CN111382787A - Target detection method based on deep learning - Google Patents
Target detection method based on deep learning Download PDFInfo
- Publication number
- CN111382787A CN111382787A CN202010150722.0A CN202010150722A CN111382787A CN 111382787 A CN111382787 A CN 111382787A CN 202010150722 A CN202010150722 A CN 202010150722A CN 111382787 A CN111382787 A CN 111382787A
- Authority
- CN
- China
- Prior art keywords
- model
- algorithm
- loss function
- data
- deep learning
- 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.)
- Pending
Links
Images
Classifications
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F18/00—Pattern recognition
- G06F18/20—Analysing
- G06F18/24—Classification techniques
-
- 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
- G06F16/55—Clustering; Classification
-
- 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
-
- 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
- G06N3/08—Learning methods
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06V—IMAGE OR VIDEO RECOGNITION OR UNDERSTANDING
- G06V2201/00—Indexing scheme relating to image or video recognition or understanding
- G06V2201/07—Target detection
Landscapes
- Engineering & Computer Science (AREA)
- Theoretical Computer Science (AREA)
- Physics & Mathematics (AREA)
- Data Mining & Analysis (AREA)
- General Physics & Mathematics (AREA)
- General Engineering & Computer Science (AREA)
- Evolutionary Computation (AREA)
- Life Sciences & Earth Sciences (AREA)
- Artificial Intelligence (AREA)
- Computing Systems (AREA)
- Software Systems (AREA)
- Molecular Biology (AREA)
- Computational Linguistics (AREA)
- Biophysics (AREA)
- Biomedical Technology (AREA)
- Mathematical Physics (AREA)
- General Health & Medical Sciences (AREA)
- Health & Medical Sciences (AREA)
- Bioinformatics & Cheminformatics (AREA)
- Bioinformatics & Computational Biology (AREA)
- Computer Vision & Pattern Recognition (AREA)
- Evolutionary Biology (AREA)
- Databases & Information Systems (AREA)
- Image Analysis (AREA)
Abstract
The invention discloses a target detection method based on deep learning, which comprises the following steps: s100, collecting data; s200, selecting an algorithm, analyzing a target to be detected, and selecting a proper algorithm; s300, cleaning data, analyzing, screening and modifying the collected data to obtain model data conforming to an algorithm; s400, data conversion; s500, building a model, and building the model according to the selected algorithm; s600, training a model, inputting the converted data into the model for training; s700, testing a model, and predicting pictures of the trained generated format; and S800, repairing the model, and improving and maintaining the model according to the test result. And installing the trained model on a test server to be used as a model test. Then, the pictures of the internet are placed on the test server, the targets concerned by the users are detected through the test server, and then the targets are classified into each type, so that the targets in the pictures can be detected and identified more accurately and rapidly.
Description
Technical Field
The invention belongs to the technical field of computer vision processing, and particularly relates to a target detection method based on deep learning.
Background
With the continuous application and improvement of network technology, the internet has become an important channel for information distribution, plays an indispensable important role in information exchange, and a series of network applications such as enterprise office, e-commerce, government open, information construction of each unit and the like are developed rapidly. However, because the number of pictures on the internet is large, it takes a lot of time to see one by one, and the pictures are all picture information that does not need attention, and a lot of manpower and material resources are consumed, which results in that the efficiency of the business and government offices is high, and a lot of time is consumed.
Disclosure of Invention
1. Technical problem to be solved by the invention
The invention aims to solve the problem of low efficiency of the conventional picture target detection.
2. Technical scheme
In order to achieve the purpose, the technical scheme provided by the invention is as follows:
the invention relates to a target detection method based on deep learning, which comprises the following steps:
s100, collecting data, namely collecting pictures of targets needing to be detected;
s200, selecting an algorithm, analyzing a target to be detected, and selecting a proper algorithm;
s300, cleaning data, analyzing, screening and modifying the collected data to obtain model data conforming to an algorithm;
s400, data conversion, namely converting the model data after data cleaning into a format required by the model;
s500, building a model, and building the model according to the selected algorithm;
s600, training a model, inputting the converted data into the model for training;
s700, testing a model, and predicting pictures of the trained generated format;
and S800, repairing the model, and improving and maintaining the model according to the test result.
Preferably, the data collection in step S100 is to take a picture of an object to be detected or to search from the internet.
Preferably, the algorithm in step S200 is selected specifically by first calculating an algorithm with an accuracy of a final result of more than 80% by mathematical theory and mathematical extrapolation, and using the algorithm as an alternative algorithm to which the algorithm is theoretically suitable, then training each alternative algorithm, and using a model with high comparison accuracy and recall as an algorithm model.
Preferably, the data cleansing in step S300 is to perform size-unified adjustment on the pictures, or perform erosion and expansion on the pictures, or perform color transformation on the pictures according to the characteristics of the collected pictures.
Preferably, the model building in step S500 is to input a code after building a framework on a server, where the code is a code matching the algorithm selected in step S200.
Preferably, the model repairing in step S800 is performed by a deep learning algorithm, where the deep learning algorithm is performed by performing convolution through a convolutional neural network, extracting features, predicting the features to obtain an area a of a prediction box, and comparing the area a of the prediction box with an area B of a real box to calculate a coefficient J, and the larger the coefficient J, the higher the correlation.
Preferably, the model repair also requires the computation of a position loss function and a classification loss function,
the position loss function is
Wherein x is the label of the real sample, c is the label of the prediction sample, L is the offset of the upper left corner of the prediction frame, g is the offset of the lower right corner of the prediction frame, n is the total number of samples, L conf is the cross entropy, α is the weight coefficient, and L loc is the cross entropy;
the classification loss function is
Wherein: l loc is cross entropy, x is a predicted sample of a training sample, c is a real sample data, n is a total number of the training samples, pos is a few samples in the training samples, i is a sequence number in the training samples, p is a real sample number, Ne is a few samples in the training samples, o is a number in the real samples, exp is an index based on e, and ij is a default few real samples.
Preferably, the threshold of the coefficient J is H, when the coefficient J is smaller than H, the prediction frame is incorrect, the threshold H is set to calculate the correct number of the prediction frames by randomly setting the threshold for several times, and the set threshold in which the correct number of the prediction frames is the largest is taken as H.
Preferably, the model is correct if the position loss function and the classification loss function are smaller than a minimum value, the minimum value of the position loss function and the classification loss function is the minimum value of the position loss function and the classification loss function determined by using a gradient descent method, and the position loss function and the classification loss function are the minimum value when the position loss function and the classification loss function are not changed within 15 minutes.
3. Advantageous effects
Compared with the prior art, the technical scheme provided by the invention has the following beneficial effects:
the invention relates to a target detection method based on deep learning, which comprises the following steps: s100, collecting data, namely collecting pictures of targets needing to be detected; s200, selecting an algorithm, analyzing a target to be detected, and selecting a proper algorithm; s300, cleaning data, analyzing, screening and modifying the collected data to obtain model data conforming to an algorithm; s400, data conversion, namely converting the model data after data cleaning into a format required by the model; s500, building a model, and building the model according to the selected algorithm; s600, training a model, inputting the converted data into the model for training; s700, testing a model, and predicting pictures of the trained generated format; and S800, repairing the model, and improving and maintaining the model according to the test result. And installing the trained model on a test server to be used as a model test. Then, the picture of the internet is placed on a test server for testing, wherein the internet can be a picture of a website, can also be a picture shot in a smart phone, and can also be a picture propagated in social apps. When network data are transmitted, the targets concerned by the users can be detected through the test server, and then the targets are classified into each class, so that the targets in the pictures can be detected and identified more accurately and quickly.
Drawings
FIG. 1 is a flow chart of the method of the present invention.
Detailed Description
In order to facilitate an understanding of the invention, the invention will now be described more fully hereinafter with reference to the accompanying drawings, in which several embodiments of the invention are shown, but which may be embodied in many different forms and are not limited to the embodiments described herein, but rather are provided for the purpose of providing a more thorough disclosure of the invention.
Unless defined otherwise, all technical and scientific terms used herein have the same meaning as commonly understood by one of ordinary skill in the art to which this invention belongs; the terminology used herein in the description of the invention is for the purpose of describing particular embodiments only and is not intended to be limiting of the invention; as used herein, the term "and/or" includes any and all combinations of one or more of the associated listed items.
Example 1
Referring to fig. 1, a target detection method based on deep learning according to this embodiment includes the following steps:
s100, collecting data, namely collecting pictures of targets needing to be detected;
s200, selecting an algorithm, analyzing a target to be detected, and selecting a proper algorithm;
s300, cleaning data, analyzing, screening and modifying the collected data to obtain model data conforming to an algorithm;
s400, data conversion, namely converting the model data after data cleaning into a format required by the model;
s500, building a model, and building the model according to the selected algorithm;
s600, training a model, inputting the converted data into the model for training;
s700, testing a model, and predicting pictures of the trained generated format;
and S800, repairing the model, and improving and maintaining the model according to the test result.
And installing the trained model on a test server to be used as a model test. Then, the picture of the internet is placed on a test server for testing, wherein the internet can be a picture of a website, can also be a picture shot in a smart phone, and can also be a picture propagated in social apps. When network data are transmitted, the targets concerned by the users can be detected through the test server, and then the targets are classified into each class, so that the targets in the pictures can be detected and identified more accurately and quickly.
The data collection in step S100 of the present embodiment is specifically to take a picture of a target to be detected or to search from the internet. The algorithm in step S200 is specifically selected as an algorithm that is calculated by mathematical theory and mathematical calculation to obtain a final result with an accuracy of more than 80%, and the final result is used as an alternative algorithm of a theoretically suitable algorithm, and then each alternative algorithm is trained, and a model with high comparison accuracy and high recall is used as an algorithm model.
The data cleaning in step S300 of this embodiment is to perform size-unified adjustment on the picture, or perform erosion and expansion on the picture, or perform color transformation on the picture according to the characteristics of the collected picture. The model building in step S500 is specifically to input a code after building a framework on a server, where the code is a code matching the algorithm selected in step S200.
In this embodiment, the model repair in step S800 is performed by a deep learning algorithm, where the deep learning algorithm is performed by performing convolution through a convolutional neural network, extracting features, predicting the features to obtain an area a of a prediction box, and comparing the area a of the prediction box with an area B of a real box to calculate a coefficient J, where the larger the coefficient J, the higher the correlation.
And the threshold value of the coefficient J is H, when the coefficient J is smaller than H, the prediction frame is incorrect, the threshold value H is set to calculate the correct number of the prediction frames by randomly setting the threshold value for a plurality of times, and the set threshold value with the maximum correct number of the prediction frames is H.
The model repair also requires position loss function and classification loss function calculations,
the position loss function is
Wherein x is the label of the real sample, c is the label of the prediction sample, L is the offset of the upper left corner of the prediction frame, g is the offset of the lower right corner of the prediction frame, n is the total number of samples, L conf is the cross entropy, α is the weight coefficient, and L loc is the cross entropy;
the classification loss function is
Wherein: l loc is cross entropy, x is a predicted sample of a training sample, c is a real sample data, n is a total number of the training samples, pos is a few samples in the training samples, i is a sequence number in the training samples, p is a real sample number, Ne is a few samples in the training samples, o is a number in the real samples, exp is an index based on e, and ij is a default few real samples.
And if the position loss function and the classification loss function are smaller than the minimum value, the model is correct, the minimum value of the position loss function and the classification loss function is the minimum value of the position loss function and the classification loss function determined by using a gradient descent method, and the position loss function and the classification loss function are the minimum value when the position loss function and the classification loss function are not changed within 15 minutes.
The above-mentioned embodiments only express a certain implementation mode of the present invention, and the description thereof is specific and detailed, but not construed as limiting the scope of the present invention; it should be noted that, for those skilled in the art, without departing from the concept of the present invention, several variations and modifications can be made, which are within the protection scope of the present invention; therefore, the protection scope of the present patent shall be subject to the appended claims.
Claims (9)
1. A target detection method based on deep learning is characterized by comprising the following steps:
s100, collecting data, namely collecting pictures of targets needing to be detected;
s200, selecting an algorithm, analyzing a target to be detected, and selecting a proper algorithm;
s300, cleaning data, analyzing, screening and modifying the collected data to obtain model data conforming to an algorithm;
s400, data conversion, namely converting the model data after data cleaning into a format required by the model;
s500, building a model, and building the model according to the selected algorithm;
s600, training a model, inputting the converted data into the model for training;
s700, testing a model, and predicting pictures of the trained generated format;
and S800, repairing the model, and improving and maintaining the model according to the test result.
2. The target detection method based on deep learning of claim 1, wherein: the data collection in step S100 is specifically to take a picture of an object to be detected or to search from the internet.
3. The target detection method based on deep learning of claim 1, wherein: the algorithm in step S200 is specifically selected by firstly calculating an algorithm with an accuracy of a final result of more than 80% by mathematical theory and mathematical calculation, and using the algorithm as an alternative algorithm of a theoretically suitable algorithm, then training each alternative algorithm, and using a model with high comparison accuracy and recall as an algorithm model.
4. The target detection method based on deep learning of claim 1, wherein: the data cleaning in step S300 is to perform size-unified adjustment on the picture, or perform corrosion and expansion on the picture, or perform color conversion on the picture according to the characteristics of the collected picture.
5. The target detection method based on deep learning of claim 1, wherein: the model building in step S500 is specifically to input a code after building a framework on a server, where the code is a code matching the algorithm selected in step S200.
6. The target detection method based on deep learning of claim 1, wherein: the model in step S800 is repaired through a deep learning algorithm, wherein the deep learning algorithm is performed by performing convolution through a convolutional neural network, extracting features, predicting the features to obtain an area a of a prediction box, comparing the area a of the prediction box with an area B of a real box, and calculating a coefficient J, and the larger the coefficient J, the higher the correlation.
7. The deep learning-based target detection method according to claim 6, wherein: the model repair also requires position loss function and classification loss function calculations,
the position loss function is
Wherein x is the label of the real sample, c is the label of the prediction sample, L is the offset of the upper left corner of the prediction frame, g is the offset of the lower right corner of the prediction frame, n is the total number of samples, L conf is the cross entropy, α is the weight coefficient, and L loc is the cross entropy;
the classification loss function is
Wherein: l loc is cross entropy, x is a predicted sample of a training sample, c is a real sample data, n is a total number of the training samples, pos is a few samples in the training samples, i is a sequence number in the training samples, p is a real sample number, Ne is a few samples in the training samples, o is a number in the real samples, exp is an index based on e, and ij is a default few real samples.
8. The deep learning-based target detection method according to claim 6, wherein: and the threshold value of the coefficient J is H, when the coefficient J is smaller than H, the prediction frame is incorrect, the threshold value H is set to calculate the correct number of the prediction frames by randomly setting the threshold value for a plurality of times, and the set threshold value with the maximum correct number of the prediction frames is taken as H.
9. The deep learning-based target detection method according to claim 7, wherein: and if the position loss function and the classification loss function are smaller than the minimum value, the model is correct, the minimum value of the position loss function and the classification loss function is the minimum value of the position loss function and the classification loss function determined by using a gradient descent method, and the position loss function and the classification loss function are the minimum value when the position loss function and the classification loss function are not changed within 15 minutes.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202010150722.0A CN111382787A (en) | 2020-03-06 | 2020-03-06 | Target detection method based on deep learning |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202010150722.0A CN111382787A (en) | 2020-03-06 | 2020-03-06 | Target detection method based on deep learning |
Publications (1)
| Publication Number | Publication Date |
|---|---|
| CN111382787A true CN111382787A (en) | 2020-07-07 |
Family
ID=71218699
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN202010150722.0A Pending CN111382787A (en) | 2020-03-06 | 2020-03-06 | Target detection method based on deep learning |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN111382787A (en) |
Cited By (1)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN112508916A (en) * | 2020-12-12 | 2021-03-16 | 江西洪都航空工业集团有限责任公司 | Road well lid damage and loss detection method based on mobile video analysis |
Citations (15)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN105719029A (en) * | 2016-03-11 | 2016-06-29 | 中国农业大学 | Combined wind power prediction method based on wind speed fluctuation characteristic extraction |
| CN107423760A (en) * | 2017-07-21 | 2017-12-01 | 西安电子科技大学 | Based on pre-segmentation and the deep learning object detection method returned |
| CN107451602A (en) * | 2017-07-06 | 2017-12-08 | 浙江工业大学 | A kind of fruits and vegetables detection method based on deep learning |
| CN107516135A (en) * | 2017-07-14 | 2017-12-26 | 浙江大学 | A kind of automation monitoring learning method for supporting multi-source data |
| CN108304787A (en) * | 2018-01-17 | 2018-07-20 | 河南工业大学 | Road target detection method based on convolutional neural networks |
| CN108363886A (en) * | 2018-03-08 | 2018-08-03 | 杭州鲁尔物联科技有限公司 | Deformation prediction method and system based on deep learning |
| CN108416695A (en) * | 2018-02-24 | 2018-08-17 | 合肥工业大学 | Electric load probability density prediction technique based on deep learning and system, medium |
| CN108681707A (en) * | 2018-05-15 | 2018-10-19 | 桂林电子科技大学 | Wide-angle model recognizing method and system based on global and local Fusion Features |
| CN109344821A (en) * | 2018-08-30 | 2019-02-15 | 西安电子科技大学 | Small target detecting method based on Fusion Features and deep learning |
| CN109446251A (en) * | 2018-09-04 | 2019-03-08 | 北京睿企信息科技有限公司 | The system and method for distributed artificial intelligence application and development |
| CN109446925A (en) * | 2018-10-08 | 2019-03-08 | 中山大学 | A kind of electric device maintenance algorithm based on convolutional neural networks |
| CN109508650A (en) * | 2018-10-23 | 2019-03-22 | 浙江农林大学 | A kind of wood recognition method based on transfer learning |
| CN110288049A (en) * | 2019-07-02 | 2019-09-27 | 北京字节跳动网络技术有限公司 | Method and apparatus for generating image recognition model |
| CN110414578A (en) * | 2019-07-16 | 2019-11-05 | 上海电机学院 | A kind of transfer learning method based on the multiple batches of training of dynamic and colour gamut conversion |
| CN110826514A (en) * | 2019-11-13 | 2020-02-21 | 国网青海省电力公司海东供电公司 | Construction site violation intelligent identification method based on deep learning |
-
2020
- 2020-03-06 CN CN202010150722.0A patent/CN111382787A/en active Pending
Patent Citations (15)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN105719029A (en) * | 2016-03-11 | 2016-06-29 | 中国农业大学 | Combined wind power prediction method based on wind speed fluctuation characteristic extraction |
| CN107451602A (en) * | 2017-07-06 | 2017-12-08 | 浙江工业大学 | A kind of fruits and vegetables detection method based on deep learning |
| CN107516135A (en) * | 2017-07-14 | 2017-12-26 | 浙江大学 | A kind of automation monitoring learning method for supporting multi-source data |
| CN107423760A (en) * | 2017-07-21 | 2017-12-01 | 西安电子科技大学 | Based on pre-segmentation and the deep learning object detection method returned |
| CN108304787A (en) * | 2018-01-17 | 2018-07-20 | 河南工业大学 | Road target detection method based on convolutional neural networks |
| CN108416695A (en) * | 2018-02-24 | 2018-08-17 | 合肥工业大学 | Electric load probability density prediction technique based on deep learning and system, medium |
| CN108363886A (en) * | 2018-03-08 | 2018-08-03 | 杭州鲁尔物联科技有限公司 | Deformation prediction method and system based on deep learning |
| CN108681707A (en) * | 2018-05-15 | 2018-10-19 | 桂林电子科技大学 | Wide-angle model recognizing method and system based on global and local Fusion Features |
| CN109344821A (en) * | 2018-08-30 | 2019-02-15 | 西安电子科技大学 | Small target detecting method based on Fusion Features and deep learning |
| CN109446251A (en) * | 2018-09-04 | 2019-03-08 | 北京睿企信息科技有限公司 | The system and method for distributed artificial intelligence application and development |
| CN109446925A (en) * | 2018-10-08 | 2019-03-08 | 中山大学 | A kind of electric device maintenance algorithm based on convolutional neural networks |
| CN109508650A (en) * | 2018-10-23 | 2019-03-22 | 浙江农林大学 | A kind of wood recognition method based on transfer learning |
| CN110288049A (en) * | 2019-07-02 | 2019-09-27 | 北京字节跳动网络技术有限公司 | Method and apparatus for generating image recognition model |
| CN110414578A (en) * | 2019-07-16 | 2019-11-05 | 上海电机学院 | A kind of transfer learning method based on the multiple batches of training of dynamic and colour gamut conversion |
| CN110826514A (en) * | 2019-11-13 | 2020-02-21 | 国网青海省电力公司海东供电公司 | Construction site violation intelligent identification method based on deep learning |
Non-Patent Citations (2)
| Title |
|---|
| 刘学军;李长云;万烂军;: "基于Spark的工业大数据处理可视化平台应用研究", no. 12, pages 114 - 115 * |
| 卢艺帆;张松海;: "基于卷积神经网络的光学遥感图像目标检测", no. 14 * |
Cited By (1)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN112508916A (en) * | 2020-12-12 | 2021-03-16 | 江西洪都航空工业集团有限责任公司 | Road well lid damage and loss detection method based on mobile video analysis |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN110188198B (en) | Anti-fraud method and device based on knowledge graph | |
| CN110659744A (en) | Training event prediction model, and method and device for evaluating operation event | |
| CN103745454B (en) | Method for evaluating image processing algorithms or systems based on rank learning | |
| CN111861909B (en) | Network fine granularity image classification method | |
| CN116911929B (en) | Advertisement service terminal and method based on big data | |
| CN117237733A (en) | Breast cancer full-slice image classification method combining self-supervision and weak supervision learning | |
| CN115099310A (en) | Method and device for training model and classifying enterprises | |
| CN117132827A (en) | Hot rolled steel strip surface defect detection method based on improved YOLOv5s network | |
| CN111382787A (en) | Target detection method based on deep learning | |
| CN112346974B (en) | Depth feature embedding-based cross-mobile application program instant defect prediction method | |
| CN114218998A (en) | Power system abnormal behavior analysis method based on hidden Markov model | |
| CN113159535A (en) | Software service performance evaluation method based on entropy weight method | |
| CN117372144A (en) | Wind control strategy intelligent method and system applied to small sample scene | |
| CN112559868A (en) | Information recall method and device, storage medium and electronic equipment | |
| CN112200862A (en) | Training method of target detection model, target detection method and device | |
| CN117078007A (en) | Multi-scale wind control system integrating scale labels and method thereof | |
| CN110705638A (en) | Credit rating prediction classification method using deep network learning fuzzy information feature technology | |
| CN108763289B (en) | Massive heterogeneous sensor format data analysis method | |
| CN115797309A (en) | Surface defect segmentation method based on two-stage incremental learning | |
| CN114387005A (en) | Arbitrage group identification method based on graph classification | |
| CN109308565B (en) | Crowd performance grade identification method and device, storage medium and computer equipment | |
| CN111369124A (en) | Image aesthetic prediction method based on self-generation global features and attention | |
| CN112116238A (en) | Satisfaction evaluation method based on index weight system design | |
| CN115827934B (en) | Enterprise portrait intelligent analysis system and method based on unified social credit code | |
| CN117057743B (en) | Building engineering project consultation cost management method and system thereof |
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 |