity in the family type information to be matched.
S680, performing iterative fusion on all initial multi-mode codes through the multi-mode model to obtain high-dimensional code information to be matched.
S690, calculating Euclidean distance between the high-dimensional coding information to be matched and the first high-dimensional coding information of each household type layout in the multi-mode household type knowledge graph database.
S6100, sorting the house type layouts according to the Euclidean distance to generate a target house type layout set.
On the other hand, as shown in fig. 5, an embodiment of the present invention provides a layout search system based on a multi-modal house type knowledge graph, including:
the first module is used for acquiring multi-mode house type information of a plurality of house types; the multi-mode house type information comprises tag information, house type images and house type attribute introduction;
the second module is used for reconstructing the house type image into house type drawings and room drawings and coding the house type drawings to generate coding data;
the third module is used for determining the function type and furniture distribution of each room in the house type according to the house type image;
a fourth module for organizing house type data according to the multi-mode house type knowledge graph; the house type data comprise multi-mode house type information, coding data, function types and furniture distribution; the multi-mode house type knowledge graph has a four-layer entity structure of a community, a house type, a room and furniture in sequence;
a fifth module, configured to perform information fusion on each household type data in the multi-mode household type knowledge graph through the multi-mode neural network, and generate first high-dimensional coding information of each household type layout;
and the sixth module is used for carrying out tag information matching or high-dimensional coding information matching in the multi-mode house type knowledge graph according to the house type information to be matched, which is input by the user, so as to obtain a target house type layout set.
On the other hand, as shown in fig. 6, an embodiment of the present invention further provides an electronic device, including: a processor and a memory; the memory is used for storing programs; the processor executes the program to implement the method as described above.
In another aspect, embodiments of the present invention also provide a computer storage medium in which a processor-executable program is stored, which when executed by a processor is configured to implement the method as above.
The embodiment of the invention has the following beneficial effects:
1. the multi-modal searching can be carried out, the humanization is better, the house type searching requirements of various requirements are met, in addition, the embodiment of the invention combines more house type information to construct a knowledge graph, and the searched result is more accurate and reasonable based on multi-modal house type knowledge matching.
2. After the method is applied, the user can obtain the corresponding layout search result according to more or less prompt information, the situation that the recommendation result is zero when the fine classification is too small is improved, the selection can be carried out through simple labels, the search can also be carried out through drawings, and the experience of the user during the search is improved.
3. The importance weights of various inputs do not need to be manually allocated.
An application example of the embodiment of the present invention is described below:
acquiring multi-mode house type information of a plurality of house types; the multi-mode house type information comprises tag information, house type images and house type attribute introduction; reconstructing the house type image into a house type drawing and a room drawing, and encoding to generate encoded data; determining the function type and furniture distribution of each room in the house type according to the house type image; organizing house type data according to the form of a multi-mode house type knowledge graph; the house type data comprise multi-mode house type information, coding data, function types and furniture distribution; the multi-mode house type knowledge graph has a four-layer entity structure of a community, a house type, a room and furniture in sequence; respectively carrying out information fusion on the house type data of each house type in the multi-mode house type knowledge graph through a multi-mode neural network to generate first high-dimensional coding information of each house type layout; and carrying out tag information matching or high-dimensional coding information matching in the multi-mode house type knowledge graph according to the house type information to be matched, which is input by the user, so as to obtain a target house type layout set.
In some alternative embodiments, the functions/acts noted in the block diagrams may occur out of the order noted in the operational illustrations. For example, two blocks shown in succession may in fact be executed substantially concurrently or the blocks may sometimes be executed in the reverse order, depending upon the functionality/acts involved. Furthermore, the embodiments presented and described in the flowcharts of the present invention are provided by way of example in order to provide a more thorough understanding of the technology. The disclosed methods are not limited to the operations and logic flows presented herein. Alternative embodiments are contemplated in which the order of various operations is changed, and in which sub-operations described as part of a larger operation are performed independently.
The functions, if implemented in the form of software functional units and sold or used as a stand-alone product, may be stored in a computer-readable storage medium. Based on this understanding, the technical solution of the present invention may be embodied essentially or in a part contributing to the prior art or in a part of the technical solution in the form of a software product stored in a storage medium, comprising several instructions for causing a computer device (which may be a personal computer, a server, a network device, etc.) to perform all or part of the steps of the method of the embodiments of the present invention. And the aforementioned storage medium includes: a U-disk, a removable hard disk, a Read-Only Memory (ROM), a random access Memory (RAM, random Access Memory), a magnetic disk, or an optical disk, or other various media capable of storing program codes.
Logic and/or steps represented in the flowcharts or otherwise described herein, e.g., a ordered listing of executable instructions for implementing logical functions, can be embodied in any computer-readable medium for use by or in connection with an instruction execution system, apparatus, or device, such as a computer-based system, processor-containing system, or other system that can fetch the instructions from the instruction execution system, apparatus, or device and execute the instructions. For the purposes of this description, a "computer-readable medium" can be any means that can contain, store, communicate, propagate, or transport the program for use by or in connection with the instruction execution system, apparatus, or device.
More specific examples (a non-exhaustive list) of the computer-readable medium would include the following: an electrical connection (electronic device) having one or more wires, a portable computer diskette (magnetic device), a Random Access Memory (RAM), a read-only memory (ROM), an erasable programmable read-only memory (EPROM or flash memory), an optical fiber device, and a portable compact disc read-only memory (CDROM). Additionally, the computer-readable medium may even be paper or other suitable medium upon which the program is printed, as the program may be electronically captured, via, for instance, optical scanning of the paper or other medium, then compiled, interpreted or otherwise processed in a suitable manner, if necessary, and then stored in a computer memory.
In the description of the present specification, a description referring to terms "one embodiment," "some embodiments," "examples," "specific examples," or "some examples," etc., means that a particular feature, structure, material, or characteristic described in connection with the embodiment or example is included in at least one embodiment or example of the present invention. In this specification, schematic representations of the above terms do not necessarily refer to the same embodiments or examples. Furthermore, the particular features, structures, materials, or characteristics described may be combined in any suitable manner in any one or more embodiments or examples.
While embodiments of the present invention have been shown and described, it will be understood by those of ordinary skill in the art that: many changes, modifications, substitutions and variations may be made to the embodiments without departing from the spirit and principles of the invention, the scope of which is defined by the claims and their equivalents.
While the preferred embodiment of the present invention has been described in detail, the present invention is not limited to the embodiments described above, and those skilled in the art can make various equivalent modifications or substitutions without departing from the spirit of the present invention, and these equivalent modifications or substitutions are included in the scope of the present invention as defined in the appended claims.
Claims (10)
1. The layout searching method based on the multi-modal house type knowledge graph is characterized by comprising the following steps of:
acquiring multi-mode house type information of a plurality of house types; the multi-mode house type information comprises tag information, house type images and house type attribute introduction;
reconstructing the house type image into a house type drawing and a room drawing, and encoding to generate encoded data;
determining the function type and furniture distribution of each room in the house type according to the house type image;
organizing house type data according to the form of a multi-mode house type knowledge graph; wherein the household type data comprises the multi-mode household type information, the coded data, the function type and the furniture distribution; the multi-mode house type knowledge graph has a four-layer entity structure of a community, a house type, a room and furniture in sequence;
the method comprises the steps of respectively carrying out information fusion on the house type data of each house type in the multi-mode house type knowledge graph through a multi-mode neural network to generate first high-dimensional coding information of each house type layout;
and according to the user input household type information to be matched, performing tag information matching or high-dimensional coding information matching in the multi-mode household type knowledge graph to obtain a target household type layout set.
2. The layout searching method based on the multi-modal house type knowledge graph according to claim 1, wherein the reconstructing the house type image into house type drawing and room drawing and encoding, generating encoded data, includes:
preprocessing the house type image; wherein the preprocessing comprises at least one of image denoising, graying and binarization;
identifying a first contour line segment in all house type images by adopting a feature detection algorithm of deep learning;
determining target contour segments representing doors, windows and walls from the first contour segments;
determining the direction, the starting point and the ending point of the target contour line segment according to the geometric characteristics of the target contour line segment;
taking the geometric feature, the direction, the starting point and the ending point as house type profile features;
inputting the house type profile features into an SVG drawing tool for image drawing and converting the SVG drawing tool into a standard PNG image to obtain a house type drawing;
dividing the house type drawing into a plurality of room drawings according to the house type outline characteristics;
and adopting an encoder of a deep learning model to encode the house drawing and the room drawing to obtain encoded data.
3. The layout searching method based on the multi-modal house type knowledge graph according to claim 1, wherein the determining the function type and the furniture distribution of each room in the house type according to the house type image comprises:
identifying a furniture bounding box in the house type image through a furniture bounding box identification model to obtain an identification result; the furniture bounding box comprises position information and boundary information of furniture; the furniture bounding box recognition model is constructed based on a convolutional neural network;
mapping the furniture bounding box in a house type drawing according to the position information;
determining the quantity and distribution of furniture in the room according to the identification result by using a classifier;
and determining the function type of the room according to the furniture quantity.
4. The layout searching method based on the multi-modal house type knowledge graph according to claim 1, wherein the organizing house type data according to the form of the multi-modal house type knowledge graph comprises:
based on historical house type experience, constructing a graph database and a vector database of a four-layer entity structure to form a multi-mode house type knowledge graph database; wherein the vector database is associated with the graph database by means of an index;
storing all characters in the multi-mode household type information into the graph database in a graph form;
the encoded data is stored in a vector database.
5. The layout searching method based on the multi-modal household pattern knowledge graph according to claim 1, wherein the generating the first high-dimensional coding information of each household pattern layout by performing information fusion on household pattern data of each household pattern in the multi-modal household pattern knowledge graph through a multi-modal neural network comprises:
carrying out random walk processing on the multi-modal house type knowledge graph, and determining attribute sets of all entities in the house type layout;
and based on a self-attention mechanism of the multi-modal neural network, carrying out aggregation processing on the attribute set to obtain the first high-dimensional coding information of the house type.
6. The layout searching method based on the multi-modal house type knowledge graph according to claim 1, wherein the performing tag information matching or high-dimensional coding information matching in the multi-modal house type knowledge graph according to the house type information to be matched input by the user to obtain the target house type layout set comprises the following steps:
searching entity nodes in the multi-mode household type knowledge graph in an index mode according to the label information in the household type information to be matched;
determining the entity range associated with each type of tag information according to the entity node;
the union sets are taken for all the layout entity ranges, and all first house type layouts which accord with the label range are obtained;
when the house type information to be matched comprises a contour drawing, converting the house type coding information to be matched into high-dimensional coding information to be matched, carrying out high-dimensional coding matching on the high-dimensional coding information and all the first house type layouts, and calculating Euclidean distance according to similarity;
and sequencing the first house type layout according to the Euclidean distance to obtain a target house type layout set.
7. The layout searching method based on the multi-modal house type knowledge graph according to claim 1, wherein the performing tag information matching or high-dimensional coding information matching in the multi-modal house type knowledge graph according to the house type information to be matched input by the user to obtain the target house type layout set comprises the following steps:
organizing data of the to-be-matched house type information according to a multi-mode house type knowledge graph;
calculating an initial multi-mode code of each entity in the family type information to be matched;
performing iterative fusion on all the initial multi-mode codes through a multi-mode model to obtain high-dimensional coding information to be matched;
calculating Euclidean distance between the high-dimensional coding information to be matched and first high-dimensional coding information of each household type layout in the multi-mode household type knowledge graph database;
and sequencing the house type layouts according to the Euclidean distance to generate a target house type layout set.
8. The layout searching method based on the multi-modal house type knowledge graph according to claim 7, wherein the calculation formula of the initial multi-modal coding is:
wherein,representing an initial multi-modal encoding; mean represents the mean of solving feature codes; v ip Feature coding of the attribute p of the entity node i; p represents the attribute of the entity node i; n is n i Representing all belonging to a physical nodei, a set of attributes of i.
9. The layout searching method based on the multi-modal knowledge graph as claimed in claim 7, wherein the calculation formula of the high-dimensional coding information is:
wherein,information aggregation coding information representing the present round; sigma is the activation function of the multimodal model; w is a weight obtained by training the multi-modal model; />Information aggregation coding representing a previous round; j represents the code of the j-th target entity node around the i-th current entity node; n is the total number of target entity nodes around the ith current entity node.
10. The layout searching system based on the multi-modal house type knowledge graph is characterized by comprising the following components:
the first module is used for acquiring multi-mode house type information of a plurality of house types; the multi-mode house type information comprises tag information, house type images and house type attribute introduction;
the second module is used for reconstructing the house type image into a house type drawing and a room drawing and encoding to generate encoded data;
the third module is used for determining the function type and furniture distribution of each room in the house type according to the house type image;
a fourth module for organizing house type data according to the multi-mode house type knowledge graph; wherein the household type data comprises the multi-mode household type information, the coded data, the function type and the furniture distribution; the multi-mode house type knowledge graph has a four-layer entity structure of a community, a house type, a room and furniture in sequence;
a fifth module, configured to perform information fusion on each household type data of each household type in the multi-mode household type knowledge graph through a multi-mode neural network, and generate first high-dimensional coding information of each household type layout;
and a sixth module, configured to perform tag information matching or high-dimensional coding information matching in the multi-mode family knowledge graph according to family information to be matched input by a user, so as to obtain a target family layout set.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202311261743.XA CN117271822B (en) | 2023-09-26 | 2023-09-26 | Layout searching method and system based on multi-modal house type knowledge graph |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202311261743.XA CN117271822B (en) | 2023-09-26 | 2023-09-26 | Layout searching method and system based on multi-modal house type knowledge graph |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN117271822A true CN117271822A (en) | 2023-12-22 |
| CN117271822B CN117271822B (en) | 2024-04-09 |
Family
ID=89215664
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN202311261743.XA Active CN117271822B (en) | 2023-09-26 | 2023-09-26 | Layout searching method and system based on multi-modal house type knowledge graph |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN117271822B (en) |
Citations (4)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN115049437A (en) * | 2022-06-30 | 2022-09-13 | 中国工商银行股份有限公司 | Household resource interaction method and device, and computer equipment |
| WO2023283834A1 (en) * | 2021-07-14 | 2023-01-19 | 北京清雷科技有限公司 | Information detection method and apparatus for indoor object, and storage medium and processor |
| CN116030184A (en) * | 2022-11-29 | 2023-04-28 | 江苏艾佳家居用品有限公司 | House type three-dimensional reconstruction method and system based on man-machine interaction, multi-mode perception and house type understanding |
| CN116578915A (en) * | 2023-04-11 | 2023-08-11 | 广州极点三维信息科技有限公司 | Structured house type analysis method and system based on graphic neural network |
-
2023
- 2023-09-26 CN CN202311261743.XA patent/CN117271822B/en active Active
Patent Citations (4)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| WO2023283834A1 (en) * | 2021-07-14 | 2023-01-19 | 北京清雷科技有限公司 | Information detection method and apparatus for indoor object, and storage medium and processor |
| CN115049437A (en) * | 2022-06-30 | 2022-09-13 | 中国工商银行股份有限公司 | Household resource interaction method and device, and computer equipment |
| CN116030184A (en) * | 2022-11-29 | 2023-04-28 | 江苏艾佳家居用品有限公司 | House type three-dimensional reconstruction method and system based on man-machine interaction, multi-mode perception and house type understanding |
| CN116578915A (en) * | 2023-04-11 | 2023-08-11 | 广州极点三维信息科技有限公司 | Structured house type analysis method and system based on graphic neural network |
Non-Patent Citations (2)
| Title |
|---|
| BANDAM, A 等: "Classification of Building Types in Germany: A Data-Driven Modeling Approach", 《DATA》, vol. 7, no. 4, 9 April 2022 (2022-04-09), pages 1 - 45 * |
| 肖丁;王乾宇;蔡铭;李秀;: "智能家居场景联动中基于知识图谱的隐式冲突检测方法研究", 计算机学报, no. 06, 18 March 2019 (2019-03-18), pages 1190 - 1204 * |
Also Published As
| Publication number | Publication date |
|---|---|
| CN117271822B (en) | 2024-04-09 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN108399428B (en) | Triple loss function design method based on trace ratio criterion | |
| CN109635150B (en) | Text generation method, device and storage medium | |
| Limketkai et al. | Relational object maps for mobile robots | |
| US20160267359A1 (en) | Image object category recognition method and device | |
| CN111695527B (en) | Mongolian online handwriting recognition method | |
| Berretti et al. | 3d mesh decomposition using reeb graphs | |
| Zhang et al. | Semantic classification of heterogeneous urban scenes using intrascene feature similarity and interscene semantic dependency | |
| JP2006520948A (en) | Method, system and data structure for searching for 3D objects | |
| Pizarro et al. | Automatic floor plan analysis and recognition | |
| Chen et al. | A multi-channel deep neural network for relation extraction | |
| Liu et al. | 3D Point cloud analysis | |
| US20220189070A1 (en) | Machine-learning for 3d object detection | |
| CN115115745A (en) | Method and system for generating self-created digital art, storage medium and electronic device | |
| Hu et al. | Curve skeleton extraction from 3D point clouds through hybrid feature point shifting and clustering | |
| Roy et al. | Offline hand-drawn circuit component recognition using texture and shape-based features | |
| CN114528368A (en) | Spatial relationship extraction method based on pre-training language model and text feature fusion | |
| CN117271822B (en) | Layout searching method and system based on multi-modal house type knowledge graph | |
| Meng et al. | Merged region based image retrieval | |
| CN111144469B (en) | End-to-end multi-sequence text recognition method based on multi-dimensional associated time sequence classification neural network | |
| CN111950646A (en) | Hierarchical knowledge model construction method and target identification method for electromagnetic image | |
| CN107798286B (en) | Hyperspectral image evolution classification method based on labeled sample position | |
| Mercioni et al. | A study on Hierarchical Clustering and the Distance metrics for Identifying Architectural Styles | |
| Tung et al. | Stroke-order-free on-line Chinese character recognition by stroke adjustment of two-layer bipartite weighted matching | |
| CN115203408A (en) | Intelligent labeling method for multi-modal test data | |
| Boyko | Efficient interfaces for accurate annotation of 3D point clouds |
Legal Events
| Date | Code | Title | Description |
|---|---|---|---|
| PB01 | Publication | ||
| PB01 | Publication | ||
| SE01 | Entry into force of request for substantive examination | ||
| SE01 | Entry into force of request for substantive examination | ||
| GR01 | Patent grant | ||
| GR01 | Patent grant |