CN114782967A - Software defect prediction method based on code visualization learning - Google Patents
Software defect prediction method based on code visualization learning Download PDFInfo
- Publication number
- CN114782967A CN114782967A CN202210276087.XA CN202210276087A CN114782967A CN 114782967 A CN114782967 A CN 114782967A CN 202210276087 A CN202210276087 A CN 202210276087A CN 114782967 A CN114782967 A CN 114782967A
- Authority
- CN
- China
- Prior art keywords
- code
- defect
- source
- source code
- defect prediction
- 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
- 230000007547 defect Effects 0.000 title claims abstract description 79
- 238000000034 method Methods 0.000 title claims abstract description 40
- 230000013016 learning Effects 0.000 title claims abstract description 25
- 238000012800 visualization Methods 0.000 title claims abstract description 20
- 238000000605 extraction Methods 0.000 claims abstract description 17
- 239000013598 vector Substances 0.000 claims abstract description 15
- 238000012549 training Methods 0.000 claims abstract description 11
- 230000002950 deficient Effects 0.000 claims description 17
- 238000009877 rendering Methods 0.000 claims description 12
- 238000013136 deep learning model Methods 0.000 claims description 10
- 239000012634 fragment Substances 0.000 claims description 10
- 230000000007 visual effect Effects 0.000 claims description 9
- 230000008569 process Effects 0.000 claims description 5
- 238000006243 chemical reaction Methods 0.000 claims description 4
- 238000007781 pre-processing Methods 0.000 claims description 4
- 238000012216 screening Methods 0.000 claims description 4
- 238000011161 development Methods 0.000 claims description 3
- 230000010354 integration Effects 0.000 claims description 2
- 239000011159 matrix material Substances 0.000 claims description 2
- 230000007246 mechanism Effects 0.000 claims description 2
- 230000031836 visual learning Effects 0.000 claims description 2
- 230000000694 effects Effects 0.000 claims 1
- 238000013135 deep learning Methods 0.000 abstract description 6
- 238000012360 testing method Methods 0.000 abstract description 3
- 238000013480 data collection Methods 0.000 abstract description 2
- 238000010586 diagram Methods 0.000 description 7
- 238000013528 artificial neural network Methods 0.000 description 2
- 238000005516 engineering process Methods 0.000 description 2
- 239000000284 extract Substances 0.000 description 2
- 238000003058 natural language processing Methods 0.000 description 2
- 238000012545 processing Methods 0.000 description 2
- 238000004458 analytical method Methods 0.000 description 1
- 230000008859 change Effects 0.000 description 1
- 238000010276 construction Methods 0.000 description 1
- 230000002349 favourable effect Effects 0.000 description 1
- 230000006870 function Effects 0.000 description 1
- 238000005259 measurement Methods 0.000 description 1
- 238000012986 modification Methods 0.000 description 1
- 230000004048 modification Effects 0.000 description 1
- 238000004904 shortening Methods 0.000 description 1
- 238000007794 visualization technique Methods 0.000 description 1
Images
Classifications
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F11/00—Error detection; Error correction; Monitoring
- G06F11/36—Preventing errors by testing or debugging software
- G06F11/3668—Software testing
- G06F11/3672—Test management
- G06F11/3684—Test management for test design, e.g. generating new test cases
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F11/00—Error detection; Error correction; Monitoring
- G06F11/36—Preventing errors by testing or debugging software
- G06F11/3668—Software testing
- G06F11/3696—Methods or tools to render software testable
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR 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; 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
- G06N3/084—Backpropagation, e.g. using gradient descent
-
- 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
- Y02P—CLIMATE CHANGE MITIGATION TECHNOLOGIES IN THE PRODUCTION OR PROCESSING OF GOODS
- Y02P90/00—Enabling technologies with a potential contribution to greenhouse gas [GHG] emissions mitigation
- Y02P90/30—Computing systems specially adapted for manufacturing
Abstract
The invention belongs to the technical field of software defect prediction, and particularly discloses a software defect prediction method based on code visualization learning, which comprises the following steps: step 1, data collection and marking; step 2, visually representing the code segments; step 3, fine-tuning the pre-training Efficientnet feature extraction model to obtain a defect prediction model according with the source code characteristics; and 4, realizing the prediction of the defect state of the source code to be measured according to the defect prediction model obtained after fine tuning. The invention can be used for constructing a new end-to-end deep learning defect prediction model by visualizing the code segments in the source code file into images so as to capture the semantic and structural information of the source file and finely adjusting the trained deep learning feature extraction model based on the images, thereby getting rid of the limitation of the natural language model limited by the vector representation of the training words, improving the performance and the general type of the software defect prediction model, reasonably distributing the test resources and improving the software quality.
Description
Technical Field
The invention relates to the technical field of software defect prediction, in particular to a software defect prediction method based on code visual learning.
Background
The software defect prediction technology can help software developers to quickly and accurately position defective software modules, and labor cost and resources are saved. Currently, most of the defect prediction technologies are built in a large amount of source code and historical development data, and a defect prediction model is built in a mode of extracting manually designed metric elements from the data or autonomously learning features by using a deep learning model. However, the artificially designed metric has reached the upper limit of the defect prediction model, and the autonomous learning feature based on the deep learning model is mainly to learn the grammatical and semantic information in the code by a natural language processing method, and although the method can show better results, the model must be regularly trained on a new data set to allow the internal word2vec model to capture new words, resulting in poor extensibility of the models in the actual software engineering project.
Patent document 1 discloses an instant software defect prediction method based on code expression learning, which solves the problem that the conventional neural network cannot solve long-distance dependence by means of a natural language model idea.
However, in patent document 1, the software defect prediction method based on model construction is based on a word vector sequence and code change, which easily results in that only a trained word vector is targeted, and a new vocabulary cannot be captured.
Patent document 2 discloses a hybrid depth defect prediction method based on code segment analysis, which takes a pre-designed key point of a defect library as an entry point of a program slice, extracts a code segment containing defect features from an open source code and expresses the code segment in a vectorization manner, and obtains a hybrid model based on a plurality of deep learning methods, so that the data processing capability and the automatic learning capability of the model can be effectively improved.
However, the defect prediction method constructed in patent document 2 is a vectorization method in which a source code is obtained by program slicing, and a model is also based on a training vector, and cannot comprehensively capture syntactic and semantic information.
The studies of patent document 1 and patent document 2 above provide a good basis for the defect prediction based on the code expression learning, but the defect prediction capability of the current code expression learning has not been sufficiently exploited, and is mainly reflected in:
1. when the self-learning codes are expressed, the self-learning codes are all represented by a natural language processing model and depend on the trained vector sequence; 2. the training of the natural language model consumes much resource and time, and the existing well-trained deep learning model is not considered.
Reference to the literature
Patent document 1 chinese invention patent application publication No. CN 111858323 a, published date: 2020.10.30;
patent document 2 chinese invention patent application publication No. CN 112035345a, published date: 2020.12.04.
disclosure of Invention
Aiming at the technical problems in the prior art, the invention provides a software defect prediction method based on code visualization learning, which improves the performance and the universality of a prediction model, thereby reasonably distributing test resources and improving the software quality.
In order to achieve the purpose, the invention adopts the following technical scheme:
a software defect prediction method based on code visualization learning comprises the following steps:
step 1, collecting required source code files, code submission and defect report information from a GitHub open source community, and realizing defect-free marking of the source code files according to the code submission and the defect report information;
wherein the defective codes and the non-defective codes marked in the source code file are taken as tag data;
step 2, realizing image visual representation of a code segment in a source code file from color, shape and structure by using four modes including a scroll Python image drawing and rendering package, a Google code-predictive javascript library, a javascript Python package and ASCII decimal conversion to obtain four source code image data;
step 3, fine-tuning an Efficientnet-v2 feature extraction model pre-trained on the basis of the ImageNet image by using the label data in the step 1 and the source code picture data in the step 2 to obtain a defect prediction model according with the characteristics of the source code;
and 4, realizing the prediction of the defect state of the detected code segment according to the defect prediction model obtained after fine tuning.
The invention has the following advantages:
as mentioned above, the invention provides a software defect prediction method based on code visualization learning, aiming at the technical problems in the prior art, the method can be used for capturing the semantic and structural information of a source file by visualizing a source code file code segment into an image, and simultaneously constructing a new end-to-end deep learning defect prediction model by finely adjusting a trained deep learning feature extraction model based on an image, so that the limitation of vector representation of a training word on the basis of a natural language model is eliminated, the performance and the universality of the software defect prediction model are improved, test resources are reasonably distributed, and the software quality is improved.
Drawings
FIG. 1 is a schematic flow chart of a software defect prediction method based on code visualization learning according to an embodiment of the present invention;
FIG. 2 is a model schematic diagram of a software defect prediction method based on code visualization learning according to an embodiment of the present invention;
FIG. 3 is a schematic diagram of a defect prediction model of a pre-trained Efficientnet-v2 feature extraction model according to an embodiment of the present invention;
FIG. 4 is a first diagram structure diagram of a code fragment visualization in the embodiment of the present invention;
FIG. 5 is a second diagram of a visualization of code snippets according to an embodiment of the present invention;
FIG. 6 is a third diagram illustrating visualization of code fragments according to an embodiment of the present invention;
FIG. 7 is a diagram illustrating a fourth structure of a code fragment visualization according to an embodiment of the present invention.
Detailed Description
The software development process inevitably has defects which seriously affect the software quality, and if the defects can be found as early as possible, the software development cost can be saved, and the software quality is improved.
In order to avoid intermediate representation such as abstract syntax trees of source codes, the source codes are represented as images by means of a strong learning function of deep learning on image feature extraction, and semantic and syntax information in the source codes can be learned in a more direct mode.
The invention is described in further detail below with reference to the following figures and detailed description:
as shown in fig. 1 and fig. 2, a software bug prediction method based on code visualization learning includes the following steps:
step 1, data collection and marking.
The present invention requires training on source code fragments and both flawless and flawless label data, and therefore, the required source code files, code submissions, and bug reporting information need to be collected first from the GitHub open source community.
And then, realizing non-defect marking of the source code file according to the code submission and the defect report information.
And 1.1, selecting a plurality of open source projects according to the Stars in the GitHub open source community and the development period of the open source projects. The number of open source items selected in this embodiment is, for example, 100 or more.
Step 1.2, preprocessing the collected open source projects, wherein the preprocessing process comprises the following steps:
a) screening out source code files in the open source project;
b) and screening out a problem report reporting the project defects in the open source project.
And step 1.3, establishing connection between the code submission and the problem report through the keyword of the defect report in the code submission, and determining the code submission for repairing the defect according to the connection established between the code submission and the problem report.
Step 1.4. according to the code submission for repairing the defect, the code submission for introducing the defect is identified by adopting the SZZ method.
And step 1.5, marking code segments between the code submission for introducing the defects and the code submission for repairing the defects in the source code file as defective codes, or marking the code segments as non-defective codes.
The defective code and the non-defective code marked through step 1.5 are taken as tag data.
And 2, visually representing the code segment.
According to the method, embedded knowledge of the pre-trained image feature extraction neural network is migrated to a defect classification task, a source code file needs to be converted into image data, and then grammatical and semantic information of a source code is captured.
The invention uses four ways to achieve the visual representation of code snippets from images of color, shape and structure.
As shown in fig. 4 to 7, four image visualization methods include a scroll Python image drawing and rendering package, a Google code-predictive javascript library, a javascript Python package, and ASCII decimal conversion.
The following describes the process of the above four visualization image representation manners in detail:
and 2.1, representing the code fragments as black and white images by using a pillow python image drawing and rendering package as a base line visual representation, and recognizing the code logic structure by learning the visual shapes of characters and words as the shapes of code blocks.
A plain text picture obtained after the scroll python image drawing and the packet rendering method is shown in fig. 4.
And 2.2, firstly, visualizing the code fragment into a rendering code text with color highlighting grammar by using a Google code-rendering javascript library, and then converting the generated rendering code text page into a PNG image by using a python package of imgkit to obtain the color grammar highlighting picture shown in the figure 5.
And 2.3, converting the code fragments into an abstract syntax tree by using a javalang Python package, and generating a grapeviz graph by using the grapeviz Python package, wherein edges in the grapeviz graph represent control flows in the codes so as to capture the sequentiality of the source codes, so that the abstract syntax tree picture shown in FIG. 6 is obtained.
And 2.4, converting the code segments into 8-bit unsigned integer vectors in an ASCII decimal system, and then generating an image from the vectors by arranging vector values of the 8-bit unsigned integer vectors in rows and columns of a matrix image to obtain an RGB picture as shown in FIG. 7.
And 3, fine-tuning the pre-training Efficientnet model.
And (3) fine-tuning the Efficientnet-v2 feature extraction model pre-trained on the basis of the ImageNet image by using the label data in the step (1) and the source code picture data in the step (2) to obtain a feature model according with the source code characteristics.
Step 3.1, the default input of the pre-trained Efficientnet-v2 feature extraction model is 224 × 224, the source code pictures obtained in the step 2 are uniformly converted into source code pictures with the size of 224 × 224, and the conversion process is as follows:
and dividing the source code picture in the step 2 into source code pictures with the sizes larger than and smaller than 224 × 224 pixels.
Source code pictures with size larger than 224 × 224 pixels are cropped to source code pictures with size 224 × 224 pixels, and source code pictures with size smaller than 224 × 224 pixels are filled with blank pixels and form source code pictures with size 224 × 224.
And 3.2, adding a full connection layer based on the Efficientnet-v2 feature extraction model to form a deep learning model, wherein the weight value of the feature layer of the deep learning model is initialized to the weight value of the Efficientnet-v2 feature extraction model, as shown in figure 3.
And 3.3, training the deep learning model constructed in the step 3.2 by using the label data obtained in the step 1 and the source code picture data preprocessed in the step 3.1 in a back propagation mode to obtain a final defect prediction model.
The method extracts the features of the source code image by finely adjusting the existing pre-training feature extraction deep learning model based on the image, is favorable for shortening the training time and extracting the measurement elements for effectively representing the grammar and the semantic information.
And 4, realizing the prediction of the defect state of the source code to be measured according to the defect prediction model obtained after fine tuning.
Step 4.1, converting the code segments of the tested project into pictures in 4 forms by using the code segment visual representation method in the step 2 for the collected code segments of the tested project;
step 4.2, the 4 types of pictures in the step 4.1 are respectively placed into the defect prediction model adjusted in the step 3, so that each picture obtains a corresponding classification result, namely whether the picture is a defect code segment or not;
step 4.3, according to the 4 classification results of the code segments of the tested project obtained in step 4.2, adopting a voting integration mechanism to obtain the final defect state of the code segments, wherein the steps are as follows:
calculating the number n of defective samples in the 4 classification resultsdAnd the number n of defect-free samplesc(ii) a If n isd≥ncIf so, the code segment of the tested item is a defective code segment; if n isd＜ncThen the code segment of the tested item is a non-defective code segment.
The method for predicting the defects of the source code image comprises the steps of visualizing code segments in a source code file into an image, independently learning feature representation in the source code image in a fine adjustment mode by utilizing an existing pre-trained feature extraction model in image processing, and further constructing a defect prediction method which is not limited to a specific word vector space.
It should be understood, however, that the description herein of specific embodiments is by way of illustration only, and not by way of limitation, the intention is to cover all modifications, equivalents, and alternatives falling within the spirit and scope of the present invention as defined by the appended claims.
Claims (7)
1. A software defect prediction method based on code visual learning is characterized by comprising the following steps:
step 1, firstly, collecting required source code files, code submission and defect report information from a GitHub open source community, and then realizing defect-free marking of the source code files according to the code submission and the defect report information;
wherein the defective codes and the non-defective codes marked in the source code file are taken as tag data;
step 2, realizing image visual representation of a code segment in a source code file from color, shape and structure by using four modes including a scroll Python image drawing and rendering package, a Google code-predictive javascript library, a javascript Python package and ASCII decimal conversion to obtain four source code image data;
step 3, fine-tuning an Efficientnet-v2 feature extraction model pre-trained on the basis of the ImageNet image by using the label data in the step 1 and the source code picture data in the step 2 to obtain a defect prediction model according with the characteristics of the source code;
and 4, realizing the prediction of the defect state of the detected code segment according to the defect prediction model obtained after fine tuning.
2. The software defect prediction method based on code visualization learning according to claim 1,
the step 1 specifically comprises the following steps:
step 1.1, selecting a plurality of open source projects according to Stars in a GitHub open source community and the development period of the open source projects;
step 1.2, preprocessing the collected open source projects, wherein the preprocessing process comprises the following steps:
a) screening out source code files in the open source project;
b) screening out a problem report reporting the project defects in the open source project;
step 1.3, establishing connection between the code submission and the problem report through the keyword of the defect report in the code submission, and determining the code submission for repairing the defect according to the connection established between the code submission and the problem report;
step 1.4, identifying the code submission of the introduced defect by adopting an SZZ method according to the code submission of the repaired defect;
step 1.5, marking code segments between code submission introducing defects and code submission repairing defects in the source code file as defective codes; otherwise, it is marked as a non-defective code.
3. The software defect prediction method based on code visualization learning according to claim 2,
the step 2 specifically comprises the following steps:
step 2.1, using a scroll python image drawing and rendering package to express the code fragments into black and white images as baseline visual representation, and recognizing the code logic structure by learning the visual shapes of characters and words as the shapes of code blocks;
step 2.2, firstly, visualizing the code segments into a rendering code text with color highlighting grammar by using a Google code-rendering javascript library, and then converting the generated rendering code text page into a PNG image by using a python package of imgkit;
step 2.3, converting the code fragments into an abstract syntax tree by using a javalang Python package, and generating a grapeviz graph by using the grapeviz Python package, wherein edges in the graph represent control flows in the codes so as to capture the sequentiality of the source codes;
step 2.4. convert the code fragments into 8-bit unsigned integer vectors in ASCII decimal and then effect the generation of an image from the vectors by arranging the vector values of the 8-bit unsigned integer vectors in the rows and columns of the matrix image.
4. The software defect prediction method based on code visualization learning according to claim 3,
the step 3 specifically comprises the following steps:
step 3.1, the default input of the pre-trained Efficientnet-v2 feature extraction model is 224 × 224, and the source code pictures obtained in the step 2 are uniformly converted into source code pictures with the size of 224 × 224;
step 3.2, adding a full connection layer based on an Efficientnet-v2 feature extraction model to form a deep learning model, and initializing the weight value of the feature layer of the deep learning model to the weight value of the Efficientnet-v2 feature extraction model;
and 3.3, training the deep learning model constructed in the step 3.2 by using the label data obtained in the step 1 and the source code picture preprocessed in the step 3.1 in a back propagation mode to obtain a final defect prediction model.
5. The software defect prediction method based on code visualization learning according to claim 4,
the step 3.1 specifically comprises the following steps:
dividing the source code picture in the step 2 into source code pictures with the sizes larger than and smaller than 224 × 224 pixels;
and the source code pictures with the size of more than 224 pixels are cut into source code pictures with the size of 224 pixels, and the source code pictures with the size of less than 224 pixels are filled with blank pixels to form source code pictures with the size of 224 pixels.
6. The software defect prediction method based on code visualization learning according to claim 4,
the step 4 specifically comprises the following steps:
step 4.1, converting the code segments of the tested project into pictures in 4 forms by using the code segment visual representation method in the step 2 for the collected code segments of the tested project;
step 4.2, the 4 types of pictures in the step 4.1 are respectively placed into the defect prediction model adjusted in the step 3, so that each picture obtains a corresponding classification result, namely whether the picture is a defect code segment or not;
and 4.3, obtaining the final defect state of the code segment by adopting a voting integration mechanism according to the 4 classification results of the code segment of the tested project obtained in the step 4.2.
7. The software defect prediction method based on code visualization learning according to claim 6,
the step 4.3 is specifically as follows:
calculating the number n of defective samples in the 4 classification resultsdAnd the number n of defect-free samplesc(ii) a If n isd≥ncIf so, the code segment of the tested item is a defective code segment; if n isd＜ncThen the code segment of the tested item is a non-defective code segment.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202210276087.XA CN114782967B (en) | 2022-03-21 | 2022-03-21 | Software defect prediction method based on code visual chemistry |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202210276087.XA CN114782967B (en) | 2022-03-21 | 2022-03-21 | Software defect prediction method based on code visual chemistry |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN114782967A true CN114782967A (en) | 2022-07-22 |
| CN114782967B CN114782967B (en) | 2024-02-20 |
Family
ID=82426263
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN202210276087.XA Active CN114782967B (en) | 2022-03-21 | 2022-03-21 | Software defect prediction method based on code visual chemistry |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN114782967B (en) |
Cited By (2)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN116662206A (en) * | 2023-07-24 | 2023-08-29 | 泰山学院 | Computer software online real-time visual debugging method and device |
| CN116775871A (en) * | 2023-06-15 | 2023-09-19 | 南京航空航天大学 | Deep learning software defect report classification method based on seBERT pre-training model |
Citations (9)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| US20180150742A1 (en) * | 2016-11-28 | 2018-05-31 | Microsoft Technology Licensing, Llc. | Source code bug prediction |
| CN110162475A (en) * | 2019-05-27 | 2019-08-23 | 浙江工业大学 | A kind of Software Defects Predict Methods based on depth migration |
| CN110597735A (en) * | 2019-09-25 | 2019-12-20 | 北京航空航天大学 | Software defect prediction method for open-source software defect feature deep learning |
| CN112711543A (en) * | 2021-03-26 | 2021-04-27 | 南京航空航天大学 | Workload perception defect prediction method based on weighted software network |
| CN112991267A (en) * | 2021-02-08 | 2021-06-18 | 上海工程技术大学 | Strip steel surface defect detection method based on improved efficientNet-RCNN |
| CN113516097A (en) * | 2021-07-29 | 2021-10-19 | 东北大学秦皇岛分校 | Plant leaf disease identification method based on improved EfficentNet-V2 |
| US20210390342A1 (en) * | 2020-06-12 | 2021-12-16 | Samasource Impact Sourcing, Inc. | Machine Learning (ML) Quality Assurance for Data Curation |
| CN114185769A (en) * | 2021-11-16 | 2022-03-15 | 南京航空航天大学 | Software defect prediction method and terminal based on bidirectional long-short term memory neural network |
| US20230274420A1 (en) * | 2020-07-06 | 2023-08-31 | Harrison-AI Pty Ltd. | Method and system for automated generation of text captions from medical images |
-
2022
- 2022-03-21 CN CN202210276087.XA patent/CN114782967B/en active Active
Patent Citations (9)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| US20180150742A1 (en) * | 2016-11-28 | 2018-05-31 | Microsoft Technology Licensing, Llc. | Source code bug prediction |
| CN110162475A (en) * | 2019-05-27 | 2019-08-23 | 浙江工业大学 | A kind of Software Defects Predict Methods based on depth migration |
| CN110597735A (en) * | 2019-09-25 | 2019-12-20 | 北京航空航天大学 | Software defect prediction method for open-source software defect feature deep learning |
| US20210390342A1 (en) * | 2020-06-12 | 2021-12-16 | Samasource Impact Sourcing, Inc. | Machine Learning (ML) Quality Assurance for Data Curation |
| US20230274420A1 (en) * | 2020-07-06 | 2023-08-31 | Harrison-AI Pty Ltd. | Method and system for automated generation of text captions from medical images |
| CN112991267A (en) * | 2021-02-08 | 2021-06-18 | 上海工程技术大学 | Strip steel surface defect detection method based on improved efficientNet-RCNN |
| CN112711543A (en) * | 2021-03-26 | 2021-04-27 | 南京航空航天大学 | Workload perception defect prediction method based on weighted software network |
| CN113516097A (en) * | 2021-07-29 | 2021-10-19 | 东北大学秦皇岛分校 | Plant leaf disease identification method based on improved EfficentNet-V2 |
| CN114185769A (en) * | 2021-11-16 | 2022-03-15 | 南京航空航天大学 | Software defect prediction method and terminal based on bidirectional long-short term memory neural network |
Non-Patent Citations (4)
| Title |
|---|
| MUHAMMAD FURQON等: "Arrhythmia Classification Using EFFICIENTNETV2with 2-D Scalogram Image Representation", 《2021 TRON SYMPOSIUM(TRONSHOW)》 * |
| 刘英辉等: "多网络联合识别辐射源个体的优化方法", 《信号处理》, no. 04 * |
| 孙旭豪等: "EfficientNet在阴虚证眼象识别中的应用研究", 《中医药信息》, no. 03 * |
| 董泽宇等: "基于特征空间的无监督换向器表面缺陷检测", 《电视技术》, vol. 45, no. 12, pages 104 - 109 * |
Cited By (3)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN116775871A (en) * | 2023-06-15 | 2023-09-19 | 南京航空航天大学 | Deep learning software defect report classification method based on seBERT pre-training model |
| CN116662206A (en) * | 2023-07-24 | 2023-08-29 | 泰山学院 | Computer software online real-time visual debugging method and device |
| CN116662206B (en) * | 2023-07-24 | 2024-02-13 | 泰山学院 | Computer software online real-time visual debugging method and device |
Also Published As
| Publication number | Publication date |
|---|---|
| CN114782967B (en) | 2024-02-20 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN110059694B (en) | Intelligent identification method for character data in complex scene of power industry | |
| CN114782967B (en) | Software defect prediction method based on code visual chemistry | |
| CN108229485B (en) | Method and apparatus for testing user interface | |
| CN111899224A (en) | Nuclear power pipeline defect detection system based on deep learning attention mechanism | |
| CN111767927A (en) | Lightweight license plate recognition method and system based on full convolution network | |
| CN112766334B (en) | Cross-domain image classification method based on pseudo label domain adaptation | |
| CN112949783B (en) | Road crack detection method based on improved U-Net neural network | |
| CN112651966A (en) | Printed circuit board micro-defect detection method based on ACYOLOV4_ CSP | |
| CN111899225A (en) | Nuclear power pipeline defect detection method based on multi-scale pyramid structure | |
| CN116049397B (en) | Sensitive information discovery and automatic classification method based on multi-mode fusion | |
| CN114663904A (en) | PDF document layout detection method, device, equipment and medium | |
| CN112967255A (en) | Shield segment defect type identification and positioning system and method based on deep learning | |
| CN113158969A (en) | Apple appearance defect identification system and method | |
| CN114972759A (en) | Remote sensing image semantic segmentation method based on hierarchical contour cost function | |
| CN115953621A (en) | Semi-supervised hyperspectral image classification method based on unreliable pseudo-label learning | |
| CN115272777A (en) | Semi-supervised image analysis method for power transmission scene | |
| US20230222643A1 (en) | Semantic deep learning and rule optimization for surface corrosion detection and evaluation | |
| CN117437647B (en) | Oracle character detection method based on deep learning and computer vision | |
| CN114550153A (en) | Terminal block image detection and identification method | |
| CN114612702A (en) | Image data annotation system and method based on deep learning | |
| CN111414855B (en) | Telegraph pole sign target detection and identification method based on end-to-end regression model | |
| CN111145314A (en) | Method for extracting place name symbol of scanning electronic map by combining place name labeling | |
| Castillo et al. | Object detection in digital documents based on machine learning algorithms | |
| CN115019243A (en) | Monitoring floater lightweight target detection method and system based on improved YOLOv3 | |
| CN115424243A (en) | Parking stall number identification method, equipment and medium based on yolov5-shufflenetv2 |
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 |