CN108596259A - A method of the artificial intelligence training dataset for object identification generates - Google Patents
A method of the artificial intelligence training dataset for object identification generates Download PDFInfo
- Publication number
- CN108596259A CN108596259A CN201810392089.9A CN201810392089A CN108596259A CN 108596259 A CN108596259 A CN 108596259A CN 201810392089 A CN201810392089 A CN 201810392089A CN 108596259 A CN108596259 A CN 108596259A
- Authority
- CN
- China
- Prior art keywords
- data
- article
- artificial intelligence
- simulated environment
- models
- 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
Classifications
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F18/00—Pattern recognition
- G06F18/20—Analysing
- G06F18/21—Design or setup of recognition systems or techniques; Extraction of features in feature space; Blind source separation
- G06F18/214—Generating training patterns; Bootstrap methods, e.g. bagging or boosting
Landscapes
- Engineering & Computer Science (AREA)
- Data Mining & Analysis (AREA)
- Theoretical Computer Science (AREA)
- Computer Vision & Pattern Recognition (AREA)
- Bioinformatics & Cheminformatics (AREA)
- Bioinformatics & Computational Biology (AREA)
- Artificial Intelligence (AREA)
- Evolutionary Biology (AREA)
- Evolutionary Computation (AREA)
- Physics & Mathematics (AREA)
- General Engineering & Computer Science (AREA)
- General Physics & Mathematics (AREA)
- Life Sciences & Earth Sciences (AREA)
- Processing Or Creating Images (AREA)
Abstract
The present invention discloses a kind of method that the artificial intelligence training dataset for object identification generates, and is related to data simulation technical field；The simulated environment that all variables are all controlled variable is established using game engine, the 3D models of the article of artificial intelligence training are generated in simulated environment, the acquisition of data is carried out by the image or video flowing that obtain article 3D models, and according to the label of the simulated environment and the article automatically generated data that are currently generated data, the data of the tape label of acquisition are generated into the artificial intelligence training dataset for carrying out object identification；Training effectiveness is effectively improved, reduces the difficulty that training data is collected, while providing high fidelity and the efficient utilization to video card.
Description
Technical field
The present invention discloses a kind of method that training dataset generates, and is related to data simulation technical field, specifically one
The method that kind is generated for the artificial intelligence training dataset of object identification.
Background technology
In artificial intelligence field, the training of artificial intelligence needs a large amount of training data, and these data need companion
There is relevant parameter to be compareed as training, therefore it is most of researcher and not no related resource to generate or obtain these data
The problem that enterprise faces.Currently in the training of the artificial intelligence of object identification, the collection of training data is to actual object mostly
The work such as label is shot and then is added to complete the generation of data, very elapsed time and manpower.
And the present invention provides a kind of method that the artificial intelligence training dataset for object identification generates, and effectively improves instruction
Practice efficiency, reduces the difficulty that training data is collected, while providing high fidelity and the efficient utilization to video card.Pass through this
One method, artificial intelligence that can be with the controllable data of a large amount of high emulation conditions of collection of high-speed and high-efficiency for 3D object identifications are instructed
Practice data set, or carries out artificial intelligence training in high simulated environment.This method reduces is put into true ring by artificial intelligence
The hsrdware requirements in border can accomplish that environmental variance is controllable under the premise of height emulates, and can effectively improve training data and collect effect
Rate.
Invention content
The present invention is directed to problem of the prior art, provides a kind of artificial intelligence training dataset generation for object identification
Method, have the characteristics that it is versatile, be easy to implement, have broad application prospects.
Concrete scheme proposed by the present invention is：
A method of the artificial intelligence training dataset for object identification generates：
The simulated environment that all variables are all controlled variable is established using game engine, artificial intelligence is generated in simulated environment
The 3D models of the article of training, by obtaining the image of article 3D models or the acquisition of video flowing progress data, and according to working as
Previous existence, at the label of the simulated environment and the article automatically generated data of data, the data generation of the tape label of acquisition was used for
Carry out the artificial intelligence training dataset of object identification.
Game engine establishes the simulated environment that all variables are all controlled variable in the method, in simulated environment
The 3D models for generating the article of artificial intelligence training obtain the image of article 3D models by adjusting the parameter of simulated environment
Or the data of video flowing.
Emulation ring is changed according to the 3D models for the article for generating artificial intelligence training in simulated environment in the method
The parameter in border obtains the image of article 3D models or the data of video flowing.
The ginseng of the article set in the simulated environment for being currently generated data according to game engine in the method
The label of number automatically generated data.
The system that a kind of artificial intelligence training dataset for object identification generates, is based on game engine, utilizes game
Engine establishes the simulated environment that all variables are all controlled variable, and the article of artificial intelligence training is generated in simulated environment
3D models, the acquisition of data is carried out by the image or video flowing that obtain article 3D models, and according to being currently generated data
The label of simulated environment and the article automatically generated data generates the data of the tape label of acquisition for carrying out object identification
Artificial intelligence training dataset.
Game engine establishes the simulated environment that all variables are all controlled variable in the system, in simulated environment
The 3D models for generating the article of artificial intelligence training obtain the image of article 3D models by adjusting the parameter of simulated environment
Or the data of video flowing.
Game engine changes according to the 3D models for the article for generating artificial intelligence training in simulated environment in the system
Become the parameter of simulated environment, obtains the image of article 3D models or the data of video flowing.
The ginseng of the article set in the simulated environment for being currently generated data according to game engine in the system
The label of number automatically generated data.
Usefulness of the present invention is：
The present invention provides a kind of method that the artificial intelligence training dataset for object identification generates, and is established using game engine
The environment of one emulation can efficiently use the performance of GPU by game engine, generate the environment of height emulation；Simulated environment
In, all variables are all controlled variable, effectively increase the controllable factor in training, and environmental parameter can be adjusted arbitrarily, be
Research and training are provided convenience, and the interference of environmental noise is reduced；After generating simulated environment, is generated and trained by game engine
The 3D models of article, carry out data acquisition, and data acquisition can be that image or video flowing can while obtaining data
According to the label of the article automatically generated data of the environment and training that are currently generated, to effectively reduce later data processing
Workload greatly reduces data collection cost.
Description of the drawings
Fig. 1 is the method for the present invention application example schematic diagram.
Fig. 2 the method for the present invention flow diagrams.
Specific implementation mode
The present invention provides a kind of method that the artificial intelligence training dataset for object identification generates：
The simulated environment that all variables are all controlled variable is established using game engine, artificial intelligence is generated in simulated environment
The 3D models of the article of training, by obtaining the image of article 3D models or the acquisition of video flowing progress data, and according to working as
Previous existence, at the label of the simulated environment and the article automatically generated data of data, the data generation of the tape label of acquisition was used for
Carry out the artificial intelligence training dataset of object identification.
The system that the artificial intelligence training dataset for object identification corresponding with the above method generates is provided simultaneously,
Based on game engine, the simulated environment that all variables are all controlled variable is established using game engine, it is raw in simulated environment
At the 3D models of the article of artificial intelligent training, obtaining for data is carried out by the image or video flowing that obtain article 3D models
It takes, and according to the label of the simulated environment and the article automatically generated data that are currently generated data, by the tape label of acquisition
Data generate the artificial intelligence training dataset for carrying out object identification.
In conjunction with attached drawing and specific implementation, the present invention will be further described.
Using the method for the present invention, the environment of an emulation, including background, geographical location, shadow are established using game engine
Effect and physical effect etc., for example a room is generated using game engine, and material is inserted in the room, add in the room
Add light source, the quantity of light source, position, intensity, the parameters such as color；
The 3D models for generating the article of training in the environment by game engine, i.e., generate the target object in the room, leads to
The parameter of adjustment simulated environment is crossed, for example is turned up or turns down the intensity of room light source, obtains the figure of all angles of target object
The video data of picture or shooting article all angles,
And the parameter of the target item set in this room in the simulated environment for being currently generated data according to game engine, than
If parameter be shooting angle, can be 45 °, 90 ° etc., the label of automatically generated data；The data of the tape label of acquisition are generated
Artificial intelligence training dataset for carrying out object identification.
Also with the method for the present invention,
When generating the 3D models of the article of training in the environment by game engine, i.e., the target object is generated in the room,
The parameter of changeable simulated environment, for example light source is not added in the room, but by changing the size in room, obtain target
The image of all angles of object or the video data for shooting article all angles,
And the parameter of the target item set in this room in the simulated environment for being currently generated data according to game engine, than
If parameter be shooting angle, can be 45 °, 90 ° etc., the label of automatically generated data；The data of the tape label of acquisition are generated
Artificial intelligence training dataset for carrying out object identification.
Game engine simulations environment and object are used by the method for the invention, directly by subject image or are regarded
Frequency stream generates a large amount of artificial intelligence training datasets for being suitable for object identification of tape label, can high-speed and high-efficiency collection it is largely high
The data of the controllable artificial intelligence training dataset for object identification of fidelity condition, can also carry out in high simulated environment
Artificial intelligence is trained.
Claims (8)
1. a kind of method that artificial intelligence training dataset for object identification generates, it is characterized in that
The simulated environment that all variables are all controlled variable is established using game engine, artificial intelligence is generated in simulated environment
The 3D models of the article of training, by obtaining the image of article 3D models or the acquisition of video flowing progress data, and according to working as
Previous existence, at the label of the simulated environment and the article automatically generated data of data, the data generation of the tape label of acquisition was used for
Carry out the artificial intelligence training dataset of object identification.
2. according to the method described in claim 1, it is characterized in that it is all the imitative of controlled variable that game engine, which establishes all variables,
True environment, the 3D models that the article of artificial intelligence training is generated in simulated environment obtain by adjusting the parameter of simulated environment
Take the image of article 3D models or the data of video flowing.
3. according to the method described in claim 1, it is characterized in that according to the article for generating artificial intelligence training in simulated environment
3D models change simulated environment parameter, obtain article 3D models image or video flowing data.
4. according to any methods of claim 1-3, it is characterized in that according to game engine in the emulation for being currently generated data
The label of the parameter automatically generated data of the article set in environment.
5. the system that a kind of artificial intelligence training dataset for object identification generates, it is characterized in that it is based on game engine, profit
The simulated environment that all variables are all controlled variable is established with game engine, artificial intelligence training is generated in simulated environment and is used
Article 3D models, the acquisition of data is carried out by the image or video flowing that obtain article 3D models, and according to being currently generated
The label of the simulated environment of data and the article automatically generated data generates the data of the tape label of acquisition for carrying out object
The artificial intelligence training dataset of body identification.
6. system according to claim 5, it is characterized in that it is all the imitative of controlled variable that game engine, which establishes all variables,
True environment, the 3D models that the article of artificial intelligence training is generated in simulated environment obtain by adjusting the parameter of simulated environment
Take the image of article 3D models or the data of video flowing.
7. system according to claim 5, it is characterized in that game engine is trained according to generation artificial intelligence in simulated environment
The 3D models of article change the parameter of simulated environment, obtain the image of article 3D models or the data of video flowing.
8. according to any systems of claim 5-7, it is characterized in that according to game engine in the emulation for being currently generated data
The label of the parameter automatically generated data of the article set in environment.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201810392089.9A CN108596259A (en) | 2018-04-27 | 2018-04-27 | A method of the artificial intelligence training dataset for object identification generates |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201810392089.9A CN108596259A (en) | 2018-04-27 | 2018-04-27 | A method of the artificial intelligence training dataset for object identification generates |
Publications (1)
| Publication Number | Publication Date |
|---|---|
| CN108596259A true CN108596259A (en) | 2018-09-28 |
Family
ID=63610834
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN201810392089.9A Pending CN108596259A (en) | 2018-04-27 | 2018-04-27 | A method of the artificial intelligence training dataset for object identification generates |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN108596259A (en) |
Cited By (5)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN109635853A (en) * | 2018-11-26 | 2019-04-16 | 深圳市玛尔仕文化科技有限公司 | The method for automatically generating artificial intelligence training sample based on computer graphics techniques |
| CN110248093A (en) * | 2019-06-20 | 2019-09-17 | 清华大学深圳研究生院 | A kind of collecting method, system and terminal device |
| CN112396076A (en) * | 2019-08-15 | 2021-02-23 | 杭州海康威视数字技术股份有限公司 | License plate image generation method and device and computer storage medium |
| CN112990121A (en) * | 2021-04-25 | 2021-06-18 | 中国人民解放军国防科技大学 | Target detection method and device, electronic equipment and storage medium |
| US11200411B2 (en) | 2019-10-16 | 2021-12-14 | The Toronto-Dominion Bank | Training a card type classifier with simulated card images |
Citations (5)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN105389584A (en) * | 2015-10-13 | 2016-03-09 | 西北工业大学 | Streetscape semantic annotation method based on convolutional neural network and semantic transfer conjunctive model |
| WO2017156243A1 (en) * | 2016-03-11 | 2017-09-14 | Siemens Aktiengesellschaft | Deep-learning based feature mining for 2.5d sensing image search |
| CN107251026A (en) * | 2014-12-22 | 2017-10-13 | 电子湾有限公司 | System and method for generating fictitious situation |
| CN107452060A (en) * | 2017-06-27 | 2017-12-08 | 西安电子科技大学 | Full angle automatic data collection generates virtual data diversity method |
| CN107451661A (en) * | 2017-06-29 | 2017-12-08 | 西安电子科技大学 | A kind of neutral net transfer learning method based on virtual image data collection |
-
2018
- 2018-04-27 CN CN201810392089.9A patent/CN108596259A/en active Pending
Patent Citations (5)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN107251026A (en) * | 2014-12-22 | 2017-10-13 | 电子湾有限公司 | System and method for generating fictitious situation |
| CN105389584A (en) * | 2015-10-13 | 2016-03-09 | 西北工业大学 | Streetscape semantic annotation method based on convolutional neural network and semantic transfer conjunctive model |
| WO2017156243A1 (en) * | 2016-03-11 | 2017-09-14 | Siemens Aktiengesellschaft | Deep-learning based feature mining for 2.5d sensing image search |
| CN107452060A (en) * | 2017-06-27 | 2017-12-08 | 西安电子科技大学 | Full angle automatic data collection generates virtual data diversity method |
| CN107451661A (en) * | 2017-06-29 | 2017-12-08 | 西安电子科技大学 | A kind of neutral net transfer learning method based on virtual image data collection |
Cited By (6)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN109635853A (en) * | 2018-11-26 | 2019-04-16 | 深圳市玛尔仕文化科技有限公司 | The method for automatically generating artificial intelligence training sample based on computer graphics techniques |
| CN110248093A (en) * | 2019-06-20 | 2019-09-17 | 清华大学深圳研究生院 | A kind of collecting method, system and terminal device |
| CN110248093B (en) * | 2019-06-20 | 2020-09-01 | 清华大学深圳研究生院 | Data acquisition method, system and terminal equipment |
| CN112396076A (en) * | 2019-08-15 | 2021-02-23 | 杭州海康威视数字技术股份有限公司 | License plate image generation method and device and computer storage medium |
| US11200411B2 (en) | 2019-10-16 | 2021-12-14 | The Toronto-Dominion Bank | Training a card type classifier with simulated card images |
| CN112990121A (en) * | 2021-04-25 | 2021-06-18 | 中国人民解放军国防科技大学 | Target detection method and device, electronic equipment and storage medium |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN108596259A (en) | A method of the artificial intelligence training dataset for object identification generates | |
| CN106251389A (en) | The method and apparatus making animation | |
| CN108845802A (en) | Unmanned plane cluster formation interactive simulation verifies system and implementation method | |
| CN104484868B (en) | The moving target of a kind of combination template matches and image outline is taken photo by plane tracking | |
| Xie et al. | Tree modeling with real tree-parts examples | |
| CN1949274A (en) | 3-D visualising method for virtual crowd motion | |
| CN103745497A (en) | Plant growth modeling method and system | |
| CN101814196A (en) | Method for designing three-dimensional cartoon toys based on pictures | |
| CN101706967A (en) | Comprehensive simulation method for realistic cloud layer | |
| CN107730501A (en) | A kind of fish body tail frequency detection method and system | |
| CN116630495B (en) | Virtual digital human model planning system based on AIGC algorithm | |
| Song et al. | Virtual Reality (VR) technology and landscape architecture | |
| Pirk et al. | Modeling plant life in computer graphics | |
| Zhou et al. | A controllable stitch layout strategy for random needle embroidery | |
| CN210072659U (en) | Image recognition module based on deep neural network | |
| CN104794750A (en) | Tree point cloud three-dimensional reconstruction method based on space colonizing algorithm | |
| CN106910233B (en) | Motion simulation method of virtual insect animation role | |
| CN101702243B (en) | Group movement implementation method based on key formation constraint and system thereof | |
| Wu et al. | An interactive virtual training system based on augmented reality | |
| Li et al. | Detection and recognition of Chinese porcelain inlay images of traditional Lingnan architectural decoration based on YOLOv4 technology | |
| CN102938162A (en) | Plant model establishing device and method based on sample | |
| Wang et al. | Design and implementation for virtual forest scene synthesis-editing system | |
| Ye et al. | An intelligent display platform of traditional farming myth’s virtual scene based on unity3D | |
| Wang et al. | Animation Generation Technology Based on Deep Learning: Opportunities and Challenges | |
| CN105160709B (en) | Sampled data conformal compressing method in surface in kind |
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 | ||
| RJ01 | Rejection of invention patent application after publication | ||
| RJ01 | Rejection of invention patent application after publication |
Application publication date: 20180928 |