CN112749788A - Super-resolution picture model generation method and device, electronic equipment and storage medium - Google Patents
Super-resolution picture model generation method and device, electronic equipment and storage medium Download PDFInfo
- Publication number
- CN112749788A CN112749788A CN202011495417.1A CN202011495417A CN112749788A CN 112749788 A CN112749788 A CN 112749788A CN 202011495417 A CN202011495417 A CN 202011495417A CN 112749788 A CN112749788 A CN 112749788A
- Authority
- CN
- China
- Prior art keywords
- resolution picture
- resolution
- generator
- picture
- super
- 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
- 238000000034 method Methods 0.000 title claims abstract description 43
- 238000012545 processing Methods 0.000 claims abstract description 9
- 238000012549 training Methods 0.000 claims description 29
- 230000006870 function Effects 0.000 claims description 21
- 238000007781 pre-processing Methods 0.000 claims description 12
- 238000013528 artificial neural network Methods 0.000 claims description 11
- 238000012360 testing method Methods 0.000 claims description 8
- 238000004590 computer program Methods 0.000 claims description 7
- 238000010586 diagram Methods 0.000 description 8
- 101100365548 Caenorhabditis elegans set-14 gene Proteins 0.000 description 4
- 230000004913 activation Effects 0.000 description 4
- 238000013527 convolutional neural network Methods 0.000 description 4
- 125000004122 cyclic group Chemical group 0.000 description 4
- 230000009286 beneficial effect Effects 0.000 description 3
- 238000004364 calculation method Methods 0.000 description 3
- 238000004891 communication Methods 0.000 description 2
- 230000000694 effects Effects 0.000 description 2
- 238000005516 engineering process Methods 0.000 description 2
- 239000000284 extract Substances 0.000 description 2
- 230000006872 improvement Effects 0.000 description 2
- 238000011176 pooling Methods 0.000 description 2
- 230000008569 process Effects 0.000 description 2
- HPTJABJPZMULFH-UHFFFAOYSA-N 12-[(Cyclohexylcarbamoyl)amino]dodecanoic acid Chemical compound OC(=O)CCCCCCCCCCCNC(=O)NC1CCCCC1 HPTJABJPZMULFH-UHFFFAOYSA-N 0.000 description 1
- 230000009471 action Effects 0.000 description 1
- 230000003213 activating effect Effects 0.000 description 1
- 238000013434 data augmentation Methods 0.000 description 1
- 238000013135 deep learning Methods 0.000 description 1
- 238000011161 development Methods 0.000 description 1
- 238000011156 evaluation Methods 0.000 description 1
- 238000002474 experimental method Methods 0.000 description 1
- 230000002349 favourable effect Effects 0.000 description 1
- 230000010365 information processing Effects 0.000 description 1
- 230000003993 interaction Effects 0.000 description 1
- 238000013507 mapping Methods 0.000 description 1
- 238000012986 modification Methods 0.000 description 1
- 230000004048 modification Effects 0.000 description 1
- 238000010606 normalization Methods 0.000 description 1
- 230000003287 optical effect Effects 0.000 description 1
- 238000005457 optimization Methods 0.000 description 1
- 238000011160 research Methods 0.000 description 1
Images
Classifications
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
- G06N3/00—Computing arrangements based on biological models
- G06N3/02—Neural networks
- G06N3/04—Architecture, e.g. interconnection topology
- G06N3/045—Combinations of networks
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
- G06N3/00—Computing arrangements based on biological models
- G06N3/02—Neural networks
- G06N3/08—Learning methods
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
- G06T3/00—Geometric image transformation in the plane of the image
- G06T3/40—Scaling the whole image or part thereof
- G06T3/4053—Super resolution, i.e. output image resolution higher than sensor resolution
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)
- Image Processing (AREA)
- Editing Of Facsimile Originals (AREA)
Abstract
The invention is applicable to the field of picture processing, and provides a super-resolution picture model generation method, a super-resolution picture model generation device, an electronic device and a storage medium, wherein the method comprises the following steps: inputting a low-resolution picture into a generator, wherein the generator generates a high-resolution picture from the low-resolution picture; inputting the generated high-resolution picture and the real high-resolution picture into a discriminator, wherein the discriminator is used for giving a discrimination result between the generated high-resolution picture and the real high-resolution picture; judging whether the judgment result is an optimal result; if not, continuing to input the low-resolution picture into the generator until the judgment result is the optimal result, and determining that the picture model is the optimal model. The method provided by the invention solves the problem that the existing super-resolution picture generation model is low in precision.
Description
Technical Field
The invention belongs to the technical field of software, relates to an image processing technology, and particularly relates to a super-resolution image model generation method and device, electronic equipment and a storage medium.
Background
The low-resolution picture contains fewer pixels, which is not beneficial for human eyes to observe the detail information of the picture, and the high-resolution picture has richer detail information, so that a picture super-resolution technology appears. The image super-resolution is to reconstruct a low-resolution image into a high-resolution image and restore the detail information lacking in the low-resolution image.
The single-picture super-resolution is that a single picture is reconstructed into a high-resolution picture from a low-resolution picture, and the traditional super-resolution picture reconstruction mostly adopts an example-based method and a dictionary storage method, wherein the external example-based method needs a large amount of sample data, and modeling data has limitations. With the research of the neural network on the super-resolution problem of a single picture, the SRCNN firstly applies the convolutional neural network to the super-resolution picture reconstruction, learns the mapping relation between LR and HR by using the convolutional network, and has great improvement in speed and quality compared with the traditional method. With the development of deep learning, some deeper network structures, such as a multi-scale resnet network, appear, compared with the SRCNN, the quality of a reconstructed picture is better, and the reconstructed picture has a clearer texture, but the problem of too smooth reconstructed picture appears. The generation countermeasure network (GAN) was proposed by GoodFollow in 2014, mainly for data augmentation, and widely used due to confidentiality and privacy of medical data and a small amount of data. And secondly, the method is also used for solving the problem of data imbalance in the tasks of action recognition and classification. In recent years, many scholars use a generation countermeasure network for super-resolution picture generation, and SRGAN reconstructs a super-resolution picture by using the generation countermeasure network, and adds a perceptual loss function to a generator objective function, thereby solving the problem of over-smoothness of the reconstructed picture and achieving a good effect.
Although the above SRGAN achieves better results, the structure is still in the gap with the best reconstructed super-resolution picture model.
Disclosure of Invention
The embodiment of the invention aims to provide a super-resolution image model generation method, and aims to solve the problem that the existing super-resolution image model is poor in reconstruction effect.
The embodiment of the invention is realized in such a way that a super-resolution picture model generation method comprises the following steps:
inputting a low-resolution picture into a generator, wherein the generator generates a high-resolution picture from the low-resolution picture;
inputting the generated high-resolution picture and the real high-resolution picture into a discriminator, wherein the discriminator is used for giving out a discrimination result between the generated high-resolution picture and the real high-resolution picture and feeding back the discrimination result to a generator;
judging whether the judgment result is an optimal result;
if not, continuing to input a low-resolution picture into the generator, and determining the picture model as an optimal model if the judgment result is an optimal result; if so, directly determining the picture model as an optimal model.
Further, when the generated high resolution picture and the real high resolution picture are inputted into the discriminator, the method further includes training a generation network model composed of the generator and the discriminator, the training step including:
training a discriminator network, inputting the generated high-resolution picture and the real high-resolution picture into a discriminator, and updating the discriminator network by the discriminator according to a target function;
and a training generator which inputs the low-resolution picture to the generator, generates the high-resolution picture, and updates the generator network according to the objective function.
Further, before inputting the low resolution picture into the generator, data preprocessing is further included, and the data preprocessing includes:
dividing a data set into a training set and a testing set;
processing the real high-resolution picture into the same pattern;
obtaining a corresponding low-resolution picture according to the high-resolution picture;
further, before training the picture model, a training environment is set up:
the experiment is completed by training on a windows10 operating system and a NVIDIA 1080Ti server, the version of CUDA is 10.2, the active program is written by using python language, the method is realized by adopting a pytorch framework, and the version of the pytorch is 1.4.
Further, before training the picture model, training detail setting is further included:
the optimized learning model method used for training is RMSProp (root mean square prop), the parameter is 0.9, the learning rate lr is 0.0005, and after a certain number of iterations reaches 3000 times in the training process, lr becomes 0.0001.
Another embodiment of the present invention also provides a super-resolution picture generating apparatus, including:
a generator for generating a high resolution picture from the input low resolution picture;
and the discriminator is used for giving out discrimination results of the generated high-resolution picture and the real high-resolution picture and feeding back the discrimination results to the generator.
The generator comprises a first part and a second part, wherein the first part adopts a convolution neural network, and the second part is a deconvolution neural network.
Further, the arbiter is further configured to update the arbiter network according to the objective function, and the generator is further configured to update the generator network according to the objective function.
Further, the present invention further includes a preprocessing module for preprocessing data, the preprocessing module includes: and dividing the data set into a training set and a testing set, processing the real high-resolution pictures in the data set into the same pattern, and obtaining corresponding low-resolution pictures according to the high-resolution pictures.
Another embodiment of the present invention further provides a computer-readable storage medium, having a computer program stored thereon, where the computer program, when executed by a processor, implements the steps in the super-resolution picture model generation method as described in any one of the above.
The invention has the beneficial effects that:
the invention relates to the field of super-resolution of pictures, in particular to a super-resolution picture model generation method, a super-resolution picture model generation device, electronic equipment and a storage medium, wherein a single low-resolution picture is reconstructed into a high-resolution picture;
in a first aspect, unlike conventional generation countermeasure networks, the method inputs a low resolution picture instead of the input gaussian noise into a generator that generates a reconstructed sharp high resolution picture. The generator adopts a convolution cyclic neural network, convolution operation not only can extract spatial features, but also can obtain a time sequence relation, namely, time features and spatial features can be simultaneously extracted, and the structure is more favorable for reconstructing a high-resolution picture;
in a second aspect, the invention inputs the generated high-resolution picture and the real high-resolution picture into a discriminator, and the discriminator gives out a discrimination result and feeds the result back to the generator. The discriminator structure finally adopts the convolution layer to replace the full-connection layer, thereby solving the problem that the picture size is not matched and the parameters need to be modified;
in the third aspect, the generator and the discriminator perform zero-sum game, and the optimal reconstruction model can obtain a clear super-resolution picture meeting the requirement.
Drawings
FIG. 1 is a flowchart of a super-resolution image model generation method according to an embodiment of the present invention;
FIG. 2 is a flow chart of a method for training a generator network according to an embodiment of the present invention;
FIG. 3 is an exemplary diagram of a wind capturing high resolution pictures of the same size according to an embodiment of the present invention;
FIG. 4 is a schematic diagram illustrating an example of obtaining low-resolution pictures with different scaling scales according to an embodiment of the present invention;
fig. 5 is a block diagram of an overall structure of a super-resolution image model generation method according to an embodiment of the present invention;
FIG. 6 is a block diagram of a generator architecture of an apparatus according to an embodiment of the present invention;
FIG. 7 is a block diagram of an embodiment of an apparatus for determining the structure of an arbiter;
fig. 8 is a block diagram of an electronic device according to an embodiment of the present invention.
Detailed Description
In order to make the objects, technical solutions and advantages of the present invention more apparent, the present invention is described in further detail below with reference to the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are merely illustrative of the invention and are not intended to limit the invention.
In the present invention, unless otherwise expressly specified or limited, the terms "mounted," "connected," "secured," and the like are to be construed broadly and can, for example, be fixedly connected, detachably connected, or integrally connected; can be mechanically or electrically connected; they may be connected directly or indirectly through intervening media, or they may be interconnected between two elements. The specific meanings of the above terms in the present invention can be understood by those skilled in the art according to specific situations. As used herein, the term "and/or" includes any and all combinations of one or more of the associated listed items.
The present invention provides
Specific implementations of the present invention are described in detail below with reference to specific examples.
Example one
Referring to fig. 1, a method for generating a super-resolution image model according to a first embodiment of the present invention includes steps S001 to S003:
step S001, inputting a low-resolution picture into a generator, wherein the generator generates a high-resolution picture from the low-resolution picture.
The generator is a generation model for generating an input low-resolution picture as a high-resolution picture. As shown in fig. 6, the generator network structure diagram includes a first part and a second part. The first part is used for extracting time features and space features of the low-resolution pictures, namely real-time space features, the first part is realized by adopting a convolution cyclic neural network, the convolution cyclic neural network is composed of convolution layers through a cyclic neural network mode and comprises 4 layers of residual error network modules, 6 convolution layers are arranged in each layer of residual error network, the size of a convolution kernel is 3 x 3, the step length is 1, and the activation function is an Lrelu function. The second part is used for amplifying a low-resolution picture into a high-resolution picture, a deconvolution network is adopted, deconvolution is also performed with convolution operation substantially, one layer of deconvolution can be amplified by 2 times, namely, the high-resolution picture reconstruction with the scaling factor of 2 is performed, the two layers of deconvolution can be performed with the high-resolution picture reconstruction with the scaling factor of 4, finally, the result is normalized to [ -1, 1] through an activating function tanh and then output, and for subsequent calculation, the result range is returned to [0, 1] finally.
Step S002, inputting the generated high resolution picture and the real high resolution picture into a discriminator, wherein the discriminator is used for giving a discrimination result between the generated high resolution picture and the real high resolution picture.
The discriminator network is essentially a two-class network that outputs the input generated high-resolution picture and the true high-resolution picture as values between (0, 1), discriminating whether the input image is true or generated.
Referring to table 1, the discriminator network is implemented by a convolutional neural network, and has 13 layers, which are: 10 convolutional layers, a global average pooling layer, a normalization layer and an output layer. The input layer of the convolutional neural network is used as the input layer of the discriminator network. Unlike conventional arbiter networks: the network finally adopts a convolution network to replace a fully connected network. By doing so, network parameters can be reduced, and the problem that parameters need to be modified due to different picture sizes is solved. The convolution kernel size of the first 8 layers of convolution layers is 3 x 3, the step size is 2 (zooming the picture), or the step size is 1 (keeping the picture size), the activation function is Lrelu function, the convolution kernel sizes of the last two layers are 1 x 1, the result is directly output after the operations, because the WGAN method is adopted for training, the last layer of Sigmoid activation function is removed, the judgment result is obtained, the generator is guided to train, and the network structure please refer to fig. 7.
Table 1 convolutional neural network of discriminators
| Layer | Operation in the discriminator |
| 1 | Input{hr,~hr} |
| 2 | {conv(64) + BN() + lrelu()} 2 |
| 4 | {conv(128) + BN() + lrelu()} 2 |
| 6 | {conv(256) + BN() + lrelu()} 2 |
| 8 | {conv(512) + BN() + lrelu()} 2 |
| 10 | {global_average_pooling()} |
| 11 | {conv(1024) + lrelu()} |
| 12 | {conv(1024)} |
| 13 | Output{out} |
And step S003, judging whether the judgment result is the optimal result, if so, performing step S004, otherwise, returning to step S001 to continue training.
In step S002, the generated high-resolution picture is output as a value between (0, 1), and then the corresponding true high-resolution picture is output as a value between (0, 1), and by combining the loss of the generator and the loss of the discriminator, when both performances are optimal, the picture generated by the generator is closest to the true high-resolution picture, and the picture model at this time is the optimal model.
Step S004: and determining the super-resolution picture generation model as an optimal model.
The invention has the beneficial effects that: the invention extracts the time characteristic and the space characteristic of the low-resolution picture through the first part of the generator, amplifies the low-resolution picture to the high-resolution picture through the second part of the generator, and finally outputs the low-resolution picture through the activation function. Enlarging the picture in the second portion helps to reduce training time and memory space. The discriminator is mainly used for discriminating whether the picture is a real picture or not, feeding the result back to the generator and guiding the generator to train, and the two game optimization models can obtain a vivid high-resolution picture sample.
Example two
In an embodiment of the present invention, referring to fig. 2, the method further includes step S100: training a generating network model consisting of a generator and a discriminator.
Further, the step S100 specifically includes:
and S101, inputting the generated high-resolution picture and the real high-resolution picture into a discriminator, and updating a discriminator network by the discriminator according to the target function.
And S102, inputting a low-resolution picture into the generator, generating a high-resolution picture by the generator, and updating the generator network according to the target function.
By the method, the discriminator network and the generator network are updated in the process of training the super-resolution picture model, and the generated network model can be continuously learned and trained so as to improve the accuracy of generating the super-resolution picture model.
EXAMPLE III
In another embodiment of the present invention, before the step S001, the method further includes the steps of:
step S0001, preprocessing the data, wherein the preprocessing comprises the following steps: and dividing the data set into a training set and a testing set, processing the real high-resolution pictures in the data set into the same pattern, and obtaining corresponding low-resolution pictures according to the high-resolution pictures.
The well-known existing data sets are all natural pictures, the data sets include VOC2012, set5, set14, BSD100 and the like, because the number of pictures in the VOC2012 is large, the VOC2012 is adopted as a training set, the data set has a corresponding test set, and in addition, the set5, set14 and BSD100 data sets are adopted as performance evaluation test sets in the invention.
Due to the fact that data sizes of the data sets are different, the data need to be screened, and pictures which do not meet the minimum width or height are deleted for subsequent operation.
As shown in fig. 3, in the BSD100 data set, although the pictures have the same size, the pictures have different patterns, and in order to meet the requirement of the input network format, the present invention rotates some pictures in the BSD100 data set, so that all the pictures have the same pattern. In addition, in VOC2012, set5 and set14 data sets, the pictures are different in size, and the pictures are cut into the same size by adopting a CenterCrop method in transform packages in a pyrrch frame, so that the training and the testing are facilitated.
According to the invention, a corresponding low-resolution picture is obtained according to a high-resolution picture, please refer to fig. 4, the set5, the set14 and the BSD100 data sets provide the low-resolution pictures with the corresponding scaling factors 2 and 4, and the VOC2012 only has the high-resolution picture.
By the mode, all high-resolution pictures are processed into the same pattern, and the pictures which do not meet the requirements are removed, so that the subsequent model training is facilitated, the calculation steps are reduced, and the training accuracy can be improved.
Example four
Another embodiment of the present invention further provides a super-resolution image generating apparatus, referring to fig. 6, the apparatus including:
a generator for generating a high resolution picture from the input low resolution picture;
and the discriminator is used for giving out discrimination results of the generated high-resolution picture and the real high-resolution picture and feeding back the discrimination results to the generator.
The generator comprises a first part and a second part, wherein the first part adopts a convolution neural network, and the second part is a deconvolution neural network.
Further, the arbiter is further configured to update the arbiter network according to the objective function, and the generator is further configured to update the generator network according to the objective function.
Further, the present invention further includes a preprocessing module for preprocessing data, the preprocessing module includes: and dividing the data set into a training set and a testing set, processing the real high-resolution pictures in the data set into the same pattern, and obtaining corresponding low-resolution pictures according to the high-resolution pictures.
EXAMPLE five
In order to solve the above technical problem, an embodiment of the present application further provides an electronic device, which is used for a super-resolution picture model. Referring to fig. 7 in detail, fig. 7 is a block diagram of a basic structure of the electronic device of the present embodiment, as shown in fig. 7.
The electronic device comprises a memory 701 and a processor 702. It is noted that only electronic device 14 having components 701 and 702 is shown, but it is understood that not all of the illustrated components are required and that more or fewer components may alternatively be implemented. As will be understood by those skilled in the art, the electronic device is a device capable of automatically performing numerical calculation and/or information processing according to a preset or stored instruction, and the hardware includes, but is not limited to, a microprocessor, an Application Specific Integrated Circuit (ASIC), a Programmable Gate Array (FPGA), a Digital Signal Processor (DSP), an embedded device, and the like.
The electronic device may be a desktop computer, a notebook, a palm computer, a cloud server, or other computing device. The electronic equipment can be in man-machine interaction with a user through a keyboard, a mouse, a remote controller, a touch panel or voice control equipment and the like.
The memory 701 is used for storing computer programs needed for the processor 702 to perform tasks.
The processor 702 may be a Central Processing Unit (CPU), controller, microcontroller, microprocessor, or other data Processing chip in some embodiments. The processor 702 is generally operative to control overall operation of the electronic device 14. In this embodiment, the processor 702 is configured to execute a computer program stored in the memory 701, for example, execute the instructions of the super-resolution image generation model method described above.
In other embodiments, the electronic device further comprises a communication interface 703 for enabling the subject to communicate with other electronic devices or a communication network.
EXAMPLE six
The present application further provides another embodiment, which is to provide a computer-readable storage medium storing a super-resolution picture model generation program, which is executable by at least one processor to cause the at least one processor to perform the steps of the super-resolution picture model generation method as described above.
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 device (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.
The above description is only for the purpose of illustrating the preferred embodiments of the present invention and is not to be construed as limiting the invention, and any modifications, equivalents and improvements made within the spirit and principle of the present invention are intended to be included within the scope of the present invention.
Claims (10)
1. A super-resolution picture model generation method is characterized by comprising the following steps:
inputting a low-resolution picture into a generator, wherein the generator generates a high-resolution picture from the low-resolution picture;
inputting the generated high-resolution picture and the real high-resolution picture into a discriminator, wherein the discriminator is used for giving a discrimination result between the generated high-resolution picture and the real high-resolution picture;
judging whether the judgment result is an optimal result;
if not, continuing to input the low-resolution pictures into the generator.
2. The super-resolution image model generation method of claim 1, further comprising determining that the image model is an optimal model if the determination result is an optimal result.
3. The super-resolution image model generation method of claim 1, wherein the generation of the high-resolution image from the low-resolution image by the generator specifically comprises: the time characteristic and the space characteristic of the low-resolution picture are extracted firstly, and then the low-resolution picture is enlarged into a high-resolution picture.
4. The super-resolution picture model generation method of claim 3, wherein when the generated high-resolution picture and the real high-resolution picture are input into the discriminator, the method further comprises: training a generating network model consisting of a generator and a discriminator.
5. The method of claim 4, wherein training the generative network model comprised of the generator and the arbiter comprises the steps of:
inputting the generated high-resolution picture and the real high-resolution picture into a discriminator, and updating a discriminator network by the discriminator according to a target function;
and inputting a low-resolution picture into the generator, generating a high-resolution picture by the generator, and updating the generator network according to the objective function.
6. The method of claim 5, wherein training the generating network model comprised of the generator and the arbiter further comprises: and preprocessing the data, wherein the preprocessing step comprises the steps of dividing a data set into a training set and a testing set, processing real high-resolution pictures in the data set into the same pattern, and obtaining corresponding low-resolution pictures according to the high-resolution pictures.
7. A super-resolution picture generation apparatus, characterized in that the apparatus comprises:
a generator for generating a high resolution picture from the input low resolution picture;
the discriminator is used for giving out discrimination results of the generated high-resolution picture and a real high-resolution picture and feeding back the discrimination results to the generator;
the generator comprises a first part and a second part, wherein the first part adopts a convolution neural network, and the second part is a deconvolution neural network.
8. The super-resolution picture generation apparatus of claim 7, wherein the discriminator is further configured to update a network of discriminators according to an objective function, and wherein the generator is further configured to update a network of generators according to the objective function.
9. An electronic device, comprising: memory, processor and computer program stored on the memory and executable on the processor, the processor implementing the steps in the super resolution picture model generation method according to any of claims 1 to 6 when executing the computer program.
10. A computer-readable storage medium, characterized in that the computer-readable storage medium has stored thereon a computer program which, when being executed by a processor, carries out the steps in the super-resolution picture model generation method according to any one of claims 1 to 6.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202011495417.1A CN112749788A (en) | 2020-12-17 | 2020-12-17 | Super-resolution picture model generation method and device, electronic equipment and storage medium |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202011495417.1A CN112749788A (en) | 2020-12-17 | 2020-12-17 | Super-resolution picture model generation method and device, electronic equipment and storage medium |
Publications (1)
| Publication Number | Publication Date |
|---|---|
| CN112749788A true CN112749788A (en) | 2021-05-04 |
Family
ID=75649287
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN202011495417.1A Pending CN112749788A (en) | 2020-12-17 | 2020-12-17 | Super-resolution picture model generation method and device, electronic equipment and storage medium |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN112749788A (en) |
Citations (7)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN109978762A (en) * | 2019-02-27 | 2019-07-05 | 南京信息工程大学 | A kind of super resolution ratio reconstruction method generating confrontation network based on condition |
| CN110189253A (en) * | 2019-04-16 | 2019-08-30 | 浙江工业大学 | A kind of image super-resolution rebuilding method generating confrontation network based on improvement |
| CN110472457A (en) * | 2018-05-10 | 2019-11-19 | 成都视观天下科技有限公司 | Low-resolution face image identification, restoring method, equipment and storage medium |
| CN110533588A (en) * | 2019-07-16 | 2019-12-03 | 中国农业大学 | Based on the root system image repair method for generating confrontation network |
| CN110930308A (en) * | 2019-11-15 | 2020-03-27 | 东南大学 | Structure searching method of image super-resolution generation network |
| CN111429355A (en) * | 2020-03-30 | 2020-07-17 | 新疆大学 | Image super-resolution reconstruction method based on generation countermeasure network |
| CN112001847A (en) * | 2020-08-28 | 2020-11-27 | 徐州工程学院 | Method for generating high-quality image by relatively generating antagonistic super-resolution reconstruction model |
-
2020
- 2020-12-17 CN CN202011495417.1A patent/CN112749788A/en active Pending
Patent Citations (7)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN110472457A (en) * | 2018-05-10 | 2019-11-19 | 成都视观天下科技有限公司 | Low-resolution face image identification, restoring method, equipment and storage medium |
| CN109978762A (en) * | 2019-02-27 | 2019-07-05 | 南京信息工程大学 | A kind of super resolution ratio reconstruction method generating confrontation network based on condition |
| CN110189253A (en) * | 2019-04-16 | 2019-08-30 | 浙江工业大学 | A kind of image super-resolution rebuilding method generating confrontation network based on improvement |
| CN110533588A (en) * | 2019-07-16 | 2019-12-03 | 中国农业大学 | Based on the root system image repair method for generating confrontation network |
| CN110930308A (en) * | 2019-11-15 | 2020-03-27 | 东南大学 | Structure searching method of image super-resolution generation network |
| CN111429355A (en) * | 2020-03-30 | 2020-07-17 | 新疆大学 | Image super-resolution reconstruction method based on generation countermeasure network |
| CN112001847A (en) * | 2020-08-28 | 2020-11-27 | 徐州工程学院 | Method for generating high-quality image by relatively generating antagonistic super-resolution reconstruction model |
Non-Patent Citations (1)
| Title |
|---|
| 张德丰: "TensorFlow深度学习从入门到进阶", 机械工业出版社, pages: 327 * |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| US10706333B2 (en) | Medical image analysis method, medical image analysis system and storage medium | |
| US11361192B2 (en) | Image classification method, computer device, and computer-readable storage medium | |
| US11302064B2 (en) | Method and apparatus for reconstructing three-dimensional model of human body, and storage medium | |
| CN108205803B (en) | Image processing method, and training method and device of neural network model | |
| US20210174072A1 (en) | Microexpression-based image recognition method and apparatus, and related device | |
| CN107145910A (en) | Performance generation system, its training method and the performance generation method of medical image | |
| US20210406579A1 (en) | Model training method, identification method, device, storage medium and program product | |
| CN111127320A (en) | Photoacoustic image super-resolution reconstruction method and device based on deep learning | |
| CN109712128A (en) | Feature point detecting method, device, computer equipment and storage medium | |
| CN108960301A (en) | A kind of ancient Yi nationality's text recognition methods based on convolutional neural networks | |
| CN112116720A (en) | Three-dimensional point cloud augmentation method and device, storage medium and computer equipment | |
| US20220292795A1 (en) | Face image processing method, electronic device, and storage medium | |
| Yang et al. | Multiscale mesh deformation component analysis with attention-based autoencoders | |
| CN113963148B (en) | Object detection method, object detection model training method and device | |
| CN110807497A (en) | Handwritten data classification method and system based on deep dynamic network | |
| DE102021124537A1 (en) | ENERGY-BASED VARIATIONAL AUTOENCODER | |
| CN112052843B (en) | Face key point detection method from coarse face to fine face | |
| Yanmin et al. | Research on ear recognition based on SSD_MobileNet_v1 network | |
| CN109241869A (en) | The recognition methods of answering card score, device and terminal device | |
| CN112749788A (en) | Super-resolution picture model generation method and device, electronic equipment and storage medium | |
| CN110728732A (en) | Image reconstruction method, device, equipment and medium | |
| CN103678888A (en) | Cardiac blood flowing indicating and displaying method based on Euler fluid simulation algorithm | |
| CN114299204B (en) | Three-dimensional cartoon character model generation method and device | |
| CN115409847A (en) | Lightweight segmentation method based on CT image | |
| CN115063847A (en) | Training method and device for facial image acquisition model |
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: 20210504 |