CN112102338A - Method and device for acquiring image sample data based on deep learning - Google Patents
Method and device for acquiring image sample data based on deep learning Download PDFInfo
- Publication number
- CN112102338A CN112102338A CN202010978516.9A CN202010978516A CN112102338A CN 112102338 A CN112102338 A CN 112102338A CN 202010978516 A CN202010978516 A CN 202010978516A CN 112102338 A CN112102338 A CN 112102338A
- Authority
- CN
- China
- Prior art keywords
- data
- tiles
- tile
- area
- verification
- 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
- G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
- G06T7/00—Image analysis
- G06T7/10—Segmentation; Edge detection
- G06T7/11—Region-based segmentation
-
- G—PHYSICS
- G06—COMPUTING; 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
- G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
- G06T2207/00—Indexing scheme for image analysis or image enhancement
- G06T2207/20—Special algorithmic details
- G06T2207/20081—Training; Learning
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
- G06T2207/00—Indexing scheme for image analysis or image enhancement
- G06T2207/20—Special algorithmic details
- G06T2207/20084—Artificial neural networks [ANN]
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
- G06T2207/00—Indexing scheme for image analysis or image enhancement
- G06T2207/20—Special algorithmic details
- G06T2207/20112—Image segmentation details
- G06T2207/20132—Image cropping
Landscapes
- Engineering & Computer Science (AREA)
- Physics & Mathematics (AREA)
- Theoretical Computer Science (AREA)
- General Physics & Mathematics (AREA)
- General Health & Medical Sciences (AREA)
- Molecular Biology (AREA)
- Biophysics (AREA)
- Computational Linguistics (AREA)
- Data Mining & Analysis (AREA)
- Evolutionary Computation (AREA)
- Artificial Intelligence (AREA)
- Biomedical Technology (AREA)
- Computing Systems (AREA)
- General Engineering & Computer Science (AREA)
- Life Sciences & Earth Sciences (AREA)
- Mathematical Physics (AREA)
- Software Systems (AREA)
- Health & Medical Sciences (AREA)
- Computer Vision & Pattern Recognition (AREA)
- Image Analysis (AREA)
Abstract
The application discloses a method and a device for acquiring image sample data based on deep learning. Wherein, the method comprises the following steps: overlapping the marked data and the image data corresponding to the marked data, wherein in the overlapping area of the marked data and the image sample data, each data point of the marked data is matched with the data point of the aligned image data; cutting out the largest area with a preset shape from the overlapped area as a usable area; performing tile rasterization on the available area to obtain N tiles with preset sizes; selecting M tiles from the N tiles as verification data tiles according to a preset verification data proportion; and performing tile cutting on the image data and the label data of the available area, taking the cut tile data corresponding to the verification data tile as verification sample data, and taking the other cut tile data as training sample data.
Description
Technical Field
The application belongs to the technical field of image processing, and particularly relates to a method and a device for acquiring image sample data based on deep learning.
Background
In image classification, segmentation, detection, and the like, image-oriented deep learning training is generally performed based on sample data. In deep learning training, the original sample data is generally divided into a training sample data set and a verification sample data set, where the proportion of the verification sample data is predetermined, for example, 20%. Deep learning training is carried out by training the sample training data set, and the accuracy of a training result of the deep learning is predicted and evaluated by verifying the sample data set.
In the related art, the training sample data set and the verification sample data set are separated into the training sample data set and the verification sample data set in a file manner based on an artificially produced independent image sample file library, where a sample file in the image sample file library may be a picture, for example, a custom-cut square picture (512 × 512 pixels).
In the process of implementing the technical solution in the embodiments of the present invention, the inventor of the present invention finds that, in the related art, in order to produce a sample library, an independent image sample file needs to be produced first, and if the file is a large file, the file needs to be cut, and then a training sample data set and a verification sample data set are divided according to a certain strategy (for example, a random proportion). When the original image file is large and the sample standard in the image is small relative to the sample file, manual collection and cutting are needed; and for samples of different classes, manual classification by class is required. Therefore, the workload of manpower is very large, and the efficiency is not high.
Disclosure of Invention
The embodiment of the application provides a method and a device for acquiring image sample data based on deep learning, which can solve the problems of large workload and low efficiency of sample data acquisition in the related art.
In order to solve the technical problem, the present application is implemented as follows:
in a first aspect, an embodiment of the present application provides a method for acquiring image sample data based on deep learning, including: overlapping the marked data with the image data corresponding to the marked data, wherein in an overlapping area of the marked data and the image sample data, each data point of the marked data is matched with an aligned data point of the image data; cutting out a largest area with a preset shape from the overlapping area as a usable area; performing tile rasterization on the available area to obtain N tiles with preset sizes, wherein N is an integer greater than 1; selecting M tiles from the N tiles as verification data tiles according to a preset verification data proportion, wherein M is an integer larger than 0 and M < N; and performing tile cutting on the image data and the label data of the available area, taking the cut tile data corresponding to the verification data tiles as verification sample data, and taking other cut tile data as training sample data, wherein the other tile data are the tile data except the verification data in the cut tile data.
In a second aspect, an embodiment of the present application provides an apparatus for acquiring image sample data based on deep learning, including: the overlapping module is used for overlapping the marked data and the image data corresponding to the marked data, wherein in the overlapping area of the marked data and the image data, each data point of the marked data is matched with the aligned data point of the image data; a first cutting module for cutting out a largest area with a preset shape from the overlapping area as a usable area; the rasterization module is used for performing tile rasterization on the available area to obtain N tiles with preset sizes, wherein N is an integer larger than 1; a selecting module, configured to select M tiles from the N tiles as verification data tiles according to a predetermined verification data ratio, where M is an integer greater than 0 and M is less than N; and the second cutting module is used for performing tile cutting on the image data and the marking data of the available area, taking the cut tile data corresponding to the verification data tile as verification sample data, and taking other cut tile data as training sample data, wherein the other cut tile data are the tile data except the verification data in the cut tile data.
In a third aspect, an embodiment of the present application provides an electronic device, which includes a processor, a memory, and a program or instructions stored on the memory and executable on the processor, and when executed by the processor, the program or instructions implement the steps of the method according to the first aspect.
In a fourth aspect, embodiments of the present application provide a readable storage medium, on which a program or instructions are stored, which when executed by a processor implement the steps of the method according to the first aspect.
In a fifth aspect, an embodiment of the present application provides a chip, where the chip includes a processor and a communication interface, where the communication interface is coupled to the processor, and the processor is configured to execute a program or instructions to implement the method according to the first aspect.
In the embodiment of the application, the annotation data and the image data corresponding to the annotation data are matched and aligned, then the maximum available area is cut out by using a preset shape, N tiles with preset sizes are obtained by performing tile rasterization on the available area, verification data tiles are selected from the N tiles, then the image data and the annotation data in the available area are subjected to tile cutting, the cut tile data corresponding to the verification data tiles are used as verification sample data, and the cut other tile data are used as training sample data. Therefore, the image data can be automatically cut into sample data, and a training sample data set and a verification sample data set are marked out, so that the manual workload is reduced, and the efficiency is improved.
Drawings
Fig. 1 is a schematic flowchart illustrating a method for acquiring image sample data based on deep learning according to an embodiment of the present disclosure;
FIG. 2 is a schematic diagram illustrating a usable area determination in an embodiment of the present application;
FIG. 3 is a schematic diagram illustrating rasterization of one usable region in an embodiment of the present application;
FIG. 4 is a schematic diagram illustrating verification of data tile selection in an embodiment of the present application;
fig. 5 is a schematic flowchart illustrating an embodiment of a method for acquiring image sample data based on deep learning according to the present application;
fig. 6 is a schematic structural diagram of an apparatus for acquiring image sample data based on deep learning according to an embodiment of the present disclosure;
fig. 7 is a schematic structural diagram of an electronic device according to an embodiment of the present application.
Detailed Description
The technical solutions in the embodiments of the present application will be clearly and completely described below with reference to the drawings in the embodiments of the present application, and it is obvious that the described embodiments are some, but not all, embodiments of the present application. All other embodiments, which can be derived by a person skilled in the art from the embodiments given herein without making any creative effort, shall fall within the protection scope of the present application.
The terms first, second and the like in the description and in the claims of the present application are used for distinguishing between similar elements and not necessarily for describing a particular sequential or chronological order. It is to be understood that the data so used is interchangeable under appropriate circumstances such that the embodiments of the application are capable of operation in sequences other than those illustrated or described herein. In addition, "and/or" in the specification and claims means at least one of connected objects, a character "/" generally means that a preceding and succeeding related objects are in an "or" relationship.
The method for acquiring image sample data based on deep learning according to the embodiments of the present application is described in detail below with reference to the accompanying drawings through specific embodiments and application scenarios thereof.
Fig. 1 is a flowchart illustrating a method for acquiring image sample data based on deep learning according to an embodiment of the present disclosure, and as shown in fig. 1, the method may include the following steps S100 to S108.
Step S100, overlapping the tagged data with the image data corresponding to the tagged data, wherein in an overlapping area between the tagged data and the image sample data, each data point of the tagged data matches with an aligned data point of the image data.
In the embodiment of the present application, in step S100, the image data to be processed is aligned with the label data thereof, that is, the label data and the image data are in an overlapping relationship, and the data points where the label data and the image data are overlapped are matched.
In this embodiment, the annotation data may be a marked image obtained by marking the image data, that is, the annotation data is sample marking data in which the image data is used as a sample. The marking data can be vector data or pixel data, wherein the vector data is a graphic data structure, and the method is used for processing and expressing point, line and plane data of the space object by recording coordinates on the basis of assuming that the description area is a continuous space; the pixel data is a data structure which represents the space object by a regular pixel array, each data in the array represents the attribute characteristic of the space object, or the pixel data is the pixel array, the row number and the column number of each pixel are used for determining the position, and the information value contained in the pixel is used for representing the characteristics of the type, the grade and the like of the space object.
In one possible implementation, the annotation data may be a single-channel image, and the value of a single pixel is used to indicate the class value of the corresponding image pixel.
In the embodiment of the present application, in step S100, when the annotation data is superimposed on the image data, the annotation data is aligned with the image data corresponding thereto. For example, if the annotation data and the image data are Geographic Information System (GIS) maps, the spatial data recorded in the annotation data and the image data are based on Geographic coordinate systems thereof, the Geographic coordinate systems adopted by the annotation data and the image data may be the same or different, if the two Geographic coordinate systems are the same, the annotation data and the image data may be aligned according to the Geographic coordinate systems, if the two Geographic coordinate systems are different, the spatial data corresponding to the annotation data and the image data may be converted into absolute Geographic longitude and latitude position data, and the annotation data and the image data may be aligned according to absolute longitude and latitude positions. If the annotation data and the image data are not GIS images, the matching of the annotation data and the image data can be determined according to the predetermined convention.
Step S102, cutting out the largest area with a preset shape from the overlapped area as a usable area.
In the embodiment of the present application, in step S102, the maximum predetermined shape is clipped in the overlapping area between the annotation data and the video data, so as to obtain the usable area.
For example, in fig. 2, if the predetermined shape is a rectangle, the dotted-line box portion is a usable area.
In the embodiment of the present application, the predetermined shape may be determined according to practical applications, for example, the predetermined shape may be a rectangle, a diamond, or a regular shape, and is not limited in this embodiment. In addition, in the embodiment of the present application, the usable area of the original image data itself is not specifically limited.
Step S104, performing tile rasterization on the available area to obtain N tiles with preset sizes, wherein N is an integer larger than 1.
In step S104, tile rasterization is performed on the available area, that is, the available area is divided into tiles of a predetermined size, where a value of the predetermined size is not limited in this embodiment and may be determined according to an actual application scenario, for example, the predetermined size may be a square with a side length of T, where a value of T may be determined according to the actual application scenario, and for example, T may be 512 pixels.
In a possible implementation manner, it is difficult to ensure that the available area is exactly N tiles, in this case, in S104, tile rasterization is performed on the available area according to the predetermined size, and the area smaller than the predetermined size after rasterization is discarded, so as to obtain N tiles of the predetermined size.
For example, after rasterizing the available area obtained in fig. 2 using tiles of size T, as shown in fig. 3, blocks that do not satisfy one size T are discarded unused and the rest are available tiles, e.g., the tiles with padding in fig. 3 are available tiles.
In the embodiment of the present application, the usable area includes at least two tile sizes, i.e., N is an integer greater than 1.
And S106, selecting M tiles from the N tiles as verification data tiles according to a preset verification data proportion, wherein M is an integer larger than 0, and M is less than N.
In step S106, M tiles are selected from the N tiles as verification data tiles according to a predetermined verification data ratio, for example, if N is 600 and the predetermined verification data ratio is 20%, M is 120.
In one possible implementation, a random selection may be used when selecting the M tiles. I.e., M indices are randomly drawn within a unique index range of N tiles. For example, after S104, N tiles may be numbered one by one to form a unique index for each tile, and then M indexes are extracted by using a random algorithm, that is, M selected tiles are obtained.
In another possible implementation, M tiles may be selected from the N tiles as verification data tiles in a spatially uniform distribution manner. I.e. M tiles are evenly distributed among N tiles. For example, if the predetermined shape is a rectangle, M tiles are selected from the length direction of the usable area, and H tiles are selected from the width direction of the usable area, where M < ═ W × H, and W/H ~ W/H, where W is the number of tiles in the length direction of the usable area, and H is the number of tiles in the width direction of the usable area.
For example, assuming that H is 3, W is 6, M is 10%, M is 2, W is 2, H is 1, and finally 2 tiles with padding as shown in fig. 4 are selected as the verification data tiles.
And S108, performing tile cutting on the image data and the label data of the available area, taking the cut tile data corresponding to the verification data tile as verification sample data, and taking other cut tile data as training sample data, wherein the other tile data are the tile data except the verification data in the cut tile data.
For example, the N tiles obtained after rasterization are clipped, M tile data corresponding to the tiles of the verification data are used as verification sample data, that is, a verification sample data set, and the other (N-M) tile data in the N clipped tile data are used as a training sample data set.
After obtaining verification sample data and training sample data, in deep learning training, training the neural network model by using the training sample data, adjusting parameters in the neural network model, storing the neural network model obtained by training, and then verifying the accuracy of the neural network model obtained by training by using the verification sample data.
In the embodiment of the application, the annotation data and the image data corresponding to the annotation data are matched and aligned, then the maximum available area is cut out by using a preset shape, N tiles with preset sizes are obtained by performing tile rasterization on the available area, verification data tiles are selected from the N tiles, then the image data and the annotation data in the available area are subjected to tile cutting, the cut tile data corresponding to the verification data tiles are used as verification sample data, and the cut other tile data are used as training sample data. Therefore, the image data can be automatically cut into sample data, and a training sample data set and a verification sample data set are marked out, so that the manual workload is reduced, and the efficiency is improved. In addition, when the verification data tiles are selected, M tiles can be selected from the N tiles as the verification data tiles in a space uniform distribution mode, and therefore the balance among sample categories can be met.
The method for acquiring image sample data based on deep learning provided by the embodiment of the present application is described below by taking a remote sensing satellite image as an example.
Fig. 5 is another schematic flow chart of the method for acquiring image sample data based on deep learning in the embodiment of the present application, as shown in fig. 5, the method may include the following steps:
step S501, obtaining a scene remote sensing satellite image and a corresponding ground feature classification label file, aligning the remote sensing satellite image with the label file space, and then cutting to obtain an available area. For example, a usable area of 10000 × 20000 pixels is clipped.
Step S502, rasterizing the available area by using tiles with preset sizes to obtain N available tiles. For example, if the tile size T is set to 500 pixels, the available area is rasterized to obtain N — 800 tiles, where the number W of the tiles in the longitudinal direction is 40, the number H of the tiles in the width direction is 20, W is 40, and all the tiles are numbered 1 to 800 in sequence.
In step S503, M verification tiles are selected from the N available tiles. For example, if the verification data percentage is 10%, M is 800 × 10% and 80, and a tile index set of 80 verification data tiles is calculated by using a random method or a spatial sampling method.
Step S504, according to the divided areas and the index set of the tiles, a training data set and a verification data set of the sample are cut.
In the method for acquiring image sample data based on deep learning according to the embodiment of the present application, the execution subject may be an apparatus for acquiring image sample data based on deep learning, or a control module in the apparatus for acquiring image sample data based on deep learning, configured to execute the method for acquiring image sample data based on deep learning. In the embodiment of the present application, an example is taken in which an obtaining apparatus for image sample data based on deep learning executes an obtaining method for loading image sample data based on deep learning, and the method for obtaining image sample data based on deep learning provided in the embodiment of the present application is described.
Fig. 6 is a schematic structural diagram of an apparatus for acquiring image sample data based on deep learning according to an embodiment of the present application, as shown in fig. 6, the apparatus mainly includes: an overlap module 601, a first crop module 602, a rasterization module 603, a pick module 604, and a second crop module 605.
In this embodiment of the present application, the overlapping module 601 is configured to overlap annotation data with image data corresponding to the annotation data, where in an overlapping area between the annotation data and the image data, each data point of the annotation data is matched with an aligned data point of the image data; a first cropping module 602, configured to crop a largest region of a predetermined shape from the overlapping region as a usable region; a rasterizing module 603, configured to perform tile rasterization on the available area to obtain N tiles of a predetermined size, where N is an integer greater than 1; a selecting module 604, configured to select M tiles from the N tiles as verification data tiles according to a predetermined verification data ratio, where M is an integer greater than 0 and M is less than N; a second clipping module 605, configured to perform tile clipping on the image data and the labeled data in the available area, use the clipped tile data corresponding to the tile of the verification data as verification sample data, and use the clipped other tile data as training sample data, where the other tile data is tile data of the clipped tile data except the verification data.
In one possible implementation, the selecting module 604 selects M tiles from the N tiles as verification data tiles, including: and randomly selecting M tiles from the N tiles to serve as verification data tiles.
In one possible implementation, the selecting module 604 selects M tiles from the N tiles as verification data tiles, including: and selecting M tiles from the N tiles as verification data tiles in a space uniform distribution mode.
In a possible implementation manner, the selecting module 604 selects M tiles from the N tiles as verification data tiles in a spatially uniform distribution manner, including: if the predetermined shape is a rectangle, selecting M tiles from the length direction of the usable area, and selecting H tiles from the width direction of the usable area, wherein M < ═ W × H, W/H ~ W/H, W is the number of tiles in the length direction of the usable area, and H is the number of tiles in the width direction of the usable area.
In one possible implementation, the rasterizing module 603 performs tile rasterization on the available area to obtain N tiles of a predetermined size, including: and performing tile rasterization on the available area according to the preset size, and abandoning the area which is smaller than the preset size after rasterization to obtain N tiles with the preset size.
In one possible implementation, the annotation data is a single-channel image, and the value of each pixel of the image is used to indicate the class value of the image pixel.
The device for acquiring image sample data based on deep learning in the embodiment of the present application may be a device, or may be a component, an integrated circuit, or a chip in a terminal. The device can be mobile electronic equipment or non-mobile electronic equipment. By way of example, the mobile electronic device may be a mobile phone, a tablet computer, a notebook computer, a palm top computer, a vehicle-mounted electronic device, a wearable device, an ultra-mobile personal computer (UMPC), a netbook or a Personal Digital Assistant (PDA), and the like, and the non-mobile electronic device may be a server, a Network Attached Storage (NAS), a Personal Computer (PC), a Television (TV), a teller machine or a self-service machine, and the like, and the embodiments of the present application are not particularly limited.
The device for acquiring image sample data based on deep learning in the embodiment of the present application may be a device having an operating system. The operating system may be an Android (Android) operating system, an ios operating system, or other possible operating systems, and embodiments of the present application are not limited specifically.
The device for acquiring image sample data based on deep learning provided by the embodiment of the application can realize each process in the method embodiments and has the same effect, and is not repeated here for avoiding repetition.
Optionally, an electronic device is further provided in an embodiment of the present application, and fig. 7 is a schematic diagram of a hardware structure of an electronic device implementing the embodiment of the present application. As shown in fig. 7, the electronic device includes a processor 701, a memory 702, and a program or an instruction stored in the memory 702 and executable on the processor 701, where the program or the instruction is executed by the processor 710 to implement the processes of the above-mentioned method for acquiring image sample data based on deep learning, and can achieve the same technical effects, and no further description is provided herein for avoiding repetition.
It should be noted that the electronic devices in the embodiments of the present application include the mobile electronic devices and the non-mobile electronic devices described above.
The embodiment of the present application further provides a readable storage medium, where a program or an instruction is stored on the readable storage medium, and when the program or the instruction is executed by a processor, the program or the instruction implements each process of the above method for acquiring image sample data based on deep learning, and can achieve the same technical effect, and in order to avoid repetition, details are not repeated here.
The processor is the processor in the electronic device described in the above embodiment. The readable storage medium includes a computer readable storage medium, such as a Read-Only Memory (ROM), a Random Access Memory (RAM), a magnetic disk or an optical disk, and so on.
The embodiment of the present application further provides a chip, where the chip includes a processor and a communication interface, the communication interface is coupled to the processor, and the processor is configured to run a program or an instruction, so as to implement each process of the above method for acquiring image sample data based on deep learning, and achieve the same technical effect, and in order to avoid repetition, the description is omitted here.
It should be understood that the chips mentioned in the embodiments of the present application may also be referred to as system-on-chip, system-on-chip or system-on-chip, etc.
It should be noted that, in this document, the terms "comprises," "comprising," or any other variation thereof, are intended to cover a non-exclusive inclusion, such that a process, method, article, or apparatus that comprises a list of elements does not include only those elements but may include other elements not expressly listed or inherent to such process, method, article, or apparatus. Without further limitation, an element defined by the phrase "comprising an … …" does not exclude the presence of other like elements in a process, method, article, or apparatus that comprises the element. Further, it should be noted that the scope of the methods and apparatus of the embodiments of the present application is not limited to performing the functions in the order illustrated or discussed, but may include performing the functions in a substantially simultaneous manner or in a reverse order based on the functions involved, e.g., the methods described may be performed in an order different than that described, and various steps may be added, omitted, or combined. In addition, features described with reference to certain examples may be combined in other examples.
Through the above description of the embodiments, those skilled in the art will clearly understand that the method of the above embodiments can be implemented by software plus a necessary general hardware platform, and certainly can also be implemented by hardware, but in many cases, the former is a better implementation manner. Based on such understanding, the technical solutions of the present application may be embodied in the form of a software product, which is stored in a storage medium (such as ROM/RAM, magnetic disk, optical disk) and includes instructions for enabling a terminal (such as a mobile phone, a computer, a server, an air conditioner, or a network device) to execute the method according to the embodiments of the present application.
While the present embodiments have been described with reference to the accompanying drawings, it is to be understood that the invention is not limited to the precise embodiments described above, which are meant to be illustrative and not restrictive, and that various changes may be made therein by those skilled in the art without departing from the spirit and scope of the invention as defined by the appended claims.
Claims (14)
1. An image sample data acquisition method based on deep learning is characterized by comprising the following steps:
overlapping the marked data with the image data corresponding to the marked data, wherein in the overlapping area of the marked data and the image data, each data point of the marked data is matched with the aligned data point of the image data;
cutting out a largest area with a preset shape from the overlapping area as a usable area;
performing tile rasterization on the available area to obtain N tiles with preset sizes, wherein N is an integer greater than 1;
selecting M tiles from the N tiles as verification data tiles according to a preset verification data proportion, wherein M is an integer larger than 0 and M < N;
and performing tile cutting on the image data and the label data of the available area, taking the cut tile data corresponding to the verification data tiles as verification sample data, and taking other cut tile data as training sample data, wherein the other tile data are the tile data except the verification data in the cut tile data.
2. The method of claim 1, wherein selecting M tiles from among the N tiles as tiles for verification data comprises:
and randomly selecting M tiles from the N tiles to serve as verification data tiles.
3. The method of claim 1, wherein selecting M tiles from among the N tiles as tiles for verification data comprises:
and selecting M tiles from the N tiles as verification data tiles in a space uniform distribution mode.
4. The method of claim 3, wherein selecting M tiles from among N tiles as tiles for verification data in a spatially uniform manner comprises:
if the predetermined shape is a rectangle, selecting M tiles from the length direction of the usable area, and selecting H tiles from the width direction of the usable area, wherein M < ═ W × H, W/H ~ W/H, W is the number of tiles in the length direction of the usable area, and H is the number of tiles in the width direction of the usable area.
5. The method of any one of claims 1 to 4, wherein tile rasterizing the available area for N tiles of a predetermined size comprises:
and performing tile rasterization on the available area according to the preset size, and abandoning the area which is smaller than the preset size after rasterization to obtain N tiles with the preset size.
6. The method of any one of claims 1 to 4, wherein the annotation data is a single-channel image, and the value of each pixel of the image is used to indicate the class value of the image pixel.
7. An image sample data acquisition device based on deep learning, comprising:
the overlapping module is used for overlapping the marked data and the image data corresponding to the marked data, wherein in the overlapping area of the marked data and the image data, each data point of the marked data is matched with the aligned data point of the image data;
a first cutting module for cutting out a largest area with a preset shape from the overlapping area as a usable area;
the rasterization module is used for performing tile rasterization on the available area to obtain N tiles with preset sizes, wherein N is an integer larger than 1;
a selecting module, configured to select M tiles from the N tiles as verification data tiles according to a predetermined verification data ratio, where M is an integer greater than 0 and M is less than N;
and the second cutting module is used for performing tile cutting on the image data and the marking data of the available area, taking the cut tile data corresponding to the verification data tile as verification sample data, and taking other cut tile data as training sample data, wherein the other cut tile data are the tile data except the verification data in the cut tile data.
8. The apparatus of claim 7, wherein said selecting module selects M tiles from N of said tiles as verification data tiles, comprising:
and randomly selecting M tiles from the N tiles to serve as verification data tiles.
9. The apparatus of claim 7, wherein said selecting module selects M tiles from N of said tiles as verification data tiles, comprising:
and selecting M tiles from the N tiles as verification data tiles in a space uniform distribution mode.
10. The apparatus of claim 9, wherein said selecting module selects M tiles from N said tiles as verification data tiles in a spatially uniform manner, comprising:
if the predetermined shape is a rectangle, selecting M tiles from the length direction of the usable area, and selecting H tiles from the width direction of the usable area, wherein M < ═ W × H, W/H ~ W/H, W is the number of tiles in the length direction of the usable area, and H is the number of tiles in the width direction of the usable area.
11. The apparatus according to any one of claims 7 to 10, wherein the rasterizing module tiles the available area into N tiles of a predetermined size, including:
and performing tile rasterization on the available area according to the preset size, and abandoning the area which is smaller than the preset size after rasterization to obtain N tiles with the preset size.
12. The apparatus according to any one of claims 7 to 10, wherein the annotation data is a single-channel image, and the value of each pixel of the image is used to indicate the class value of the image pixel.
13. An electronic device comprising a processor, a memory and a program or instructions stored on the memory and executable on the processor, wherein the program or instructions, when executed by the processor, implement the steps of the method for acquiring image sample data based on deep learning according to claims 1-6.
14. A readable storage medium, storing thereon a program or instructions, which when executed by a processor, implement the steps of the method for acquiring image sample data based on deep learning according to claims 1-6.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202010978516.9A CN112102338A (en) | 2020-09-17 | 2020-09-17 | Method and device for acquiring image sample data based on deep learning |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202010978516.9A CN112102338A (en) | 2020-09-17 | 2020-09-17 | Method and device for acquiring image sample data based on deep learning |
Publications (1)
| Publication Number | Publication Date |
|---|---|
| CN112102338A true CN112102338A (en) | 2020-12-18 |
Family
ID=73759738
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN202010978516.9A Pending CN112102338A (en) | 2020-09-17 | 2020-09-17 | Method and device for acquiring image sample data based on deep learning |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN112102338A (en) |
Cited By (2)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN113223042A (en) * | 2021-05-19 | 2021-08-06 | 自然资源部国土卫星遥感应用中心 | Intelligent acquisition method and equipment for remote sensing image deep learning sample |
| WO2024098396A1 (en) * | 2022-11-11 | 2024-05-16 | 杭州数梦工场科技有限公司 | Method and apparatus for determining overlap of spatial data, and device and storage medium |
Citations (8)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN109670060A (en) * | 2018-12-10 | 2019-04-23 | 北京航天泰坦科技股份有限公司 | A kind of remote sensing image semi-automation mask method based on deep learning |
| CN110070548A (en) * | 2019-04-22 | 2019-07-30 | 杨勇 | A kind of deep learning training sample optimization method |
| US20190258824A1 (en) * | 2018-02-22 | 2019-08-22 | International Business Machines Corporation | Identification of optimal data utility-preserving anonymization techniques by evaluation of a plurality of anonymization techniques on sample data sets that correspond to different anonymization categories |
| CN110348506A (en) * | 2019-07-03 | 2019-10-18 | 广州大学 | Land use classes method, storage medium and calculating equipment based on remote sensing images |
| CN110569699A (en) * | 2018-09-07 | 2019-12-13 | 阿里巴巴集团控股有限公司 | Method and device for carrying out target sampling on picture |
| CN110674735A (en) * | 2019-09-23 | 2020-01-10 | 中国科学院地理科学与资源研究所 | Agricultural facility remote sensing extraction method and device based on fine classification |
| US10540572B1 (en) * | 2019-01-31 | 2020-01-21 | StradVision, Inc. | Method for auto-labeling training images for use in deep learning network to analyze images with high precision, and auto-labeling device using the same |
| CN110990516A (en) * | 2019-12-13 | 2020-04-10 | 苏州中科图新网络科技有限公司 | Map data processing method and device and server |
-
2020
- 2020-09-17 CN CN202010978516.9A patent/CN112102338A/en active Pending
Patent Citations (9)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| US20190258824A1 (en) * | 2018-02-22 | 2019-08-22 | International Business Machines Corporation | Identification of optimal data utility-preserving anonymization techniques by evaluation of a plurality of anonymization techniques on sample data sets that correspond to different anonymization categories |
| CN110569699A (en) * | 2018-09-07 | 2019-12-13 | 阿里巴巴集团控股有限公司 | Method and device for carrying out target sampling on picture |
| CN109670060A (en) * | 2018-12-10 | 2019-04-23 | 北京航天泰坦科技股份有限公司 | A kind of remote sensing image semi-automation mask method based on deep learning |
| US10540572B1 (en) * | 2019-01-31 | 2020-01-21 | StradVision, Inc. | Method for auto-labeling training images for use in deep learning network to analyze images with high precision, and auto-labeling device using the same |
| CN111507335A (en) * | 2019-01-31 | 2020-08-07 | 斯特拉德视觉公司 | Method and device for automatically labeling training images for deep learning network |
| CN110070548A (en) * | 2019-04-22 | 2019-07-30 | 杨勇 | A kind of deep learning training sample optimization method |
| CN110348506A (en) * | 2019-07-03 | 2019-10-18 | 广州大学 | Land use classes method, storage medium and calculating equipment based on remote sensing images |
| CN110674735A (en) * | 2019-09-23 | 2020-01-10 | 中国科学院地理科学与资源研究所 | Agricultural facility remote sensing extraction method and device based on fine classification |
| CN110990516A (en) * | 2019-12-13 | 2020-04-10 | 苏州中科图新网络科技有限公司 | Map data processing method and device and server |
Cited By (2)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN113223042A (en) * | 2021-05-19 | 2021-08-06 | 自然资源部国土卫星遥感应用中心 | Intelligent acquisition method and equipment for remote sensing image deep learning sample |
| WO2024098396A1 (en) * | 2022-11-11 | 2024-05-16 | 杭州数梦工场科技有限公司 | Method and apparatus for determining overlap of spatial data, and device and storage medium |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN109960742B (en) | Local information searching method and device | |
| US20200380263A1 (en) | Detecting key frames in video compression in an artificial intelligence semiconductor solution | |
| CN111260037B (en) | Convolution operation method and device of image data, electronic equipment and storage medium | |
| CN111612822B (en) | Object tracking method, device, computer equipment and storage medium | |
| EP3543910B1 (en) | Cloud detection in aerial imagery | |
| CN107644423B (en) | Scene segmentation-based video data real-time processing method and device and computing equipment | |
| US20210200984A1 (en) | Information Processing Apparatus, Information Processing Method and Non-Transitory Computer-Readable Storage Medium | |
| CN112102338A (en) | Method and device for acquiring image sample data based on deep learning | |
| CN112307853A (en) | Detection method of aerial image, storage medium and electronic device | |
| CN111104813A (en) | Two-dimensional code image key point detection method and device, electronic equipment and storage medium | |
| CN104520848A (en) | Searching for events by attendants | |
| US9165220B2 (en) | Image object recognition based on a feature vector with context information | |
| CN113516666A (en) | Image cropping method and device, computer equipment and storage medium | |
| CN112183321A (en) | Method and device for optimizing machine learning model, computer equipment and storage medium | |
| CN110188815B (en) | Feature point sampling method, device, equipment and storage medium | |
| CN109829421B (en) | Method and device for vehicle detection and computer readable storage medium | |
| CN113298042B (en) | Remote sensing image data processing method and device, storage medium and computer equipment | |
| CN112668675B (en) | Image processing method and device, computer equipment and storage medium | |
| CN108573510B (en) | Grid map vectorization method and device | |
| CN111741329B (en) | Video processing method, device, equipment and storage medium | |
| CN117319662A (en) | Image compression and decompression method and system for human body target recognition | |
| CN112926426A (en) | Ship identification method, system, equipment and storage medium based on monitoring video | |
| US8755606B2 (en) | Systems and methods for efficient feature extraction accuracy using imperfect extractors | |
| CN116610140A (en) | Unmanned aerial vehicle control method, unmanned aerial vehicle control device, computer equipment and storage medium | |
| CN112487943B (en) | Key frame de-duplication method and device and electronic equipment |
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 |