CN113269197A - Certificate image vertex coordinate regression system and identification method based on semantic segmentation - Google Patents
Certificate image vertex coordinate regression system and identification method based on semantic segmentation Download PDFInfo
- Publication number
- CN113269197A CN113269197A CN202110451208.5A CN202110451208A CN113269197A CN 113269197 A CN113269197 A CN 113269197A CN 202110451208 A CN202110451208 A CN 202110451208A CN 113269197 A CN113269197 A CN 113269197A
- Authority
- CN
- China
- Prior art keywords
- module
- semantic segmentation
- vertex
- regression
- coordinate regression
- 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
Images
Classifications
-
- G—PHYSICS
- G06—COMPUTING OR CALCULATING; COUNTING
- G06V—IMAGE OR VIDEO RECOGNITION OR UNDERSTANDING
- G06V10/00—Arrangements for image or video recognition or understanding
- G06V10/20—Image preprocessing
- G06V10/28—Quantising the image, e.g. histogram thresholding for discrimination between background and foreground patterns
-
- G—PHYSICS
- G06—COMPUTING OR CALCULATING; COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F18/00—Pattern recognition
- G06F18/20—Analysing
- G06F18/24—Classification techniques
- G06F18/241—Classification techniques relating to the classification model, e.g. parametric or non-parametric approaches
-
- G—PHYSICS
- G06—COMPUTING OR CALCULATING; COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F18/00—Pattern recognition
- G06F18/20—Analysing
- G06F18/24—Classification techniques
- G06F18/241—Classification techniques relating to the classification model, e.g. parametric or non-parametric approaches
- G06F18/2415—Classification techniques relating to the classification model, e.g. parametric or non-parametric approaches based on parametric or probabilistic models, e.g. based on likelihood ratio or false acceptance rate versus a false rejection rate
-
- G—PHYSICS
- G06—COMPUTING OR CALCULATING; 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
-
- G—PHYSICS
- G06—COMPUTING OR CALCULATING; 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
- G06N3/047—Probabilistic or stochastic networks
-
- G—PHYSICS
- G06—COMPUTING OR CALCULATING; 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
- G06N3/048—Activation functions
-
- G—PHYSICS
- G06—COMPUTING OR CALCULATING; COUNTING
- G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
- G06N3/00—Computing arrangements based on biological models
- G06N3/02—Neural networks
- G06N3/08—Learning methods
-
- G—PHYSICS
- G06—COMPUTING OR CALCULATING; COUNTING
- G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
- G06T7/00—Image analysis
- G06T7/10—Segmentation; Edge detection
- G06T7/11—Region-based segmentation
-
- G—PHYSICS
- G06—COMPUTING OR CALCULATING; COUNTING
- G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
- G06T7/00—Image analysis
- G06T7/70—Determining position or orientation of objects or cameras
- G06T7/73—Determining position or orientation of objects or cameras using feature-based methods
-
- G—PHYSICS
- G06—COMPUTING OR CALCULATING; 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 OR CALCULATING; COUNTING
- G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
- G06T2207/00—Indexing scheme for image analysis or image enhancement
- G06T2207/10—Image acquisition modality
- G06T2207/10004—Still image; Photographic image
-
- G—PHYSICS
- G06—COMPUTING OR CALCULATING; COUNTING
- G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
- G06T2207/00—Indexing scheme for image analysis or image enhancement
- G06T2207/20—Special algorithmic details
- G06T2207/20081—Training; Learning
-
- G—PHYSICS
- G06—COMPUTING OR CALCULATING; COUNTING
- G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
- G06T2207/00—Indexing scheme for image analysis or image enhancement
- G06T2207/20—Special algorithmic details
- G06T2207/20084—Artificial neural networks [ANN]
-
- Y—GENERAL TAGGING OF NEW TECHNOLOGICAL DEVELOPMENTS; GENERAL TAGGING OF CROSS-SECTIONAL TECHNOLOGIES SPANNING OVER SEVERAL SECTIONS OF THE IPC; TECHNICAL SUBJECTS COVERED BY FORMER USPC CROSS-REFERENCE ART COLLECTIONS [XRACs] AND DIGESTS
- Y02—TECHNOLOGIES OR APPLICATIONS FOR MITIGATION OR ADAPTATION AGAINST CLIMATE CHANGE
- Y02D—CLIMATE CHANGE MITIGATION TECHNOLOGIES IN INFORMATION AND COMMUNICATION TECHNOLOGIES [ICT], I.E. INFORMATION AND COMMUNICATION TECHNOLOGIES AIMING AT THE REDUCTION OF THEIR OWN ENERGY USE
- Y02D10/00—Energy efficient computing, e.g. low power processors, power management or thermal management
Landscapes
- Engineering & Computer Science (AREA)
- Theoretical Computer Science (AREA)
- Physics & Mathematics (AREA)
- General Physics & Mathematics (AREA)
- Data Mining & Analysis (AREA)
- General Engineering & Computer Science (AREA)
- Life Sciences & Earth Sciences (AREA)
- Artificial Intelligence (AREA)
- Evolutionary Computation (AREA)
- Computer Vision & Pattern Recognition (AREA)
- Health & Medical Sciences (AREA)
- Software Systems (AREA)
- Biophysics (AREA)
- Biomedical Technology (AREA)
- General Health & Medical Sciences (AREA)
- Molecular Biology (AREA)
- Computing Systems (AREA)
- Computational Linguistics (AREA)
- Mathematical Physics (AREA)
- Multimedia (AREA)
- Probability & Statistics with Applications (AREA)
- Bioinformatics & Cheminformatics (AREA)
- Bioinformatics & Computational Biology (AREA)
- Evolutionary Biology (AREA)
- Image Analysis (AREA)
Abstract
The invention discloses a certificate image vertex coordinate regression system based on semantic segmentation, which comprises a coordinate regression module, wherein the coordinate regression module is arranged behind the semantic segmentation module and mainly uses a plurality of semantic segmentation networks as backbone networks by using the coordinate regression module to directly obtain the polygon vertex coordinates of a certificate area, and the coordinate regression module finally uses the maximum index in a thermodynamic diagram to obtain the coordinates of each vertex.
Description
The technical field is as follows:
the invention belongs to the technical field of image target detection, and particularly relates to a certificate image vertex coordinate regression system and a certificate image vertex coordinate identification method based on semantic segmentation.
Background art:
the certificate image semantic segmentation is mainly characterized in that manual labeling is carried out by collecting certain certificate pictures (including but not limited to motor vehicle registration certificates, driving certificates and contract equivalents), sub-regions of different types are divided in the pictures, and a deep neural network semantic segmentation model is trained by using the pictures and labeled data. When the method is applied, a new picture is input into the model, and the segmentation result of the target sub-region is automatically obtained. The size of the segmentation result is the same as that of the input picture, wherein the value of each pixel point is indexed by the type (such as certificate or background) of the pixel point, and the vertex coordinates of the sub-region are calculated by an image post-processing algorithm (such as OpenCV (open computer vision library) to calculate the minimum circumscribed polygon, and the structure is shown in FIG. 1. The defects of the conventional technology are specifically: (1) when a general semantic segmentation model is used for segmentation of a certificate area, post-processing operations such as noise removal, minimum circumscribed rectangle solving and the like are needed to obtain the polygon vertex coordinates of the certificate, and the post-processing operations cannot be accelerated by a GPU together with a neural network and are easily interfered by a segmentation result; (2) when the number and the quality of the training pictures are insufficient, a model with strong generalization capability cannot be obtained, and the anti-blocking capability is limited; (3) using multi-class masks as a single label makes it difficult for a simple model to learn accurate edge and vertex information.
The information disclosed in this background section is only for enhancement of understanding of the general background of the invention and should not be taken as an acknowledgement or any form of suggestion that this information forms the prior art already known to a person skilled in the art.
The invention content is as follows:
the invention aims to provide a certificate image vertex coordinate regression system based on semantic segmentation, thereby overcoming the defects in the prior art.
In order to achieve the purpose, the invention provides a certificate image vertex coordinate regression system based on semantic segmentation, which comprises a coordinate regression module, wherein the coordinate regression module is arranged behind the semantic segmentation module.
Preferably, in the above technical solution, the semantic segmentation module performs semantic segmentation according to the backbone network output structure by using a semantic segmentation model;
and the coordinate regression module is used for directly obtaining all vertex coordinates of the certificate area through network calculation.
Preferably, in the above technical solution, the semantic segmentation module adopts deep lab, UNet, PSPNet or other semantic segmentation models.
Preferably, in the above technical solution, the coordinate regression module includes a feature map convolution module, a mask attention module, an attention fusion module, and a thermodynamic diagram calculation module;
the feature map convolution module is communicated with a network layer used for calculating a final segmentation result in the semantic segmentation module, comprises a plurality of 3x3 convolution layers, ReLu activation layers and Batch Normalization layers, and obtains an output result which has the same width and height as the input feature map and has the number of channels as the number of regression vertex coordinates, and sends the output result to the attention fusion module;
the mask attention module is communicated with a semantic segmentation result output end in the semantic segmentation module, global pooling is carried out after a plurality of 3x3 convolutions, ReLu activation and Batch Normalization layers, a middle feature layer with the width and height of 1 and the number of channels as the number of vertex-return coordinates is obtained, and an output result of the module is obtained after passing through two full connection layers and is transmitted to the attention fusion module;
the attention fusion module performs pixel-by-pixel addition on the output characteristic graphs of the characteristic graph convolution module and the mask attention module, and then obtains an output result through a sigmoid activation layer and outputs the output result to the thermodynamic diagram calculation module;
the input end of the thermodynamic diagram calculation module is in communication connection with the output end of the attention fusion module, the width and the height of the thermodynamic diagram calculation module are the same as those of the semantic segmentation network feature diagram, the number of channels is the number of regression vertex coordinates, and each channel of the feature diagram represents a thermodynamic diagram of each vertex; and (3) directly performing loss calculation on the input and the label during training, and calculating the maximum value index of each channel through an argmax function during reasoning, namely each predicted vertex coordinate.
A certificate image vertex coordinate regression recognition method based on semantic segmentation uses a coordinate regression module to take various semantic segmentation networks as backbone networks to directly obtain polygon vertex coordinates of a certificate area, and the coordinate regression module finally uses a maximum value index in a thermodynamic diagram to obtain the coordinates of each vertex.
Preferably, in the above technical solution, the semantic segmentation network features and the semantic segmentation results are given by a general semantic segmentation model, and the coordinate regression module is disposed behind the semantic segmentation model.
Preferably, in the above technical solution, the feature graph convolution module semantically segments network features to obtain an output to the next process, where the width and height are the same as those of the input feature graph, and the number of channels is the number of regression vertex coordinates;
the mask attention module performs global pooling on semantic segmentation results after a plurality of 3x3 convolutions, ReLu activation and Batch Normalization layers to obtain a middle feature layer with the width and height of 1 and the number of channels as the number of vertex coordinates, and outputs the semantic segmentation results to the next process after passing through two full connection layers;
the attention fusion module performs pixel-by-pixel addition on the output characteristic graphs of the characteristic graph convolution module and the mask attention module, and then obtains an output result through a sigmoid activation layer to the next process;
the thermodynamic diagram calculation module obtains an output result of the attention fusion module, the width and the height of the output result are the same as those of a semantic segmentation network feature diagram, the number of channels is the number of regression vertex coordinates, and each channel of the feature diagram represents a thermodynamic diagram of each vertex; directly performing loss calculation on the input and the label during training, and calculating the maximum index of each channel through an argmax function during reasoning, namely predicting each vertex coordinate;
the labels of the coordinate regression module are vertex thermodynamic diagrams, and can be automatically generated from the vertexes of the minimum circumscribed polygon according to the existing semantic segmentation label diagram before training; the Smooth L1 is used as a loss function to calculate the loss of the label and network output thermodynamic diagrams during training. The coordinate regression module can be used as an improvement module to carry out independent training, and can also be used for carrying out weighted calculation together with a loss function of the original semantic segmentation model. The coordinate thermodynamic diagram has content independence and can provide anti-occlusion capability in an area.
Compared with the prior art, the invention has the following beneficial effects:
(1) the coordinate regression module has strong adaptability, and can be added after most of semantic segmentation models and modified according to the output structure of the backbone network;
(2) under the condition of existing vertex coordinates or masks, data do not need to be marked again, and supervision training is directly carried out after conversion;
(3) the coordinate regression module directly obtains all vertex coordinates of the certificate area through network calculation, is not limited by fixed shapes such as rectangles and the like, and can resist deformation interference of various images;
(4) the coordinate points have content independence, and can provide anti-blocking capability in the certificate image when being finely adjusted together with the backbone network, and correct errors of the segmentation network.
(5) The external rectangle is further calculated without a post-processing method, calculation can be accelerated through the GPU, the process complexity and tool dependence during application deployment are reduced, and the calculation speed is remarkably improved.
Description of the drawings:
FIG. 1 is a prior art schematic;
FIG. 2 is a schematic flow chart of the present invention;
FIG. 3 is a schematic diagram of a network architecture according to the present invention;
FIG. 4 is a schematic diagram of the regression method of the present invention.
The specific implementation mode is as follows:
the following detailed description of specific embodiments of the invention is provided, but it should be understood that the scope of the invention is not limited to the specific embodiments.
Throughout the specification and claims, unless explicitly stated otherwise, the word "comprise", or variations such as "comprises" or "comprising", will be understood to imply the inclusion of a stated element or component but not the exclusion of any other element or component.
The invention provides a certificate image vertex coordinate regression method based on semantic segmentation, which is characterized in that a semantic segmentation database and an existing semantic segmentation model aiming at certificate images such as motor vehicle registration certificates, driving certificates and the like are constructed by means of data generated by users in app in the automobile industry of a vehicle 300, and the improved semantic segmentation method is adopted to obtain the vertex coordinates of the certificates in the images. The patent adopts a certificate image vertex coordinate regression method based on semantic segmentation, and specifically comprises the following steps:
1. an innovative coordinate regression module is used, various semantic segmentation networks are used as backbone networks, the polygon vertex coordinates of a certificate area are directly obtained, and post-processing operations such as external polygon solving are replaced. The coordinate regression module finally uses the maximum index in the thermodynamic diagram to obtain the coordinate of each vertex, an image post-processing algorithm is not needed to calculate the circumscribed rectangle, dependence on an image processing library is reduced during deployment, and high speed and high efficiency are achieved.
2. The overall network structure is shown in fig. 3 and fig. 4, wherein the semantic segmentation network features and the semantic segmentation results are given by a general semantic segmentation model, such as deep lab, UNet, PSPNet, and the like. The coordinate regression module comprises a feature map convolution module, a mask attention module, an attention fusion module, a thermodynamic diagram calculation module and the like and is arranged behind the semantic segmentation module.
3. The input of the feature map convolution module is a semantic segmentation network feature, namely a network layer used for calculating a final segmentation result in a semantic segmentation model, and the feature map convolution module comprises a plurality of 3x3 convolution layers, ReLu activation layers and Batch Normalization layers and obtains output with the same width and height as the input feature map and the channel number being the regression vertex coordinate number.
4. The input of the mask attention module is a semantic segmentation result, global pooling is carried out after a plurality of 3x3 convolution layers, ReLu activation layers and Batch Normalization layers, a middle feature layer with the width and height of 1 and the number of channels as the number of vertex-homing coordinates is obtained, and the output of the module is obtained after the middle feature layer passes through two full-connection layers.
5. And the attention fusion module performs pixel-by-pixel addition on the output characteristic graphs of the characteristic graph convolution module and the mask attention module, and outputs the result through a sigmoid activation layer.
6. The input of the thermodynamic diagram calculation module is the output of the attention fusion module, the width and the height of the thermodynamic diagram are the same as those of the semantic segmentation network feature diagram, the number of channels is the number of regression vertex coordinates, and each channel of the feature diagram represents a thermodynamic diagram of each vertex. And (3) directly performing loss calculation on the input and the label during training, and calculating the maximum value index of each channel through an argmax function during reasoning, namely each predicted vertex coordinate.
7. The labels of the coordinate regression module are vertex thermodynamic diagrams, and can be automatically generated from the vertexes of the minimum circumscribed polygon according to the existing semantic segmentation label diagram before training. The Smooth L1 is used as a loss function to calculate the loss of the label and network output thermodynamic diagrams during training. The coordinate regression module can be used as an improvement module to carry out independent training, and can also be used for carrying out weighted calculation together with a loss function of the original semantic segmentation model. The coordinate thermodynamic diagram has content independence and can provide anti-occlusion capability in an area.
8. To verify the effect of the coordinate regression module, a comparative test was performed using 198 motor vehicle registration certificate pictures containing 4 semantic segmentation classes, with the following results:
the foregoing descriptions of specific exemplary embodiments of the present invention have been presented for purposes of illustration and description. It is not intended to limit the invention to the precise form disclosed, and obviously many modifications and variations are possible in light of the above teaching. The exemplary embodiments were chosen and described in order to explain certain principles of the invention and its practical application to enable one skilled in the art to make and use various exemplary embodiments of the invention and various alternatives and modifications as are suited to the particular use contemplated. It is intended that the scope of the invention be defined by the claims and their equivalents.
Claims (7)
1. A certificate image vertex coordinate regression system based on semantic segmentation is characterized in that: the system comprises a coordinate regression module which is arranged behind a semantic segmentation module.
2. The semantic segmentation based certificate image vertex coordinate regression system of claim 1, wherein:
the semantic segmentation module is used for performing semantic segmentation according to the backbone network output structure by adopting a semantic segmentation model;
and the coordinate regression module is used for directly obtaining all vertex coordinates of the certificate area through network calculation.
3. The semantic segmentation based certificate image vertex coordinate regression system of claim 1 or 2, characterized in that: the semantic segmentation module adopts deep Lab, UNet, PSPNet or other semantic segmentation models.
4. The semantic segmentation based certificate image vertex coordinate regression system of claim 1 or 2, characterized in that: the coordinate regression module comprises a feature map convolution module, a mask attention module, an attention fusion module and a thermodynamic diagram calculation module;
the feature map convolution module is communicated with a network layer used for calculating a final segmentation result in the semantic segmentation module, comprises a plurality of 3x3 convolution layers, ReLu activation layers and Batch Normalization layers, and obtains an output result which has the same width and height as the input feature map and has the number of channels as the number of regression vertex coordinates, and sends the output result to the attention fusion module;
the mask attention module is communicated with a semantic segmentation result output end in the semantic segmentation module, global pooling is carried out after a plurality of 3x3 convolutions, ReLu activation and Batch Normalization layers, a middle feature layer with the width and height of 1 and the number of channels as the number of vertex-return coordinates is obtained, and an output result of the module is obtained after passing through two full connection layers and is transmitted to the attention fusion module;
the attention fusion module performs pixel-by-pixel addition on the output characteristic graphs of the characteristic graph convolution module and the mask attention module, and then obtains an output result through a sigmoid activation layer and outputs the output result to the thermodynamic diagram calculation module;
the input end of the thermodynamic diagram calculation module is in communication connection with the output end of the attention fusion module, the width and the height of the thermodynamic diagram calculation module are the same as those of the semantic segmentation network feature diagram, the number of channels is the number of regression vertex coordinates, and each channel of the feature diagram represents a thermodynamic diagram of each vertex; and (3) directly performing loss calculation on the input and the label during training, and calculating the maximum value index of each channel through an argmax function during reasoning, namely each predicted vertex coordinate.
5. A certificate image vertex coordinate regression identification method based on semantic segmentation is characterized by comprising the following steps: and the coordinate regression module is used for taking various semantic segmentation networks as backbone networks to directly obtain the polygon vertex coordinates of the certificate area, and finally, the coordinate regression module is used for solving the coordinates of each vertex by using the maximum value index in the thermodynamic diagram.
6. The regression recognition method for certificate image vertex coordinates based on semantic segmentation as claimed in claim 1, wherein the semantic segmentation network features and the semantic segmentation result are given by a common semantic segmentation model, and the coordinate regression module is arranged behind the semantic segmentation model.
7. The certificate image vertex coordinate regression recognition method based on semantic segmentation as claimed in claim 1, wherein the feature graph convolution module semantically segments network features to obtain output to the next process, wherein the width and the height of the input feature graph are the same, and the number of channels is the number of regression vertex coordinates;
the mask attention module performs global pooling on semantic segmentation results after a plurality of 3x3 convolutions, ReLu activation and Batch Normalization layers to obtain a middle feature layer with the width and height of 1 and the number of channels as the number of vertex coordinates, and outputs the semantic segmentation results to the next process after passing through two full connection layers;
the attention fusion module performs pixel-by-pixel addition on the output characteristic graphs of the characteristic graph convolution module and the mask attention module, and then obtains an output result through a sigmoid activation layer to the next process;
the thermodynamic diagram calculation module obtains an output result of the attention fusion module, the width and the height of the output result are the same as those of a semantic segmentation network feature diagram, the number of channels is the number of regression vertex coordinates, and each channel of the feature diagram represents a thermodynamic diagram of each vertex; directly performing loss calculation on the input and the label during training, and calculating the maximum index of each channel through an argmax function during reasoning, namely predicting each vertex coordinate;
the labels of the coordinate regression module are vertex thermodynamic diagrams, and can be automatically generated from the vertexes of the minimum circumscribed polygon according to the existing semantic segmentation label diagram before training; the Smooth L1 is used as a loss function to calculate the loss of the label and network output thermodynamic diagrams during training.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202110451208.5A CN113269197B (en) | 2021-04-25 | 2021-04-25 | Certificate image vertex coordinate regression system and identification method based on semantic segmentation |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202110451208.5A CN113269197B (en) | 2021-04-25 | 2021-04-25 | Certificate image vertex coordinate regression system and identification method based on semantic segmentation |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN113269197A true CN113269197A (en) | 2021-08-17 |
| CN113269197B CN113269197B (en) | 2024-03-08 |
Family
ID=77229460
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN202110451208.5A Active CN113269197B (en) | 2021-04-25 | 2021-04-25 | Certificate image vertex coordinate regression system and identification method based on semantic segmentation |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN113269197B (en) |
Citations (28)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN109711413A (en) * | 2018-12-30 | 2019-05-03 | 陕西师范大学 | Image Semantic Segmentation Method Based on Deep Learning |
| CN109886286A (en) * | 2019-01-03 | 2019-06-14 | 武汉精测电子集团股份有限公司 | Target detection method, target detection model and system based on cascade detectors |
| CN110008915A (en) * | 2019-04-11 | 2019-07-12 | 电子科技大学 | System and method for dense human pose estimation based on mask-RCNN |
| CN110163193A (en) * | 2019-03-25 | 2019-08-23 | 腾讯科技（深圳）有限公司 | Image processing method, device, computer readable storage medium and computer equipment |
| CN110223298A (en) * | 2019-05-27 | 2019-09-10 | 东南大学 | Semantic segmentation innovatory algorithm based on point cloud local structure |
| WO2019192397A1 (en) * | 2018-04-04 | 2019-10-10 | 华中科技大学 | End-to-end recognition method for scene text in any shape |
| CN110781850A (en) * | 2019-10-31 | 2020-02-11 | 深圳金信诺高新技术股份有限公司 | Semantic segmentation system and method for road recognition, and computer storage medium |
| CN110837836A (en) * | 2019-11-05 | 2020-02-25 | 中国科学技术大学 | Semi-supervised semantic segmentation method based on maximized confidence |
| CN110866930A (en) * | 2019-11-18 | 2020-03-06 | 北京云聚智慧科技有限公司 | Semantic segmentation auxiliary labeling method and device |
| CN111462126A (en) * | 2020-04-08 | 2020-07-28 | 武汉大学 | Semantic image segmentation method and system based on edge enhancement |
| CN111476067A (en) * | 2019-01-23 | 2020-07-31 | 腾讯科技（深圳）有限公司 | Character recognition method and device for image, electronic equipment and readable storage medium |
| AU2020101229A4 (en) * | 2020-07-02 | 2020-08-06 | South China University Of Technology | A Text Line Recognition Method in Chinese Scenes Based on Residual Convolutional and Recurrent Neural Networks |
| CN111738055A (en) * | 2020-04-24 | 2020-10-02 | 浙江大学城市学院 | Multi-category text detection system and bill form detection method based on the system |
| CN111815570A (en) * | 2020-06-16 | 2020-10-23 | 浙江大华技术股份有限公司 | Regional intrusion detection method and related device thereof |
| CN111902825A (en) * | 2018-03-23 | 2020-11-06 | 多伦多大学管理委员会 | Polygonal object labeling system and method for training object labeling system |
| CN111912531A (en) * | 2020-07-31 | 2020-11-10 | 郑州奥腾网络科技有限公司 | Dynamic human body trunk temperature analysis method based on intelligent thermal imaging video analysis |
| CN111985466A (en) * | 2020-08-19 | 2020-11-24 | 上海海事大学 | A kind of container dangerous goods sign identification method |
| CN112149496A (en) * | 2020-08-10 | 2020-12-29 | 浙江科技学院 | Real-time road scene segmentation method based on convolutional neural network |
| WO2021000702A1 (en) * | 2019-06-29 | 2021-01-07 | 华为技术有限公司 | Image detection method, device, and system |
| CN112330589A (en) * | 2020-09-18 | 2021-02-05 | 北京沃东天骏信息技术有限公司 | Method and device for estimating pose and computer readable storage medium |
| JP2021022367A (en) * | 2019-07-29 | 2021-02-18 | 富士通株式会社 | Image processing method and information processor |
| CN112378591A (en) * | 2020-11-02 | 2021-02-19 | 林国桥 | Air tightness detection laser pose self-adaptive adjusting method based on computer vision |
| CN112508956A (en) * | 2020-11-05 | 2021-03-16 | 浙江科技学院 | Road scene semantic segmentation method based on convolutional neural network |
| WO2021056705A1 (en) * | 2019-09-23 | 2021-04-01 | 平安科技（深圳）有限公司 | Method for detecting damage to outside of human body on basis of semantic segmentation network, and related device |
| CN112598004A (en) * | 2020-12-21 | 2021-04-02 | 安徽七天教育科技有限公司 | English composition test paper layout analysis method based on scanning |
| CN112651316A (en) * | 2020-12-18 | 2021-04-13 | 上海交通大学 | Two-dimensional and three-dimensional multi-person attitude estimation system and method |
| CN112668522A (en) * | 2020-12-31 | 2021-04-16 | 华南理工大学 | Human body key point and human body mask combined detection network and method |
| CN112700426A (en) * | 2021-01-07 | 2021-04-23 | 浙江科技学院 | Method for detecting salient object in complex environment |
-
2021
- 2021-04-25 CN CN202110451208.5A patent/CN113269197B/en active Active
Patent Citations (28)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN111902825A (en) * | 2018-03-23 | 2020-11-06 | 多伦多大学管理委员会 | Polygonal object labeling system and method for training object labeling system |
| WO2019192397A1 (en) * | 2018-04-04 | 2019-10-10 | 华中科技大学 | End-to-end recognition method for scene text in any shape |
| CN109711413A (en) * | 2018-12-30 | 2019-05-03 | 陕西师范大学 | Image Semantic Segmentation Method Based on Deep Learning |
| CN109886286A (en) * | 2019-01-03 | 2019-06-14 | 武汉精测电子集团股份有限公司 | Target detection method, target detection model and system based on cascade detectors |
| CN111476067A (en) * | 2019-01-23 | 2020-07-31 | 腾讯科技（深圳）有限公司 | Character recognition method and device for image, electronic equipment and readable storage medium |
| CN110163193A (en) * | 2019-03-25 | 2019-08-23 | 腾讯科技（深圳）有限公司 | Image processing method, device, computer readable storage medium and computer equipment |
| CN110008915A (en) * | 2019-04-11 | 2019-07-12 | 电子科技大学 | System and method for dense human pose estimation based on mask-RCNN |
| CN110223298A (en) * | 2019-05-27 | 2019-09-10 | 东南大学 | Semantic segmentation innovatory algorithm based on point cloud local structure |
| WO2021000702A1 (en) * | 2019-06-29 | 2021-01-07 | 华为技术有限公司 | Image detection method, device, and system |
| JP2021022367A (en) * | 2019-07-29 | 2021-02-18 | 富士通株式会社 | Image processing method and information processor |
| WO2021056705A1 (en) * | 2019-09-23 | 2021-04-01 | 平安科技（深圳）有限公司 | Method for detecting damage to outside of human body on basis of semantic segmentation network, and related device |
| CN110781850A (en) * | 2019-10-31 | 2020-02-11 | 深圳金信诺高新技术股份有限公司 | Semantic segmentation system and method for road recognition, and computer storage medium |
| CN110837836A (en) * | 2019-11-05 | 2020-02-25 | 中国科学技术大学 | Semi-supervised semantic segmentation method based on maximized confidence |
| CN110866930A (en) * | 2019-11-18 | 2020-03-06 | 北京云聚智慧科技有限公司 | Semantic segmentation auxiliary labeling method and device |
| CN111462126A (en) * | 2020-04-08 | 2020-07-28 | 武汉大学 | Semantic image segmentation method and system based on edge enhancement |
| CN111738055A (en) * | 2020-04-24 | 2020-10-02 | 浙江大学城市学院 | Multi-category text detection system and bill form detection method based on the system |
| CN111815570A (en) * | 2020-06-16 | 2020-10-23 | 浙江大华技术股份有限公司 | Regional intrusion detection method and related device thereof |
| AU2020101229A4 (en) * | 2020-07-02 | 2020-08-06 | South China University Of Technology | A Text Line Recognition Method in Chinese Scenes Based on Residual Convolutional and Recurrent Neural Networks |
| CN111912531A (en) * | 2020-07-31 | 2020-11-10 | 郑州奥腾网络科技有限公司 | Dynamic human body trunk temperature analysis method based on intelligent thermal imaging video analysis |
| CN112149496A (en) * | 2020-08-10 | 2020-12-29 | 浙江科技学院 | Real-time road scene segmentation method based on convolutional neural network |
| CN111985466A (en) * | 2020-08-19 | 2020-11-24 | 上海海事大学 | A kind of container dangerous goods sign identification method |
| CN112330589A (en) * | 2020-09-18 | 2021-02-05 | 北京沃东天骏信息技术有限公司 | Method and device for estimating pose and computer readable storage medium |
| CN112378591A (en) * | 2020-11-02 | 2021-02-19 | 林国桥 | Air tightness detection laser pose self-adaptive adjusting method based on computer vision |
| CN112508956A (en) * | 2020-11-05 | 2021-03-16 | 浙江科技学院 | Road scene semantic segmentation method based on convolutional neural network |
| CN112651316A (en) * | 2020-12-18 | 2021-04-13 | 上海交通大学 | Two-dimensional and three-dimensional multi-person attitude estimation system and method |
| CN112598004A (en) * | 2020-12-21 | 2021-04-02 | 安徽七天教育科技有限公司 | English composition test paper layout analysis method based on scanning |
| CN112668522A (en) * | 2020-12-31 | 2021-04-16 | 华南理工大学 | Human body key point and human body mask combined detection network and method |
| CN112700426A (en) * | 2021-01-07 | 2021-04-23 | 浙江科技学院 | Method for detecting salient object in complex environment |
Non-Patent Citations (2)
| Title |
|---|
| CALIMANUT-IONUT CIRA，等: "A Deep Learning-Based Solution for Large-Scale Extraction of the Secondary Road Network from High-Resolution Aerial Orthoimagery", 《APPLIED SCIENCES-BASEL》, vol. 10, no. 20, pages 7272 * |
| 方鹏飞: "基于卷积神经网络框架的室内物体目标检测和姿态估计", 《中国优秀硕士学位论文全文数据库 信息科技》, no. 2, pages 138 - 1637 * |
Also Published As
| Publication number | Publication date |
|---|---|
| CN113269197B (en) | 2024-03-08 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN110427937B (en) | Inclined license plate correction and indefinite-length license plate identification method based on deep learning | |
| CN109886121B (en) | Human face key point positioning method for shielding robustness | |
| JP6853560B2 (en) | A method for auto-labeling a training image to be used for learning a deep learning network that analyzes a high-precision image, and an auto-labeling device using this {METHOD FOR AUTO-LABELING TRAINING IMAGES FOR USE IN DEEP LEARNING NETWORK TOAL IMAGES WITH HIGH PRECISION, AND AUTO-LABELING DEVICE USING THE SAMEM} | |
| CN112116599B (en) | Method and system for semantic segmentation of sputum smear Mycobacterium tuberculosis based on weakly supervised learning | |
| CN108334881B (en) | License plate recognition method based on deep learning | |
| CN107301383A (en) | A kind of pavement marking recognition methods based on Fast R CNN | |
| CN111598030A (en) | Method and system for detecting and segmenting vehicle in aerial image | |
| CN111160352A (en) | Workpiece metal surface character recognition method and system based on image segmentation | |
| CN104517103A (en) | Traffic sign classification method based on deep neural network | |
| CN112446356B (en) | Method for detecting text with arbitrary shape in natural scene based on multiple polar coordinates | |
| CN110659601B (en) | Dense vehicle detection method for remote sensing images based on deep fully convolutional network based on central points | |
| CN111191611B (en) | Traffic sign identification method based on deep learning | |
| CN107808376B (en) | A Deep Learning-Based Hand Raised Detection Method | |
| CN112633149B (en) | Domain-adaptive foggy-day image target detection method and device | |
| CN110781744A (en) | A small-scale pedestrian detection method based on multi-level feature fusion | |
| CN109858327B (en) | Character segmentation method based on deep learning | |
| CN109657538B (en) | Scene segmentation method and system based on context information guidance | |
| CN107516316A (en) | A Method for Segmenting Static Human Image by Introducing Focusing Mechanism in FCN | |
| CN112085728A (en) | Submarine pipeline and leakage point detection method | |
| CN112464731A (en) | Traffic sign detection and identification method based on image processing | |
| CN110956119B (en) | Method for detecting target in image | |
| CN114913498A (en) | Parallel multi-scale feature aggregation lane line detection method based on key point estimation | |
| CN111161213A (en) | A Knowledge Graph-based Image Classification Method for Industrial Product Defects | |
| CN114445620A (en) | An Object Segmentation Method to Improve Mask R-CNN | |
| CN118314563A (en) | Text detection method based on rotary frame |
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 |