CN108491889A - Image, semantic dividing method, device and computer readable storage medium - Google Patents
Image, semantic dividing method, device and computer readable storage medium Download PDFInfo
- Publication number
- CN108491889A CN108491889A CN201810281858.8A CN201810281858A CN108491889A CN 108491889 A CN108491889 A CN 108491889A CN 201810281858 A CN201810281858 A CN 201810281858A CN 108491889 A CN108491889 A CN 108491889A
- Authority
- CN
- China
- Prior art keywords
- pictures
- image
- semantic segmentation
- semantic
- existing
- 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
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F18/00—Pattern recognition
- G06F18/20—Analysing
- G06F18/24—Classification techniques
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- 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
Landscapes
- Engineering & Computer Science (AREA)
- Theoretical Computer Science (AREA)
- Data Mining & Analysis (AREA)
- Physics & Mathematics (AREA)
- Life Sciences & Earth Sciences (AREA)
- Artificial Intelligence (AREA)
- General Physics & Mathematics (AREA)
- General Engineering & Computer Science (AREA)
- Evolutionary Computation (AREA)
- Bioinformatics & Computational Biology (AREA)
- Evolutionary Biology (AREA)
- Computer Vision & Pattern Recognition (AREA)
- Bioinformatics & Cheminformatics (AREA)
- Computational Linguistics (AREA)
- Biomedical Technology (AREA)
- Biophysics (AREA)
- Health & Medical Sciences (AREA)
- General Health & Medical Sciences (AREA)
- Molecular Biology (AREA)
- Computing Systems (AREA)
- Mathematical Physics (AREA)
- Software Systems (AREA)
- Image Analysis (AREA)
Abstract
The invention discloses a kind of image, semantic dividing method, device and computer readable storage medium, described image semantic segmentation method includes：Existing pictures are expanded, new pictures are obtained；Data training is carried out to initial semantic segmentation algorithm by the new pictures, obtains target semanteme parted pattern；Test pictures are obtained, the test pictures are pre-processed；Pretreated test pictures will be passed through and input the target semanteme parted pattern, obtain semantic segmentation result.In the present invention, increases pictures covering scene range, test pictures are pre-processed, the scope of application of semantic segmentation algorithm is on the one hand improved, on the other hand improve the predictablity rate of semantic segmentation algorithm.
Description
Technical field
The present invention relates to technical field of machine vision more particularly to image, semantic dividing method, devices and computer-readable
Storage medium.
Background technology
With the rapid development of automatic Pilot, also developed rapidly with the relevant technology of automatic Pilot, wherein be based on
The semantic segmentation technology of deep learning is used widely in automatic Pilot.Semantic segmentation technology can carry out pixel to target
The classification of grade, such Pixel-level classification are of great significance to the environment sensing in automatic Pilot scene.Since semantic segmentation is calculated
Method associated data set is broad on road surface mostly, and the good situation of weather acquires, therefore semantic segmentation algorithm is applicable in scene
It is the good situation of weather.However as the development of automatic Pilot technology, autonomous driving vehicle needs the scene adapted to also more next
More, semantic segmentation algorithm can fail when identification carries the road surface picture of hot spot at present, be easy the spot identification on road surface
At pavement strip.
Invention content
The main purpose of the present invention is to provide a kind of image, semantic dividing method, device and computer-readable storage mediums
Matter, it is intended to it solves semantic segmentation algorithm in the prior art and applies under uneven illumination shimming scape, the lower technology of predictablity rate
Problem.
To achieve the above object, the present invention provides a kind of image, semantic dividing method, described image semantic segmentation method packet
It includes：
Existing pictures are expanded, new pictures are obtained；
Data training is carried out to initial semantic segmentation algorithm by the new pictures, obtains target semanteme parted pattern；
Test pictures are obtained, the test pictures are pre-processed；
Pretreated test pictures will be passed through and input the target semanteme parted pattern, obtain semantic segmentation result.
Optionally, described to be expanded to existing pictures, obtaining new pictures includes：
To in existing pictures existing picture carry out contrast enhancement processing and will target scene picture be added described in
Existing pictures obtain new pictures.
Optionally, described that data training is carried out to initial semantic segmentation algorithm by the new pictures, obtain target language
Adopted parted pattern includes：
The new pictures are inputted into initial semantic segmentation algorithm, obtain training result；
By the training result, parameter optimization is carried out to the initial semantic segmentation algorithm, obtains target semantic segmentation
Model.
Optionally, the acquisition test pictures, carrying out pretreatment to the test pictures includes：
Test pictures are obtained, contrast enhancement processing is carried out to the test pictures and white balance is handled.
Optionally, described to pass through the pretreated test pictures input target semanteme parted pattern, obtain semantic point
Cutting result includes later：
Obtain the markup information of the test pictures；
The semantic segmentation result is compared with the markup information, obtains comparing result, and export the comparison
As a result.
In addition, to achieve the above object, the present invention also provides a kind of image, semantic segmenting device, described image semantic segmentations
Device includes：The image, semantic point that memory, processor and being stored in can be run on the memory and on the processor
Program is cut, described image semantic segmentation program realizes image, semantic dividing method as described above when being executed by the processor
Step.
In addition, to achieve the above object, it is described computer-readable the present invention also provides a kind of computer readable storage medium
Image, semantic segmentation procedure is stored on storage medium, described image semantic segmentation program realizes institute as above when being executed by processor
The step of image, semantic dividing method stated.
In the present invention, existing pictures are expanded, obtain new pictures, by new pictures to initial semantic segmentation
Algorithm carries out data training, obtains target semanteme parted pattern, obtains test pictures, is pre-processed to test pictures, will be through
Pretreated test pictures input target semanteme parted pattern is crossed, semantic segmentation result is obtained.In the present invention, increases pictures and cover
Lid scene domain, pre-processes test pictures, on the one hand improves the scope of application of semantic segmentation algorithm, on the other hand carries
The high predictablity rate of semantic segmentation algorithm.
Description of the drawings
Fig. 1 is the image, semantic segmenting device structural schematic diagram for the hardware running environment that the embodiment of the present invention is related to；
Fig. 2 is the flow diagram of image, semantic dividing method first embodiment of the present invention；
Fig. 3 is the operating process schematic diagram of one embodiment of image, semantic dividing method of the present invention；
Fig. 4 is semantic segmentation result schematic diagram in one embodiment of image, semantic dividing method of the present invention.
The embodiments will be further described with reference to the accompanying drawings for the realization, the function and the advantages of the object of the present invention.
Specific implementation mode
It should be appreciated that the specific embodiments described herein are merely illustrative of the present invention, it is not intended to limit the present invention.
As shown in Figure 1, the image, semantic segmenting device knot for the hardware running environment that Fig. 1, which is the embodiment of the present invention, to be related to
Structure schematic diagram.
Image, semantic of embodiment of the present invention segmenting device can be PC, can also be smart mobile phone, tablet computer, portable meter
Calculation machine etc. has the terminal device of certain data-handling capacity.
As shown in Figure 1, the image, semantic segmenting device may include：Processor 1001, such as CPU, network interface 1004,
User interface 1003, memory 1005, communication bus 1002.Wherein, communication bus 1002 is for realizing between these components
Connection communication.User interface 1003 may include display screen (Display), input unit such as keyboard (Keyboard), optional
User interface 1003 can also include standard wireline interface and wireless interface.Network interface 1004 may include optionally standard
Wireline interface, wireless interface (such as WI-FI interfaces).Memory 1005 can be high-speed RAM memory, can also be stable
Memory (non-volatile memory), such as magnetic disk storage.Memory 1005 optionally can also be independently of aforementioned
The storage device of processor 1001.
It will be understood by those skilled in the art that image, semantic segmenting device structure shown in Fig. 1 is not constituted to image
The restriction of semantic segmentation device may include either combining certain components or different than illustrating more or fewer components
Component is arranged.
As shown in Figure 1, as may include that operating system, network are logical in a kind of memory 1005 of computer storage media
Believe module, Subscriber Interface Module SIM and image, semantic segmentation procedure.
In image, semantic segmenting device shown in Fig. 1, network interface 1004 is mainly used for connecting background server, and rear
Platform server is into row data communication；User interface 1003 is mainly used for connecting client (user terminal), and data are carried out with client
Communication；And processor 1001 can be used for calling the image, semantic segmentation procedure stored in memory 1005, and execute following behaviour
Make：
Existing pictures are expanded, new pictures are obtained；
Data training is carried out to initial semantic segmentation algorithm by the new pictures, obtains target semanteme parted pattern；
Test pictures are obtained, the test pictures are pre-processed；
Pretreated test pictures will be passed through and input the target semanteme parted pattern, obtain semantic segmentation result.
Further, described to be expanded to existing pictures, obtaining new pictures includes：
To in existing pictures existing picture carry out contrast enhancement processing and will target scene picture be added described in
Existing pictures obtain new pictures.
Further, described that data training is carried out to initial semantic segmentation algorithm by the new pictures, obtain target
Semantic segmentation model includes：
The new pictures are inputted into initial semantic segmentation algorithm, obtain training result；
By the training result, parameter optimization is carried out to the initial semantic segmentation algorithm, obtains target semantic segmentation
Model.
Further, the acquisition test pictures, carrying out pretreatment to the test pictures includes：
Test pictures are obtained, contrast enhancement processing is carried out to the test pictures and white balance is handled.
Further, described to pass through the pretreated test pictures input target semanteme parted pattern, obtain semanteme
Include after segmentation result：
Obtain the markup information of the test pictures；
The semantic segmentation result is compared with the markup information, obtains comparing result, and export the comparison
As a result.
It is the flow diagram of image, semantic dividing method first embodiment of the present invention with reference to Fig. 2, Fig. 2.
In one embodiment, image, semantic dividing method includes：
Step S10 expands existing pictures, obtains new pictures；
In the present embodiment, existing pictures are expanded to obtain new pictures include：To existing in existing pictures
Picture carries out contrast enhancement processing, and uneven illumination shimming scape picture is added in existing pictures (such as at tree shade bottom
The road surface picture of hot spot is carried down).
In the present embodiment, the existing picture in existing pictures refers to broad and weather is good adopts on road surface
The picture of collection, there is no the situations that uneven illumination is even in existing picture.For example, existing picture is concentrated with several road surface pictures,
These pictures are without hot spot.Carrying out contrast enhancement processing to these pictures, (contrast enhancing is by the brightness in image
Value range stretches or is compressed into the specified brightness display range of display system, to improve the whole or local contrast of image).
Be equivalent to the display effect that hot spot is added in picture, that is, by these without hot spot road surface pictorial simulation at band light
The road surface picture of spot.And the road surface picture with hot spot is added in existing pictures, in the present embodiment, carry the road of hot spot
Face picture can shoot the road surface picture for carrying hot spot under tree shade under road surface picture or other barriers with hot spot.
I.e. in the present embodiment, new pictures include under the uneven illumination shimming scape picture really acquired, such as tree shade with hot spot
Road surface picture, and simulate the obtained road surface picture with hot spot.
Step S20 carries out data training to initial semantic segmentation algorithm by the new pictures, obtains target semanteme point
Cut model；
In the present embodiment, semantic segmentation refers to the image understanding of Pixel-level, i.e., to belonging to each pixel mark in image
Classification.Before deep learning is widely used in computer vision field, people generally use TextonForest and
The method of Random Forest based classifiers carries out semantic segmentation.CNN(Convolutional Neural
Network, convolutional neural networks) image recognition is not only facilitated, equally achieve success in the semantic segmentation problem of image.
A kind of common semantic segmentation method is image block classification (patch classification) in deep learning method, that is, is utilized
Image block around pixel classifies to each pixel.The reason of be that network model generally comprises full articulamentum (fully
Connect layer), and the image of fixed size is required to input.2014, Long of University of California Berkeley et al.
It is proposed full convolutional network (Fully Convolutional Networks, FCN) so that convolutional neural networks need not connect entirely
Layer can realize intensive Pixel-level classification, to classify CNN frameworks as current popular Pixel-level.Due to being not required to
Want full articulamentum, it is possible to semantic segmentation carried out to the image of arbitrary size, and it is faster than conventional method it is upper very much.It
Afterwards, the almost all of advanced method in semantic segmentation field is all based on what the model was extended.In the present embodiment, for semanteme
The selection of partitioning algorithm is selected according to actual conditions, and this is not restricted.
In the present embodiment, by new pictures (including the uneven illumination shimming scape picture really acquired as described above, and
Simulate the obtained road surface picture with hot spot) it is input to initial semantic segmentation algorithm as data, obtain training result.Pass through
The training result, using gradient descent algorithm, (gradient decline is one kind of iterative method, can be used for solving least square problem
(linear and nonlinear can).When solving the model parameter, i.e. unconstrained optimization problem of machine learning algorithm, gradient declines
(Gradient Descent) is commonly used one of method, and another common method is least square method.It is damaged solving
When losing functional minimum value, can by gradient descent method come iterative solution step by step, the loss function that is minimized and
Model parameter value.In turn, it if we need to solve the maximum value of loss function, at this moment just needs to be changed with gradient rise method
Dai Liao.In machine learning, two kinds of gradient descent methods are developed based on basic gradient descent method, respectively under stochastic gradient
Drop method and batch gradient descent method) parameter optimization is carried out to initial semantic segmentation algorithm, until obtaining target semanteme parted pattern.
In the present embodiment, data training is carried out to initial semantic segmentation algorithm by new pictures, obtains target semanteme parted pattern
Specific embodiment, reference can be made to model training method in prior art machine learning, this will not be repeated here.
Step S30 obtains test pictures, is pre-processed to the test pictures；
In the present embodiment, test pictures refer to uneven illumination and the picture under scene, such as the road surface picture with hot spot.It surveys
Attempting piece cannot be identical with the uneven illumination shimming scape picture really acquired in step S10.
In the present embodiment, it includes that contrast compression processing and white balance are handled to carry out pretreatment to test pictures.It reduces
The contrast difference of test pictures is road surface to reduce target semanteme parted pattern by the spot identification in test pictures
The probability of graticule.
Step S40 will pass through pretreated test pictures and input the target semanteme parted pattern, and obtain semantic segmentation knot
Fruit.
In the present embodiment, by target semanteme parted pattern to carrying out image, semantic point by pretreated test pictures
It cuts, i.e., Pixel-level classification is carried out to the content in test pictures, the different objects in test pictures are partitioned into from the angle of pixel,
Each pixel in test pictures is labeled to get to semantic segmentation result.
It is the operating process schematic diagram of one embodiment of image, semantic dividing method of the present invention with reference to Fig. 3, Fig. 3.The present embodiment
In, contrast enhancement processing is carried out to existing picture, acquires uneven illumination shimming scape picture, it will be by contrast enhancement processing
Picture and uneven illumination shimming scape picture afterwards inputs initial semantic segmentation algorithm, and parameter is carried out to initial semantic segmentation algorithm
Optimization, obtains target semanteme parted pattern.Test pictures are pre-processed, pretreatment include contrast compression processing and in vain
Balance Treatment will pass through pretreated picture and input target semanteme parted pattern, by target semanteme parted pattern to passing through
Pretreated test pictures carry out image, semantic segmentation, obtain semantic segmentation result.
In the present embodiment, existing pictures are expanded, obtain new pictures, by new pictures to initial semantic point
It cuts algorithm and carries out data training, obtain target semanteme parted pattern, obtain test pictures, test pictures are pre-processed, it will
Target semanteme parted pattern is inputted by pretreated test pictures, obtains semantic segmentation result.Through this embodiment, increase figure
Piece collection covering scene range, pre-processes test pictures, on the one hand improves the scope of application of semantic segmentation algorithm, another
Aspect improves the predictablity rate of semantic segmentation algorithm.
Further, in one embodiment of image, semantic dividing method of the present invention, step S10 includes：
To in existing pictures existing picture carry out contrast enhancement processing and will target scene picture be added described in
Existing pictures obtain new pictures.
In the present embodiment, existing pictures are expanded to obtain new pictures include：To existing in existing pictures
Picture carries out contrast enhancement processing, and uneven illumination shimming scape picture (i.e. target scene figure is added in existing pictures
Piece, such as the road surface picture with hot spot under tree shade).
In the present embodiment, the existing picture in existing pictures refers to broad and weather is good adopts on road surface
The picture of collection, there is no the situations that uneven illumination is even in existing picture.For example, existing picture is concentrated with several road surface pictures,
These pictures are without hot spot.Carrying out contrast enhancement processing to these pictures, (contrast enhancing is by the brightness in image
Value range stretches or is compressed into the specified brightness display range of display system, to improve the whole or local contrast of image).
Be equivalent to the display effect that hot spot is added in picture, that is, by these without hot spot road surface pictorial simulation at band light
The road surface picture of spot.And the road surface picture with hot spot is added in existing pictures, in the present embodiment, carry the road of hot spot
Face picture can shoot the road surface picture for carrying hot spot under tree shade under road surface picture or other barriers with hot spot.
I.e. in the present embodiment, new pictures include under the uneven illumination shimming scape picture really acquired, such as tree shade with hot spot
Road surface picture, and simulate the obtained road surface picture with hot spot.
The present embodiment increases pictures covering scene range, improves by carrying out data amplification to original pictures
The scope of application of semantic segmentation algorithm.
Further, in one embodiment of image, semantic dividing method of the present invention, step S20 includes：
The new pictures are inputted into initial semantic segmentation algorithm, obtain training result；
By the training result, parameter optimization is carried out to the initial semantic segmentation algorithm, obtains target semantic segmentation
Model.
In the present embodiment, semantic segmentation refers to the image understanding of Pixel-level, i.e., to belonging to each pixel mark in image
Classification.Before deep learning is widely used in computer vision field, people generally use TextonForest and
The method of Random Forest based classifiers carries out semantic segmentation.CNN(Convolutional Neural
Network, convolutional neural networks) image recognition is not only facilitated, equally achieve success in the semantic segmentation problem of image.
A kind of common semantic segmentation method is image block classification (patch classification) in deep learning method, that is, is utilized
Image block around pixel classifies to each pixel.The reason of be that network model generally comprises full articulamentum (fully
Connect layer), and the image of fixed size is required to input.2014, Long of University of California Berkeley et al.
It is proposed full convolutional network (Fully Convolutional Networks, FCN) so that convolutional neural networks need not connect entirely
Layer can realize intensive Pixel-level classification, to classify CNN frameworks as current popular Pixel-level.Due to being not required to
Want full articulamentum, it is possible to semantic segmentation carried out to the image of arbitrary size, and it is faster than conventional method it is upper very much.It
Afterwards, the almost all of advanced method in semantic segmentation field is all based on what the model was extended.In the present embodiment, for semanteme
The selection of partitioning algorithm is selected according to actual conditions, and this is not restricted.
In the present embodiment, by new pictures (including the uneven illumination shimming scape picture really acquired as described above, and
Simulate the obtained road surface picture with hot spot) it is input to initial semantic segmentation algorithm as data, obtain training result.Pass through
The training result, using gradient descent algorithm, (gradient decline is one kind of iterative method, can be used for solving least square problem
(linear and nonlinear can).When solving the model parameter, i.e. unconstrained optimization problem of machine learning algorithm, gradient declines
(Gradient Descent) is commonly used one of method, and another common method is least square method.It is damaged solving
When losing functional minimum value, can by gradient descent method come iterative solution step by step, the loss function that is minimized and
Model parameter value.In turn, it if we need to solve the maximum value of loss function, at this moment just needs to be changed with gradient rise method
Dai Liao.In machine learning, two kinds of gradient descent methods are developed based on basic gradient descent method, respectively under stochastic gradient
Drop method and batch gradient descent method) parameter optimization is carried out to initial semantic segmentation algorithm, until obtaining target semanteme parted pattern.
In the present embodiment, data training is carried out to initial semantic segmentation algorithm by new pictures, obtains target semanteme parted pattern
Specific embodiment, reference can be made to model training method in prior art machine learning, this will not be repeated here.
In the present embodiment, initial semantic segmentation algorithm is trained by new pictures, obtains target semantic segmentation mould
Type reduces probability of the target semanteme parted pattern by spot identification for pavement strip.
Further, in one embodiment of image, semantic dividing method of the present invention, step S30 includes：
Test pictures are obtained, contrast enhancement processing is carried out to the test pictures and white balance is handled.
In the present embodiment, test pictures refer to uneven illumination and the picture under scene, such as the road surface picture with hot spot.It surveys
Attempting piece cannot be identical with the uneven illumination shimming scape picture really acquired in step S10.
In the present embodiment, it includes that contrast compression processing and white balance are handled to carry out pretreatment to test pictures.It reduces
The contrast difference of test pictures, improves semantic segmentation accuracy rate, reduces target semanteme parted pattern by test pictures
In spot identification be pavement strip probability.
Further, in one embodiment of image, semantic dividing method of the present invention, include after step S40：
Obtain the markup information of the test pictures；
The semantic segmentation result is compared with the markup information, obtains comparing result, and export the comparison
As a result.
In the present embodiment, by target semanteme parted pattern to carrying out image, semantic point by pretreated test pictures
It cuts, i.e., Pixel-level classification is carried out to the content in test pictures, the different objects in test pictures are partitioned into from the angle of pixel,
Each pixel in test pictures is labeled to get to semantic segmentation result.It is image language of the present invention with reference to Fig. 4, Fig. 4
Semantic segmentation result schematic diagram in one embodiment of adopted dividing method.If as shown in figure 4, semantic segmentation result by object in Fig. 41,
Object 2, object 3 are labeled as pavement strip, and object 4, object 5 are labeled as hot spot.If the markup information of the test pictures is：It is right
As 1, object 2, object 3 be pavement strip, object 4, object 5 be hot spot, then comparing result be：Semantic segmentation result accuracy
100%.If object in Fig. 41, object 2, object 3, object 4 are labeled as pavement strip by semantic segmentation result, object 5 is marked
For hot spot.If the markup information of the test pictures is：Object 1, object 2, object 3 are pavement strip, and object 4, object 5 are light
Spot, then comparing result be：Semantic segmentation result accuracy 80%.
In the present embodiment, semantic segmentation result is compared with markup information, obtains comparing result, and exports comparison knot
Fruit so that related personnel can know that target semanteme parted pattern carries out test pictures the standard of semantic segmentation by comparing result
True property, can using comparing result as target semanteme parted pattern can practical application judgment criteria.
In addition, the embodiment of the present invention also proposes a kind of computer readable storage medium, the computer readable storage medium
On be stored with image, semantic segmentation procedure, described image semantic segmentation program realizes image as described above when being executed by processor
The step of semantic segmentation method.
Each implementation of the specific embodiment of computer readable storage medium of the present invention and above-mentioned image, semantic dividing method
Example is essentially identical, and this will not be repeated here.
It should be noted that herein, the terms "include", "comprise" or its any other variant are intended to non-row
His property includes, so that process, method, article or system including a series of elements include not only those elements, and
And further include other elements that are not explicitly listed, or further include for this process, method, article or system institute it is intrinsic
Element.In the absence of more restrictions, the element limited by sentence "including a ...", it is not excluded that including this
There is also other identical elements in the process of element, method, article or system.
The embodiments of the present invention are for illustration only, can not represent the quality of embodiment.
Through the above description of the embodiments, those skilled in the art can be understood that above-described embodiment side
Method can add the mode of required general hardware platform to realize by software, naturally it is also possible to by hardware, but in many cases
The former is more preferably embodiment.Based on this understanding, technical scheme of the present invention substantially in other words does the prior art
Going out the part of contribution can be expressed in the form of software products, which is stored in one as described above
In storage medium (such as ROM/RAM, magnetic disc, CD), including some instructions use so that a station terminal equipment (can be mobile phone,
Computer, server, air conditioner or network equipment etc.) execute method described in each embodiment of the present invention.
It these are only the preferred embodiment of the present invention, be not intended to limit the scope of the invention, it is every to utilize this hair
Equivalent structure or equivalent flow shift made by bright specification and accompanying drawing content is applied directly or indirectly in other relevant skills
Art field, is included within the scope of the present invention.
Claims (10)
1. a kind of image, semantic dividing method, which is characterized in that described image semantic segmentation method includes：
Existing pictures are expanded, new pictures are obtained；
Data training is carried out to initial semantic segmentation algorithm by the new pictures, obtains target semanteme parted pattern；
Test pictures are obtained, the test pictures are pre-processed；
Pretreated test pictures will be passed through and input the target semanteme parted pattern, obtain semantic segmentation result.
2. image, semantic dividing method as described in claim 1, which is characterized in that it is described that existing pictures are expanded,
Obtaining new pictures includes：
Contrast enhancement processing is carried out to the existing picture in existing pictures and target scene picture is added described existing
Pictures obtain new pictures.
3. image, semantic dividing method as described in claim 1, which is characterized in that it is described by the new pictures to initial
Semantic segmentation algorithm carries out data training, and obtaining target semanteme parted pattern includes：
The new pictures are inputted into initial semantic segmentation algorithm, obtain training result；
By the training result, parameter optimization is carried out to the initial semantic segmentation algorithm, obtains target semanteme parted pattern.
4. image, semantic dividing method as described in claim 1, which is characterized in that the acquisition test pictures, to the survey
Attempt piece carry out pretreatment include：
Test pictures are obtained, contrast enhancement processing is carried out to the test pictures and white balance is handled.
5. image, semantic dividing method as described in claim 1, which is characterized in that described to pass through pretreated test pictures
The target semanteme parted pattern is inputted, obtaining semantic segmentation result includes later：
Obtain the markup information of the test pictures；
The semantic segmentation result is compared with the markup information, obtains comparing result, and export the comparing result.
6. a kind of image, semantic segmenting device, which is characterized in that described image semantic segmentation device includes：Memory, processor
And it is stored in the image, semantic segmentation procedure that can be run on the memory and on the processor, described image semantic segmentation
Program realizes following steps when being executed by the processor：
Existing pictures are expanded, new pictures are obtained；
Data training is carried out to initial semantic segmentation algorithm by the new pictures, obtains target semanteme parted pattern；
Test pictures are obtained, the test pictures are pre-processed；
Pretreated test pictures will be passed through and input the target semanteme parted pattern, obtain semantic segmentation result.
7. image, semantic segmenting device as claimed in claim 6, which is characterized in that described image semantic segmentation program is described
Processor also realizes following steps when executing：
Contrast enhancement processing is carried out to the existing picture in existing pictures and target scene picture is added described existing
Pictures obtain new pictures.
8. image, semantic segmenting device as claimed in claim 6, which is characterized in that described image semantic segmentation program is described
Processor also realizes following steps when executing：
The new pictures are inputted into initial semantic segmentation algorithm, obtain training result；
By the training result, parameter optimization is carried out to the initial semantic segmentation algorithm, obtains target semanteme parted pattern.
9. image, semantic segmenting device as claimed in claim 6, which is characterized in that described image semantic segmentation program is described
The step of processor also realizes image, semantic dividing method as described in claim 4 or 5 when executing.
10. a kind of computer readable storage medium, which is characterized in that be stored with image language on the computer readable storage medium
Adopted segmentation procedure is realized when described image semantic segmentation program is executed by processor as described in any one of claim 1 to 5
The step of image, semantic dividing method.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201810281858.8A CN108491889A (en) | 2018-04-02 | 2018-04-02 | Image, semantic dividing method, device and computer readable storage medium |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201810281858.8A CN108491889A (en) | 2018-04-02 | 2018-04-02 | Image, semantic dividing method, device and computer readable storage medium |
Publications (1)
| Publication Number | Publication Date |
|---|---|
| CN108491889A true CN108491889A (en) | 2018-09-04 |
Family
ID=63317978
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN201810281858.8A Pending CN108491889A (en) | 2018-04-02 | 2018-04-02 | Image, semantic dividing method, device and computer readable storage medium |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN108491889A (en) |
Cited By (29)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN109299716A (en) * | 2018-08-07 | 2019-02-01 | 北京市商汤科技开发有限公司 | Training method, image partition method, device, equipment and the medium of neural network |
| CN109377509A (en) * | 2018-09-26 | 2019-02-22 | 深圳前海达闼云端智能科技有限公司 | Method, apparatus, storage medium and the equipment of image, semantic segmentation mark |
| CN109492396A (en) * | 2018-11-12 | 2019-03-19 | 杭州安恒信息技术股份有限公司 | Malware Gene Detecting method and apparatus based on semantic segmentation |
| CN110769258A (en) * | 2019-11-05 | 2020-02-07 | 山东浪潮人工智能研究院有限公司 | Image compression method and system for multi-semantic region of specific scene |
| CN110874598A (en) * | 2019-11-05 | 2020-03-10 | 西南交通大学 | Highway water mark detection method based on deep learning |
| CN112560879A (en) * | 2019-09-10 | 2021-03-26 | 中科星图股份有限公司 | Data enhancement method, system and storage medium under multi-data mode environment |
| CN112840376A (en) * | 2018-10-15 | 2021-05-25 | 华为技术有限公司 | Image processing method, device and equipment |
| CN112949516A (en) * | 2021-03-09 | 2021-06-11 | 深圳海翼智新科技有限公司 | Recognition method and device for quilt kicking behavior |
| US11403069B2 (en) | 2017-07-24 | 2022-08-02 | Tesla, Inc. | Accelerated mathematical engine |
| US11409692B2 (en) | 2017-07-24 | 2022-08-09 | Tesla, Inc. | Vector computational unit |
| US11487288B2 (en) | 2017-03-23 | 2022-11-01 | Tesla, Inc. | Data synthesis for autonomous control systems |
| US11537811B2 (en) | 2018-12-04 | 2022-12-27 | Tesla, Inc. | Enhanced object detection for autonomous vehicles based on field view |
| US11561791B2 (en) | 2018-02-01 | 2023-01-24 | Tesla, Inc. | Vector computational unit receiving data elements in parallel from a last row of a computational array |
| US11562231B2 (en) | 2018-09-03 | 2023-01-24 | Tesla, Inc. | Neural networks for embedded devices |
| US11567514B2 (en) | 2019-02-11 | 2023-01-31 | Tesla, Inc. | Autonomous and user controlled vehicle summon to a target |
| US11610117B2 (en) | 2018-12-27 | 2023-03-21 | Tesla, Inc. | System and method for adapting a neural network model on a hardware platform |
| WO2023050651A1 (en) * | 2021-09-29 | 2023-04-06 | 平安科技（深圳）有限公司 | Semantic image segmentation method and apparatus, and device and storage medium |
| US11636333B2 (en) | 2018-07-26 | 2023-04-25 | Tesla, Inc. | Optimizing neural network structures for embedded systems |
| US11665108B2 (en) | 2018-10-25 | 2023-05-30 | Tesla, Inc. | QoS manager for system on a chip communications |
| US11681649B2 (en) | 2017-07-24 | 2023-06-20 | Tesla, Inc. | Computational array microprocessor system using non-consecutive data formatting |
| US11734562B2 (en) | 2018-06-20 | 2023-08-22 | Tesla, Inc. | Data pipeline and deep learning system for autonomous driving |
| US11748620B2 (en) | 2019-02-01 | 2023-09-05 | Tesla, Inc. | Generating ground truth for machine learning from time series elements |
| US11790664B2 (en) | 2019-02-19 | 2023-10-17 | Tesla, Inc. | Estimating object properties using visual image data |
| US11816585B2 (en) | 2018-12-03 | 2023-11-14 | Tesla, Inc. | Machine learning models operating at different frequencies for autonomous vehicles |
| US11841434B2 (en) | 2018-07-20 | 2023-12-12 | Tesla, Inc. | Annotation cross-labeling for autonomous control systems |
| US11893393B2 (en) | 2017-07-24 | 2024-02-06 | Tesla, Inc. | Computational array microprocessor system with hardware arbiter managing memory requests |
| US11893774B2 (en) | 2018-10-11 | 2024-02-06 | Tesla, Inc. | Systems and methods for training machine models with augmented data |
| US12014553B2 (en) | 2019-02-01 | 2024-06-18 | Tesla, Inc. | Predicting three-dimensional features for autonomous driving |
| US12136030B2 (en) | 2023-03-16 | 2024-11-05 | Tesla, Inc. | System and method for adapting a neural network model on a hardware platform |
Citations (9)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN104850845A (en) * | 2015-05-30 | 2015-08-19 | 大连理工大学 | Traffic sign recognition method based on asymmetric convolution neural network |
| CN105631440A (en) * | 2016-02-22 | 2016-06-01 | 清华大学 | Vulnerable road user joint detection method |
| CN106530305A (en) * | 2016-09-23 | 2017-03-22 | 北京市商汤科技开发有限公司 | Semantic segmentation model training and image segmentation method and device, and calculating equipment |
| CN106886801A (en) * | 2017-04-14 | 2017-06-23 | 北京图森未来科技有限公司 | A kind of image, semantic dividing method and device |
| CN107301383A (en) * | 2017-06-07 | 2017-10-27 | 华南理工大学 | A kind of pavement marking recognition methods based on Fast R CNN |
| CN107403183A (en) * | 2017-07-21 | 2017-11-28 | 桂林电子科技大学 | The intelligent scissor method that conformity goal is detected and image segmentation is integrated |
| CN107424159A (en) * | 2017-07-28 | 2017-12-01 | 西安电子科技大学 | Image, semantic dividing method based on super-pixel edge and full convolutional network |
| CN107610141A (en) * | 2017-09-05 | 2018-01-19 | 华南理工大学 | A kind of remote sensing images semantic segmentation method based on deep learning |
| CN107862293A (en) * | 2017-09-14 | 2018-03-30 | 北京航空航天大学 | Radar based on confrontation generation network generates colored semantic image system and method |
-
2018
- 2018-04-02 CN CN201810281858.8A patent/CN108491889A/en active Pending
Patent Citations (9)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN104850845A (en) * | 2015-05-30 | 2015-08-19 | 大连理工大学 | Traffic sign recognition method based on asymmetric convolution neural network |
| CN105631440A (en) * | 2016-02-22 | 2016-06-01 | 清华大学 | Vulnerable road user joint detection method |
| CN106530305A (en) * | 2016-09-23 | 2017-03-22 | 北京市商汤科技开发有限公司 | Semantic segmentation model training and image segmentation method and device, and calculating equipment |
| CN106886801A (en) * | 2017-04-14 | 2017-06-23 | 北京图森未来科技有限公司 | A kind of image, semantic dividing method and device |
| CN107301383A (en) * | 2017-06-07 | 2017-10-27 | 华南理工大学 | A kind of pavement marking recognition methods based on Fast R CNN |
| CN107403183A (en) * | 2017-07-21 | 2017-11-28 | 桂林电子科技大学 | The intelligent scissor method that conformity goal is detected and image segmentation is integrated |
| CN107424159A (en) * | 2017-07-28 | 2017-12-01 | 西安电子科技大学 | Image, semantic dividing method based on super-pixel edge and full convolutional network |
| CN107610141A (en) * | 2017-09-05 | 2018-01-19 | 华南理工大学 | A kind of remote sensing images semantic segmentation method based on deep learning |
| CN107862293A (en) * | 2017-09-14 | 2018-03-30 | 北京航空航天大学 | Radar based on confrontation generation network generates colored semantic image system and method |
Non-Patent Citations (6)
| Title |
|---|
| ADAM PASZKE等: "ENet: A Deep Neural Network Architecture for Real-Time Semantic Segmentation", 《ARXIV:1606.02147V1 [CS.CV]》 * |
| EDUARDO ROMERA等: "Efficient ConvNet for Real-time Semantic Segmentation", 《RESEARCHGATE》 * |
| EDUARDO ROMERA等: "ERFNet: Efficient Residualfor Real-Time Semantic Segmentation Factorized ConvNet", 《IEEE TRANSACTIONS ON INTELLIGENT TRANSPORTATION SYSTEMS》 * |
| EVAN SHELHAMER等: "Fully Convolutional Networks for Semantic Segmentation", 《IEEE TRANSACTIONS ON PATTERN ANALYSIS AND MICHINE INTELLIGENCE》 * |
| TOBIAS POHLEN等: "Full-Resolution Residual Networks for Semantic Segmentation in Street Scenes", 《ARXIV:1611.08323V2 [CS.CV]》 * |
| VIJAY BADRINARAYANAN等: "SegNet: A Deep ConvolutionArchitecture for Image Segmentational Encoder-Decoder", 《IEEE TRANSACTIONS ON PATTERN ANALYSIS AND MACHINE INTELLIGENCE》 * |
Cited By (38)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| US12020476B2 (en) | 2017-03-23 | 2024-06-25 | Tesla, Inc. | Data synthesis for autonomous control systems |
| US11487288B2 (en) | 2017-03-23 | 2022-11-01 | Tesla, Inc. | Data synthesis for autonomous control systems |
| US11409692B2 (en) | 2017-07-24 | 2022-08-09 | Tesla, Inc. | Vector computational unit |
| US12086097B2 (en) | 2017-07-24 | 2024-09-10 | Tesla, Inc. | Vector computational unit |
| US11893393B2 (en) | 2017-07-24 | 2024-02-06 | Tesla, Inc. | Computational array microprocessor system with hardware arbiter managing memory requests |
| US11681649B2 (en) | 2017-07-24 | 2023-06-20 | Tesla, Inc. | Computational array microprocessor system using non-consecutive data formatting |
| US11403069B2 (en) | 2017-07-24 | 2022-08-02 | Tesla, Inc. | Accelerated mathematical engine |
| US11797304B2 (en) | 2018-02-01 | 2023-10-24 | Tesla, Inc. | Instruction set architecture for a vector computational unit |
| US11561791B2 (en) | 2018-02-01 | 2023-01-24 | Tesla, Inc. | Vector computational unit receiving data elements in parallel from a last row of a computational array |
| US11734562B2 (en) | 2018-06-20 | 2023-08-22 | Tesla, Inc. | Data pipeline and deep learning system for autonomous driving |
| US11841434B2 (en) | 2018-07-20 | 2023-12-12 | Tesla, Inc. | Annotation cross-labeling for autonomous control systems |
| US11636333B2 (en) | 2018-07-26 | 2023-04-25 | Tesla, Inc. | Optimizing neural network structures for embedded systems |
| US12079723B2 (en) | 2018-07-26 | 2024-09-03 | Tesla, Inc. | Optimizing neural network structures for embedded systems |
| CN109299716B (en) * | 2018-08-07 | 2021-07-06 | 北京市商汤科技开发有限公司 | Neural network training method, image segmentation method, device, equipment and medium |
| CN109299716A (en) * | 2018-08-07 | 2019-02-01 | 北京市商汤科技开发有限公司 | Training method, image partition method, device, equipment and the medium of neural network |
| US11562231B2 (en) | 2018-09-03 | 2023-01-24 | Tesla, Inc. | Neural networks for embedded devices |
| US11983630B2 (en) | 2018-09-03 | 2024-05-14 | Tesla, Inc. | Neural networks for embedded devices |
| CN109377509A (en) * | 2018-09-26 | 2019-02-22 | 深圳前海达闼云端智能科技有限公司 | Method, apparatus, storage medium and the equipment of image, semantic segmentation mark |
| US11893774B2 (en) | 2018-10-11 | 2024-02-06 | Tesla, Inc. | Systems and methods for training machine models with augmented data |
| US12026863B2 (en) | 2018-10-15 | 2024-07-02 | Huawei Technologies Co., Ltd. | Image processing method and apparatus, and device |
| CN112840376A (en) * | 2018-10-15 | 2021-05-25 | 华为技术有限公司 | Image processing method, device and equipment |
| US11665108B2 (en) | 2018-10-25 | 2023-05-30 | Tesla, Inc. | QoS manager for system on a chip communications |
| CN109492396A (en) * | 2018-11-12 | 2019-03-19 | 杭州安恒信息技术股份有限公司 | Malware Gene Detecting method and apparatus based on semantic segmentation |
| US11816585B2 (en) | 2018-12-03 | 2023-11-14 | Tesla, Inc. | Machine learning models operating at different frequencies for autonomous vehicles |
| US11908171B2 (en) | 2018-12-04 | 2024-02-20 | Tesla, Inc. | Enhanced object detection for autonomous vehicles based on field view |
| US11537811B2 (en) | 2018-12-04 | 2022-12-27 | Tesla, Inc. | Enhanced object detection for autonomous vehicles based on field view |
| US11610117B2 (en) | 2018-12-27 | 2023-03-21 | Tesla, Inc. | System and method for adapting a neural network model on a hardware platform |
| US11748620B2 (en) | 2019-02-01 | 2023-09-05 | Tesla, Inc. | Generating ground truth for machine learning from time series elements |
| US12014553B2 (en) | 2019-02-01 | 2024-06-18 | Tesla, Inc. | Predicting three-dimensional features for autonomous driving |
| US11567514B2 (en) | 2019-02-11 | 2023-01-31 | Tesla, Inc. | Autonomous and user controlled vehicle summon to a target |
| US11790664B2 (en) | 2019-02-19 | 2023-10-17 | Tesla, Inc. | Estimating object properties using visual image data |
| CN112560879A (en) * | 2019-09-10 | 2021-03-26 | 中科星图股份有限公司 | Data enhancement method, system and storage medium under multi-data mode environment |
| CN110874598A (en) * | 2019-11-05 | 2020-03-10 | 西南交通大学 | Highway water mark detection method based on deep learning |
| CN110769258A (en) * | 2019-11-05 | 2020-02-07 | 山东浪潮人工智能研究院有限公司 | Image compression method and system for multi-semantic region of specific scene |
| CN110874598B (en) * | 2019-11-05 | 2022-09-27 | 西南交通大学 | Highway water mark detection method based on deep learning |
| CN112949516A (en) * | 2021-03-09 | 2021-06-11 | 深圳海翼智新科技有限公司 | Recognition method and device for quilt kicking behavior |
| WO2023050651A1 (en) * | 2021-09-29 | 2023-04-06 | 平安科技（深圳）有限公司 | Semantic image segmentation method and apparatus, and device and storage medium |
| US12136030B2 (en) | 2023-03-16 | 2024-11-05 | Tesla, Inc. | System and method for adapting a neural network model on a hardware platform |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN108491889A (en) | Image, semantic dividing method, device and computer readable storage medium | |
| Nogueira et al. | Dynamic multicontext segmentation of remote sensing images based on convolutional networks | |
| CN110533097B (en) | Image definition recognition method and device, electronic equipment and storage medium | |
| CN103578119B (en) | Target detection method in Codebook dynamic scene based on superpixels | |
| US20200034648A1 (en) | Method and apparatus for segmenting sky area, and convolutional neural network | |
| CN108229575A (en) | For detecting the method and apparatus of target | |
| CN109472193A (en) | Method for detecting human face and device | |
| US11301711B2 (en) | Systems and methods for enhancing real-time image recognition | |
| US12106218B2 (en) | Deep forecasted human behavior from digital content | |
| US20160275343A1 (en) | System and method for recognizing offensive images | |
| CN111833372B (en) | Foreground target extraction method and device | |
| CN110717420A (en) | Cultivated land extraction method and system based on remote sensing image and electronic equipment | |
| CN113627402B (en) | Image identification method and related device | |
| CN110033018A (en) | Shape similarity judgment method, device and computer readable storage medium | |
| CN111325271A (en) | Image classification method and device | |
| CN106339719A (en) | Image identification method and image identification device | |
| CN108364278A (en) | A kind of rock core crack extract method and system | |
| CN110020653A (en) | Image, semantic dividing method, device and computer readable storage medium | |
| US11495020B2 (en) | Systems and methods for stream recognition | |
| CN107622278A (en) | A kind of patent image-recognizing method based on Android phone platform | |
| CN108229271A (en) | Interpret the method, apparatus and electronic equipment of remote sensing images | |
| CN108664906B (en) | Method for detecting content in fire scene based on convolutional network | |
| CN113221991A (en) | Method for re-labeling data set by utilizing deep learning | |
| CN111178181B (en) | Traffic scene segmentation method and related device | |
| CN116958729A (en) | Training of object classification model, object classification method, device and storage medium |
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: 20180904 |