CN113989549A - Semi-supervised learning image classification optimization method and system based on pseudo labels - Google Patents

Semi-supervised learning image classification optimization method and system based on pseudo labels Download PDF

Info

Publication number
CN113989549A
CN113989549A CN202111227379.6A CN202111227379A CN113989549A CN 113989549 A CN113989549 A CN 113989549A CN 202111227379 A CN202111227379 A CN 202111227379A CN 113989549 A CN113989549 A CN 113989549A
Authority
CN
China
Prior art keywords
image classification
label
pseudo
data set
classification model
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
Application number
CN202111227379.6A
Other languages
Chinese (zh)
Inventor
陈英鹏
许野平
刘辰飞
张朝瑞
席道亮
高朋
Current Assignee (The listed assignees may be inaccurate. Google has not performed a legal analysis and makes no representation or warranty as to the accuracy of the list.)
Synthesis Electronic Technology Co Ltd
Original Assignee
Synthesis Electronic Technology Co Ltd
Priority date (The priority date 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 date listed.)
Filing date
Publication date
Application filed by Synthesis Electronic Technology Co Ltd filed Critical Synthesis Electronic Technology Co Ltd
Priority to CN202111227379.6A priority Critical patent/CN113989549A/en
Publication of CN113989549A publication Critical patent/CN113989549A/en
Pending legal-status Critical Current

Links

Images

Classifications

    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06F—ELECTRIC DIGITAL DATA PROCESSING
    • G06F18/00—Pattern recognition
    • G06F18/20—Analysing
    • G06F18/24—Classification techniques
    • G06F18/241—Classification techniques relating to the classification model, e.g. parametric or non-parametric approaches
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06F—ELECTRIC DIGITAL DATA PROCESSING
    • G06F18/00—Pattern recognition
    • G06F18/20—Analysing
    • G06F18/21—Design or setup of recognition systems or techniques; Extraction of features in feature space; Blind source separation
    • G06F18/214—Generating training patterns; Bootstrap methods, e.g. bagging or boosting

Landscapes

  • Engineering & Computer Science (AREA)
  • Data Mining & Analysis (AREA)
  • Theoretical Computer Science (AREA)
  • Computer Vision & Pattern Recognition (AREA)
  • Bioinformatics & Cheminformatics (AREA)
  • Bioinformatics & Computational Biology (AREA)
  • Artificial Intelligence (AREA)
  • Evolutionary Biology (AREA)
  • Evolutionary Computation (AREA)
  • Physics & Mathematics (AREA)
  • General Engineering & Computer Science (AREA)
  • General Physics & Mathematics (AREA)
  • Life Sciences & Earth Sciences (AREA)
  • Image Analysis (AREA)

Abstract

The invention discloses a semi-supervised learning image classification optimization method and system based on a pseudo label, which comprises the following steps: acquiring an image to be classified and preprocessing the image; recognizing the images by using the trained image classification model to obtain an image classification result; firstly, training an image classification model based on a labeled image sample data set to obtain an initial image classification model; obtaining a prediction set corresponding to the image sample data set without the label by using the initial image classification model, and generating a pseudo label data set based on the prediction set; and training the initial image classification model by using the image sample data set without the label and the pseudo label data set to obtain a trained image classification model. The image classification is carried out by the semi-supervised learning method based on the pseudo label, and the pseudo label can be automatically marked on the image sample data without the label under the condition that the label image sample data is limited.

Description

Semi-supervised learning image classification optimization method and system based on pseudo labels
Technical Field
The invention relates to the technical field of image classification, in particular to a semi-supervised learning image classification optimization method and system based on a pseudo label.
Background
The statements in this section merely provide background information related to the present disclosure and may not necessarily constitute prior art.
In recent years, convolutional neural networks have made a major breakthrough in image classification technology, and the design of convolutional neural networks has become more and more complex. The prediction precision of the convolutional neural network is closely related to the size and the accuracy of the sample data set; generally, the larger the label sample set is, the better the model performance obtained by training is, and the greater the labor cost consumption corresponding to the label data set is.
In a real application scene, limited by the limitation of tag data, the semi-supervised learning mode is more and more concerned. By constructing a loss function suitable for the unlabeled samples, the semi-supervised learning enables only a small number of labeled samples and certain unlabeled samples to be needed in the training process of the model, and the requirement on labeled samples can be reduced.
In the prior art, on the premise of limited label data (for example, taking identification of dress of factory staff as an example, in such a scene, generally, collected image samples of dress of staff are limited, and a large amount of manpower is needed for marking the image sample data), a pseudo label is generated from the data without the label; by using the pseudo label mode, the target domain picture can be better utilized, and the inter-domain difference is reduced, so that the network performance is improved; however, the validity problem of the pseudo label is not considered, and actually, a part of the pseudo labels predicted according to the training model often have false labels, so that the model training is wrong, and the accuracy is reduced.
Disclosure of Invention
In order to solve the problems, the invention provides a pseudo label-based semi-supervised learning image classification optimization method and system, which aim at generating a pseudo label for unlabelled image sample data and simultaneously limit the loss weight of the pseudo label under the condition that the pseudo label has a wrong label, so that the network classification performance is better while manpower marking is saved.
In some embodiments, the following technical scheme is adopted:
a semi-supervised learning image classification optimization method based on pseudo labels comprises the following steps:
acquiring an image to be classified and preprocessing the image;
recognizing the images by using the trained image classification model to obtain an image classification result;
firstly, training an image classification model based on a labeled image sample data set to obtain an initial image classification model; obtaining a prediction set corresponding to the image sample data set without the label by using the initial image classification model, and generating a pseudo label data set based on the prediction set; and training the initial image classification model by using the image sample data set without the label and the pseudo label data set to obtain a trained image classification model.
As a further scheme, generating a pseudo tag data set based on the prediction set specifically includes:
and recording the element value with the maximum numerical value in the prediction set as 1, and recording the rest elements as 0 to obtain a pseudo label data set of the non-label data set.
As a further scheme, when the initial image classification model is trained by using an image sample data set without a label and a pseudo label data set, a loss function is designed into a logarithm function and an exponential function, wherein the loss function is designed into a logarithm function and an exponential function through a parameter t1Constraining the bounds of the logarithmic function; by the parameter t2Limiting the decay rate of the exponential function.
As a further scheme, the training of the initial image classification model by using an image sample data set without a label and a pseudo label data set specifically includes:
initializing network weight, and keeping a verification set of an initial image classification model unchanged;
inputting an image sample data set without a label and a corresponding pseudo label data set into an initial image classification model, and controlling a loss function in a training process through set parameters;
normalizing the predicted value corresponding to each image sample data without a label;
a trained loss function is obtained.
As a further scheme, the loss function in the training process is controlled by the set parameters, specifically: designing the loss function into a logarithmic function and an exponential function, wherein the parameter t is passed1Constraining the bounds of the logarithmic function; by the parameter t2Limiting the decay rate of the exponential function.
As a further scheme, normalizing the predicted value corresponding to each unlabeled image sample data specifically includes:
obtaining the maximum value of all predicted values, making a difference between the predicted value corresponding to each unlabeled image sample data and the maximum value, and iterating each difference value according to a set rule to obtain a final iteration data set;
and respectively substituting the final iteration data into the exponential function and summing, substituting the reciprocal of the sum into the logarithmic function, and making a difference between the obtained result and the maximum value to obtain a predicted value after normalization.
As a further scheme, the loss function after training is obtained, specifically:
and taking the pseudo label data as a true value of image sample data without a label, and constructing a loss function after training by using the true value and a predicted value after normalization.
In other embodiments, the following technical solutions are adopted:
a semi-supervised learning image classification optimizing system based on pseudo labels comprises:
the image acquisition module is used for acquiring an image to be classified;
the image classification module is used for identifying the images by utilizing the trained image classification model to obtain an image classification result;
firstly, training an image classification model based on a labeled image sample data set to obtain an initial image classification model; obtaining a prediction set corresponding to the image sample data set without the label by using the initial image classification model, and generating a pseudo label data set based on the prediction set; and training the initial image classification model by using the image sample data set without the label and the pseudo label data set to obtain a trained image classification model.
In other embodiments, the following technical solutions are adopted:
a terminal device comprising a processor and a memory, the processor being arranged to implement instructions; the memory is for storing a plurality of instructions adapted to be loaded and executed by the processor in a pseudo-label based semi-supervised learning image classification optimization method.
In other embodiments, the following technical solutions are adopted:
a computer readable storage medium, wherein a plurality of instructions are stored, the instructions are suitable for being loaded by a processor of a terminal device and executing the pseudo label based semi-supervised learning image classification optimization method.
Compared with the prior art, the invention has the beneficial effects that:
(1) the image classification is carried out by the semi-supervised learning method based on the pseudo label, and the pseudo label can be automatically marked on the image sample data without the label under the condition that the label image sample data is limited.
(2) According to the invention, the loss function is designed into two parts, the boundary constraint and the tail attenuation limitation are further utilized to optimize aiming at the error label in the pseudo label, the loss weight is limited, the influence brought by the error label can be well controlled, and the network performance is further improved.
Additional features and advantages of the invention will be set forth in part in the description which follows, and in part will be obvious from the description, or may be learned by practice of the invention.
Drawings
Fig. 1 is a schematic diagram of a training process of an image classification model in an embodiment of the present invention.
Detailed Description
It should be noted that the following detailed description is exemplary and is intended to provide further explanation of the disclosure. Unless defined otherwise, all technical and scientific terms used herein have the same meaning as commonly understood by one of ordinary skill in the art to which this application belongs.
It is noted that the terminology used herein is for the purpose of describing particular embodiments only and is not intended to be limiting of example embodiments according to the present application. As used herein, the singular forms "a", "an" and "the" are intended to include the plural forms as well, and it should be understood that when the terms "comprises" and/or "comprising" are used in this specification, they specify the presence of stated features, steps, operations, devices, components, and/or combinations thereof, unless the context clearly indicates otherwise.
Example one
In one or more embodiments, a pseudo label-based semi-supervised learning image classification optimization method is disclosed, and specifically includes the following processes:
acquiring an image to be classified, carrying out preprocessing operations such as normalization and the like, and identifying the image by using a trained image classification model to obtain an image classification result; referring to fig. 1, the training process of the image classification model specifically includes:
s101: firstly, training an image classification model based on a labeled image sample data set to obtain an initial image classification model;
assume a tagged dataset as SlClass number C, unlabeled dataset Su(ii) a Will have label data SlDivided into training set SltAnd a verification set SlvAnd training an image classification model to obtain an initial image classification model M1。
S102: obtaining a prediction set corresponding to the image sample data set without the label by using the initial image classification model, and generating a pseudo label data set based on the prediction set;
will not have the label data set SuInput to an initial image classification modelM1In (3), obtain the corresponding prediction set Pu(ii) a Wherein, Pui＝{pu0,pu1,…,pu(c-1)}。
Using prediction set PuGenerating a label-free dataset SuPseudo tag data set PlIn which P isliI.e., P is taken as {0,1, …,0}uiThe maximum value in (1) and the rest are all 0, the symbol function can be used to represent:
Figure BDA0003314783540000061
s103: and training the initial image classification model by using the image sample data set without the label and the pseudo label data set to obtain a trained image classification model.
In this embodiment, the process of training the initial image classification model specifically includes:
s1031: initializing the weight of the initial image classification model and keeping a verification set SlvThe change is not changed;
s1032: image sample data set S without labeluAnd a corresponding pseudo tag data set PlInputting the image data into an initial image classification model for training;
s1033: in the training process, the loss function is designed into a logarithmic function and an exponential function, wherein the parameter t is passed1Constraining the bounds of the logarithmic function; by the parameter t2Limiting the decay rate of the exponential function. Therefore, the influence caused by the error label can be well controlled, and the network performance is further improved.
In particular, the parameter t1Is a parameter between 0 and 1, and the smaller the value of the parameter, the more the constraint on the limit of the logarithmic function is, and the specific logarithmic function is:
Figure BDA0003314783540000062
parameter t2Is a ginseng of 1 or moreThe larger the value of the exponential function is, the thicker the tail of the exponential function is, and the slower the attenuation is, and the specific exponential function is:
Figure BDA0003314783540000071
s1034: suppose a certain unlabeled data SuiThe corresponding predicted value is PuiThen need to be applied to PuiThe values were normalized by the following specific steps:
(1) obtaining PuiMaximum of (d), i.e. mu ═ Max (P)ui)；
(2) Each P isuiBy difference with said maximum, i.e. norm0 ═ Pui-mu；
(3) Further, norm0 is initialized with j 0, and N iterations are performed according to the following procedure, where N is a set positive integer value.
j+＝1 ①
temp＝Sum(F2(norm,t2)) ②
Figure BDA0003314783540000074
(4) The predicted value after normalization is obtained as:
Figure BDA0003314783540000072
where Max is the maximum and Sum is the Sum.
S1035: taking the pseudo label data as a true value of image sample data without a label, and constructing a loss function after training by using the true value and a predicted value after normalization; the loss function minimizes the impact of noise (false labeling) on the training results by limiting the boundaries of the predicted values.
In this embodiment, let:
temp1＝F1(y,t1)-F1(Norm(y^),t1)*y
Figure BDA0003314783540000073
L＝temp1-temp2
wherein L is the loss function after training, y is the true value, and y ^ is the predicted value.
And finally, obtaining a trained image classification model, and inputting the image data to be classified into the trained image classification model to obtain an image classification result.
In some embodiments, unlabeled image sample data is input into a trained image classification model, so that a prediction result of the unlabeled image sample can be directly obtained, and thus, automatic labeling of the unlabeled image sample can be realized.
In the embodiment, for example, the dressing identification of the factory staff is taken as an example, in such a scene, the samples capable of being collected are limited, and a large amount of manpower is also required for marking data; by utilizing the semi-supervised learning image classification optimization method based on the pseudo label, training an image classification model by utilizing the sample images of the staffs wearing the labels and the staffs wearing the labels to obtain the trained image classification model; obtaining an image of the clothing of the employee, inputting the image into a trained image classification model, and obtaining a classification result of the clothing of the employee, such as: wearing or not wearing the tool.
Example two
In one or more embodiments, disclosed is a pseudo-label based semi-supervised learning image classification optimization system, comprising:
the image acquisition module is used for acquiring an image to be classified;
the image classification module is used for identifying the images by utilizing the trained image classification model to obtain an image classification result;
firstly, training an image classification model based on a labeled image sample data set to obtain an initial image classification model; obtaining a prediction set corresponding to the image sample data set without the label by using the initial image classification model, and generating a pseudo label data set based on the prediction set; and training the initial image classification model by using the image sample data set without the label and the pseudo label data set to obtain a trained image classification model.
It should be noted that, the specific implementation process of each module described above has been described in the first embodiment, and is not described in detail here.
EXAMPLE III
In one or more embodiments, a terminal device is disclosed, which includes a server including a memory, a processor, and a computer program stored on the memory and executable on the processor, and the processor implements the pseudo-label based semi-supervised learning image classification optimization method in the first embodiment when executing the program. For brevity, no further description is provided herein.
It should be understood that in this embodiment, the processor may be a central processing unit CPU, and the processor may also be other general purpose processors, digital signal processors DSP, application specific integrated circuits ASIC, off-the-shelf programmable gate arrays FPGA or other programmable logic devices, discrete gate or transistor logic devices, discrete hardware components, and so on. A general purpose processor may be a microprocessor or the processor may be any conventional processor or the like.
The memory may include both read-only memory and random access memory, and may provide instructions and data to the processor, and a portion of the memory may also include non-volatile random access memory. For example, the memory may also store device type information.
In implementation, the steps of the above method may be performed by integrated logic circuits of hardware in a processor or instructions in the form of software.
Example four
In one or more implementations, a computer-readable storage medium is disclosed, in which a plurality of instructions are stored, the instructions being adapted to be loaded by a processor of a terminal device and to execute the pseudo-label based semi-supervised learning image classification optimization method described in the first embodiment.
Although the embodiments of the present invention have been described with reference to the accompanying drawings, it is not intended to limit the scope of the present invention, and it should be understood by those skilled in the art that various modifications and variations can be made without inventive efforts by those skilled in the art based on the technical solution of the present invention.

Claims (10)

1. A semi-supervised learning image classification optimization method based on pseudo labels is characterized by comprising the following steps:
acquiring an image to be classified and preprocessing the image;
recognizing the images by using the trained image classification model to obtain an image classification result;
firstly, training an image classification model based on a labeled image sample data set to obtain an initial image classification model; obtaining a prediction set corresponding to the image sample data set without the label by using the initial image classification model, and generating a pseudo label data set based on the prediction set; and training the initial image classification model by using the image sample data set without the label and the pseudo label data set to obtain a trained image classification model.
2. The pseudo-label-based semi-supervised learning image classification optimization method according to claim 1, wherein generating a pseudo-label data set based on the prediction set specifically comprises:
and recording the element value with the maximum numerical value in the prediction set as 1, and recording the rest elements as 0 to obtain a pseudo label data set of the non-label data set.
3. The pseudo-label-based semi-supervised learning image classification optimization method of claim 1, wherein when an initial image classification model is trained by using an unlabeled image sample data set and a pseudo-label data set, a loss function is designed into a logarithmic function and an exponential function, wherein a parameter t is used for designing the loss function1Constraining the bounds of the logarithmic function; by the parameter t2Limiting the decay rate of the exponential function.
4. The pseudo-label-based semi-supervised learning image classification optimization method of claim 1, wherein an initial image classification model is trained by using an unlabeled image sample data set and a pseudo-label data set, and specifically comprises:
initializing network weight, and keeping a verification set of an initial image classification model unchanged;
inputting an image sample data set without a label and a corresponding pseudo label data set into an initial image classification model, and controlling a loss function in a training process through set parameters;
normalizing the predicted value corresponding to each image sample data without a label;
a trained loss function is obtained.
5. The pseudo-label-based semi-supervised learning image classification optimization method of claim 4, wherein a loss function in a training process is controlled through set parameters, specifically: designing the loss function into a logarithmic function and an exponential function, wherein the parameter t is passed1Constraining the bounds of the logarithmic function; by the parameter t2Limiting the decay rate of the exponential function.
6. The pseudo-label-based semi-supervised learning image classification optimization method according to claim 5, wherein the normalization of the predicted value corresponding to each unlabeled image sample data specifically comprises:
obtaining the maximum value of all predicted values, making a difference between the predicted value corresponding to each unlabeled image sample data and the maximum value, and iterating each obtained difference value according to a set rule to obtain a final iteration data set;
and respectively substituting the final iteration data into the exponential function and summing, substituting the reciprocal of the sum into the logarithmic function, and making a difference between the obtained result and the maximum value to obtain a predicted value after normalization.
7. The pseudo-label-based semi-supervised learning image classification optimization method of claim 4, wherein a trained loss function is obtained, and specifically comprises:
and taking the pseudo label data as a true value of image sample data without a label, and constructing a loss function after training by using the true value and a predicted value after normalization.
8. A semi-supervised learning image classification optimization system based on pseudo labels is characterized by comprising the following steps:
the image acquisition module is used for acquiring an image to be classified;
the image classification module is used for identifying the images by utilizing the trained image classification model to obtain an image classification result;
firstly, training an image classification model based on a labeled image sample data set to obtain an initial image classification model; obtaining a prediction set corresponding to the image sample data set without the label by using the initial image classification model, and generating a pseudo label data set based on the prediction set; and training the initial image classification model by using the image sample data set without the label and the pseudo label data set to obtain a trained image classification model.
9. A terminal device comprising a processor and a memory, the processor being arranged to implement instructions; the memory is configured to store a plurality of instructions adapted to be loaded by the processor and to perform the pseudo-label based semi-supervised learning image classification optimization method of any one of claims 1 to 7.
10. A computer-readable storage medium having stored thereon a plurality of instructions adapted to be loaded by a processor of a terminal device and to execute the pseudo-label based semi-supervised learning image classification optimization method of any one of claims 1-7.
CN202111227379.6A 2021-10-21 2021-10-21 Semi-supervised learning image classification optimization method and system based on pseudo labels Pending CN113989549A (en)

Priority Applications (1)

Application Number Priority Date Filing Date Title
CN202111227379.6A CN113989549A (en) 2021-10-21 2021-10-21 Semi-supervised learning image classification optimization method and system based on pseudo labels

Applications Claiming Priority (1)

Application Number Priority Date Filing Date Title
CN202111227379.6A CN113989549A (en) 2021-10-21 2021-10-21 Semi-supervised learning image classification optimization method and system based on pseudo labels

Publications (1)

Publication Number Publication Date
CN113989549A true CN113989549A (en) 2022-01-28

Family

ID=79739973

Family Applications (1)

Application Number Title Priority Date Filing Date
CN202111227379.6A Pending CN113989549A (en) 2021-10-21 2021-10-21 Semi-supervised learning image classification optimization method and system based on pseudo labels

Country Status (1)

Country Link
CN (1) CN113989549A (en)

Cited By (4)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN115272777A (en) * 2022-09-26 2022-11-01 山东大学 Semi-supervised image analysis method for power transmission scene
CN115471717A (en) * 2022-09-20 2022-12-13 北京百度网讯科技有限公司 Model semi-supervised training and classification method and device, equipment, medium and product
CN116050428A (en) * 2023-03-07 2023-05-02 腾讯科技（深圳）有限公司 Intention recognition method, device, equipment and storage medium
CN117372819A (en) * 2023-12-07 2024-01-09 神思电子技术股份有限公司 Target detection increment learning method, device and medium for limited model space

Cited By (6)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN115471717A (en) * 2022-09-20 2022-12-13 北京百度网讯科技有限公司 Model semi-supervised training and classification method and device, equipment, medium and product
CN115272777A (en) * 2022-09-26 2022-11-01 山东大学 Semi-supervised image analysis method for power transmission scene
CN115272777B (en) * 2022-09-26 2022-12-23 山东大学 Semi-supervised image analysis method for power transmission scene
CN116050428A (en) * 2023-03-07 2023-05-02 腾讯科技（深圳）有限公司 Intention recognition method, device, equipment and storage medium
CN117372819A (en) * 2023-12-07 2024-01-09 神思电子技术股份有限公司 Target detection increment learning method, device and medium for limited model space
CN117372819B (en) * 2023-12-07 2024-02-20 神思电子技术股份有限公司 Target detection increment learning method, device and medium for limited model space

Similar Documents

Publication Publication Date Title
CN109741332B (en) Man-machine cooperative image segmentation and annotation method
CN113989549A (en) Semi-supervised learning image classification optimization method and system based on pseudo labels
CN111985229B (en) Sequence labeling method and device and computer equipment
CN110069709B (en) Intention recognition method, device, computer readable medium and electronic equipment
CN112132179A (en) Incremental learning method and system based on small number of labeled samples
CN110363049B (en) Method and device for detecting, identifying and determining categories of graphic elements
CN111127364B (en) Image data enhancement strategy selection method and face recognition image data enhancement method
CN113128478B (en) Model training method, pedestrian analysis method, device, equipment and storage medium
CN111079847B (en) Remote sensing image automatic labeling method based on deep learning
CN113449821B (en) Intelligent training method, device, equipment and medium fusing semantics and image characteristics
CN107797989A (en) Enterprise name recognition methods, electronic equipment and computer-readable recording medium
CN109766435A (en) The recognition methods of barrage classification, device, equipment and storage medium
CN115080749B (en) Weak supervision text classification method, system and device based on self-supervision training
CN114330588A (en) Picture classification method, picture classification model training method and related device
CN116245097A (en) Method for training entity recognition model, entity recognition method and corresponding device
CN109359664A (en) The efficient Checking model construction method and system of self-teaching update training sample
CN111783688A (en) Remote sensing image scene classification method based on convolutional neural network
CN116342906A (en) Cross-domain small sample image recognition method and system
CN115080748B (en) Weak supervision text classification method and device based on learning with noise label
CN116681961A (en) Weak supervision target detection method based on semi-supervision method and noise processing
CN112347957A (en) Pedestrian re-identification method and device, computer equipment and storage medium
CN111091198A (en) Data processing method and device
CN113886602B (en) Domain knowledge base entity identification method based on multi-granularity cognition
CN113377884B (en) Event corpus purification method based on multi-agent reinforcement learning
CN113240565B (en) Target identification method, device, equipment and storage medium based on quantization 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