CN105117429A - Scenario image annotation method based on active learning and multi-label multi-instance learning - Google Patents
Scenario image annotation method based on active learning and multi-label multi-instance learning Download PDFInfo
- Publication number
- CN105117429A CN105117429A CN201510473322.2A CN201510473322A CN105117429A CN 105117429 A CN105117429 A CN 105117429A CN 201510473322 A CN201510473322 A CN 201510473322A CN 105117429 A CN105117429 A CN 105117429A
- Authority
- CN
- China
- Prior art keywords
- label
- scene image
- image
- many examples
- individual
- 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.)
- Granted
Links
Classifications
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F16/00—Information retrieval; Database structures therefor; File system structures therefor
- G06F16/50—Information retrieval; Database structures therefor; File system structures therefor of still image data
- G06F16/58—Retrieval characterised by using metadata, e.g. metadata not derived from the content or metadata generated manually
- G06F16/5866—Retrieval characterised by using metadata, e.g. metadata not derived from the content or metadata generated manually using information manually generated, e.g. tags, keywords, comments, manually generated location and time information
-
- 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
- G06F18/2411—Classification techniques relating to the classification model, e.g. parametric or non-parametric approaches based on the proximity to a decision surface, e.g. support vector machines
Landscapes
- Engineering & Computer Science (AREA)
- Theoretical Computer Science (AREA)
- Data Mining & Analysis (AREA)
- General Physics & Mathematics (AREA)
- Physics & Mathematics (AREA)
- General Engineering & Computer Science (AREA)
- Life Sciences & Earth Sciences (AREA)
- Databases & Information Systems (AREA)
- Library & Information Science (AREA)
- Artificial Intelligence (AREA)
- Bioinformatics & Cheminformatics (AREA)
- Bioinformatics & Computational Biology (AREA)
- Computer Vision & Pattern Recognition (AREA)
- Evolutionary Biology (AREA)
- Evolutionary Computation (AREA)
- Image Analysis (AREA)
Abstract
The present invention is directed to two fundamental characteristics of a scene image: (1) the scene image often containing complex semantics; and (2) a great number of manual annotation images taking high labor cost. The invention further discloses a scene image annotation method based on an active learning and a multi-label and multi-instance learning. The method comprises: training an initial classification model on the basis of a label image; predicting a label to an unlabeled image; calculating a confidence of the classification model; selecting an unlabeled image with the greatest uncertainty; experts carrying on a manual annotation on the image; updating an image set; and stopping when an algorithm meeting the requirements. An active learning strategy utilized by the method ensures accuracy of the classification model, and significantly reduces the quantity of the scenario image needed to be manually annotated, thereby decreasing the annotation cost. Moreover, according to the method, the image is converted to a multi-label and multi-instance data, complex semantics of the image has a reasonable demonstration, and accuracy of image annotation is improved.
Description
Technical field
The present invention relates to scene image label technology field, particularly relate to a kind of scene image mask method based on Active Learning and many labels multi-instance learning.
background technology:
Along with the development of infotech and the progress of Internet service, all kinds of websites such as news, social activity and commodity transaction obtain significant progress, and internet all produces the scene picture of magnanimity every day.These scene pictures have following two basic characteristics.On the one hand, single width scene image not only reflects a content, may relate to multiple theme, semantic more complicated.Such as, a pair, about the image in street, may relate to multiple different themes such as pedestrian, road, vehicle, trees, sky, buildings.
On the other hand, a large amount of scene images that internet produces, do not have the tag along sort that fully can describe image content.For example, user may upload a picture with scenes at social networks, but the text description that photo content is not detailed.Semantic complicated for these, and do not possess the magnanimity scene image of tag along sort, how to utilize these pictures, for Internet user provides relevant service, this is the core missions of scene image mark.The object of scene image mark is, by there being the study of label scene image, is given accurate tag along sort without label scene image, enables them provide service for Internet user.
Traditional image labeling method is having some limitations property in Internet scene image labeling.First, traditional image labeling method regards single vector as piece image.As mentioned above, a secondary scene image may comprise several themes, if piece image is converted into single vector, may the semanteme of accurate description scene image, and also cannot accurately mark scene image.Secondly, traditional image labeling method needs a large amount of label scene images that has to carry out learning classification model.In order to set up the disaggregated model of pinpoint accuracy, often needing expert to pass through artificial notation methods, marking a considerable amount of scene image and carrying out train classification models.The scene image that artificial mark is a large amount of, needs to expend huge human and material resources.Therefore, a kind of based on there being the efficient automatic scene image labeling technology of label image urgently to propose on a small quantity.
Summary of the invention
The object of the invention is to solve two basic characteristics for scene image, scene image may comprise multiple content area, semantic complicated, it is converted into single vector-quantities cannot Precise Representation scene image theme, and a large amount of scene pictures of internet do not possess tag along sort, a kind of scene image mask method based on many examples Multi-label learning and Active Learning of problem such as mark cost intensive etc.
To achieve these goals, present invention employs following technical scheme:
Based on the scene image mask method of Active Learning and many labels multi-instance learning, comprise the steps,
(1) a collection of scene image without label is obtained.Randomly draw a small amount of scene image, by the artificial notation methods of expert, give these scene image tag along sorts;
(2) having label scene image and being converted into many sample datas without label scene image, every width image regards example bag more than as, and an example of many examples bag is regarded in each region as;
(3) there being label scene image to regard training set as on a small quantity, according to the number of labels of scene image, several preliminary classification models are trained;
(4) utilize the disaggregated model set up, to marking without label scene image in sample set, each image may have multiple label;
(5) according to the annotation results without label scene image, the confidence level of each disaggregated model is calculated;
(6) confidence level of combining classification model, from without selecting the maximum image of a uncertainty label scene image, and gives expert and marks this scene image;
(7) scene image marked through expert being removed from concentrating without label image data, being placed with label scene image data collection, and train classification models again;
(8) judge whether the degree of accuracy of this model reaches the degree of accuracy required by user, or whether iteration wheel number reaches the number of times that user specifies, if do not reach requirement, returns (3); Otherwise terminate and output category model.
The present invention utilizes active learning strategies, while guarantee disaggregated model degree of accuracy, greatly reduces the scene image quantity needing artificial mark, thus reduces mark cost.Meanwhile, the present invention is converted into the many sample datas of many labels image, makes the complicated semanteme of image obtain reasonable representation, improves the degree of accuracy of image labeling.
Accompanying drawing explanation
Fig. 1 is the process flow diagram of the training marking model of the embodiment of the present invention.
Embodiment
Below in conjunction with specific embodiment, illustrate the present invention further, these embodiments should be understood only be not used in for illustration of the present invention and limit the scope of the invention, after having read the present invention, the amendment of those skilled in the art to the various equivalent form of value of the present invention has all fallen within the application's claims limited range.
Fig. 1 is the process flow diagram of the scene image mask method model based on Active Learning and many labels multi-instance learning of the embodiment of the present invention.As shown in Figure 1, the scene image mask method that the present invention relates to comprises following process:
The first step, obtains a collection of scene image without label.Randomly draw a small amount of scene image, by the artificial notation methods of expert, give these scene image tag along sorts.Because a secondary scene image may comprise different contents, relate to multiple theme, therefore piece image may have several tag along sorts.In image collection, suppose that the maximum number of tag along sort is k.By above-mentioned steps, scene image set is originally divided into two set again, and a set comprises label scene image on a small quantity, and another one set comprises remaining a large amount of without label scene image.
Second step, having label scene image and being converted into many sample datas without label scene image.Because scene image may relate to multiple theme, semantic complicated, if a secondary scene image is converted into single vector, the complexity being difficult to Description Image is exactly semantic.Therefore, need scene image to be converted into many sample datas.Specifically, the classical way of field of image recognition can be used, as BlobworldSystem etc., image be cut into several regions according to different contents.Then, each image-region is extracted to the features such as color, texture, shape, an image-region is converted into an example vector.In this way, a sub-picture has been cut into several regions.One sub-picture regards example bag more than as, and the example of many examples bag is regarded in a region as.
3rd step, there being label scene image to regard training set as on a small quantity, according to k tag along sort of scene image, trains k preliminary classification model.For each tag along sort, the image with this label is regarded as positive class data, the image without this label is regarded as negative class data, train initial many example classification model.
4th step, utilizes k the disaggregated model set up, predicts the label without label scene image.Through k disaggregated model, each pair will obtain k tag along sort without label scene image.For i-th disaggregated model, if the value of tag along sort is 1, represent that this scene image comprises the picture material of the i-th class; If the value of tag along sort is 0, represent that this scene image does not comprise the picture material of the i-th class.
5th step, according to the annotation results without label scene image, calculates the confidence level of each disaggregated model.With reference to transductive SVM (TransductiveSupportVectorMachine, TSVM) thought, given one group independent identically distributed have the training sample of label and another group from same distribution without exemplar, when sample is abundant, can the corresponding ratio estimated without exemplar positive in exemplar according to the positive exemplar proportion had in exemplar.For this reason, should be close with the ratio shared by the positive exemplar had in exemplar without exemplar proportion positive in exemplar.Based on this thought, propose the criterion of a kind of disaggregated model to prediction label confidence level, first utilize and have the training of label many examples bag
kindividual sorter, recycling obtains
kindividual sorter is classified to without label many examples bag, obtains its prediction label.Assuming that
xrepresent instance space,
yrepresent tally set space, given
n l individual have label many examples bag
with
n u individual without label many examples bag
.Target is that study obtains objective function
f mIML : 2 x →
2 y .Wherein,
a corresponding example collection,
,
for
x i corresponding one group of tag set
y i1 ,
y i2 ...,
y il ,
y ik =0,1} (
k=1,2 ...,
l), here,
n i represent many examples bag
x i in containing the number of example,
lrepresent the label number in many examples bag.On this basis,
kthe confidence level of individual disaggregated model
c k can be defined as:
In above formula,
i [] be an indicator function (indicatorfunction), satisfied [] specified criteria then its value is 1, otherwise value is 0;
y l ik represent the
kin individual sorter
ithe individual label having label many examples bag,
y u ik represent the
kin individual sorter
ithe individual label without label many examples bag.
indicate and wrap in without the many examples of label
kthe mean value of the positive label predicted in individual sorter,
be shown with the many examples of label and wrap in
kthe mean value of positive label in individual sorter.Therefore, confidence level
c k less, illustrate without exemplar proportion positive in label many examples bag and have the ratio shared by label many examples bag more not close, namely confidence level is lower, otherwise then confidence level is higher.
6th step, according to
minimum classification distance selection strategy,combining classification model credibility, from without selecting the maximum image of a uncertainty label scene image, and gives expert and marks this image.It is generally acknowledged, sample distance lineoid is more closely larger by the possibility of misclassification, and uncertain larger, the quantity of information that sample packages contains is also more, and also namely sample is more valuable.Therefore, by calculating the distance of many examples bag distance lineoid, and considering that the confidence level of disaggregated model to many examples bag is weighed as one, proposing minimum classification distance strategy.For this reason, first define the minor increment of many examples bag and lineoid, as follows:
In above formula,
f k (X ij )represent many examples bag
x i in
jindividual example is
kthe classification function output valve of individual SVM classifier,
represent example
x ij for
kthe lineoid distance of individual SVM classifier.
represent many examples bag X
imiddle distance
kindividual SVM classifier lineoid example farthest, according to the definition of multi-instance learning, at least containing a positive example in each positive closure, and distance classification plane example to be farthest the possibility of positive example larger, therefore, utilize this example to represent many examples bag at its place.For
lindividual sorter, in conjunction with confidence level presented above
c k , with many examples bag that classification plane is nearer, its uncertainty is also larger, also namely to the effect that classifier performance is improved most.
Based on above analysis, selection strategy represents as follows:
In Active Learning, many examples bag of most worthy is exactly the most uncertain sample of sorter, therefore the many examples bag calculated according to selection strategy and the distance of separation vessel lineoid, the minimum many examples bag of chosen distance joins training set and trains, and will improve the performance of sorter.
7th step, removing the scene image marked through expert, being placed with label scene image data collection from concentrating without label image data, and train classification models again;
8th step, judges whether the degree of accuracy of this model reaches the degree of accuracy required by user, or whether iteration wheel number reaches the number of times that user specifies, if do not reach requirement, returns the 3rd step; Otherwise terminate and output category model.
Above-described embodiment of the present invention, does not form limiting the scope of the present invention.Any amendment done within the spirit and principles in the present invention, equivalent replacement and improvement etc., all should be included within claims of the present invention.
Claims (5)
1., based on the scene image mask method of Active Learning and many labels multi-instance learning, it is characterized in that, comprise the steps,
The first step, obtain a collection of scene image without label; Randomly draw a small amount of scene image, by the artificial notation methods of expert, give these scene image tag along sorts, the maximum number of tag along sort is k, k >=2, original scene image set is divided into again two set, a set comprises label scene image on a small quantity, and another one set comprises remaining a large amount of without label scene image;
Second step, having label scene image and being converted into many sample datas without label scene image, every width image regards example bag more than as, and an example of many examples bag is regarded in each region as;
3rd step, there being label scene image to regard training set as on a small quantity, according to the number of labels of scene image, train several preliminary classification models;
The disaggregated model that 4th step, utilization have been set up, to marking without label scene image in sample set, each image may have multiple label;
5th step, according to the annotation results without label scene image, calculate the confidence level of each disaggregated model;
The confidence level of the 6th step, combining classification model, from without selecting the maximum image of a uncertainty label scene image, and gives expert and marks this scene image;
7th step, the scene image marked through expert to be removed from concentrating without label image data, being placed with label scene image data collection, and train classification models again;
8th step, judge whether the degree of accuracy of this model reaches the degree of accuracy required by user, or whether iteration wheel number reaches the number of times that user specifies, if do not reach requirement, returns the 3rd step; Otherwise terminate and output category model.
2. the scene image mask method based on Active Learning and many labels multi-instance learning according to claim 1, it is characterized in that, the concrete grammar of the 3rd step is: regard training set as there being label scene image on a small quantity, according to k tag along sort of scene image, train k preliminary classification model, for each tag along sort, the image with this label is regarded as positive class data, the image without this label is regarded as negative class data, trains initial many example classification model.
3. the scene image mask method based on Active Learning and many labels multi-instance learning according to claim 2, is characterized in that, in the 4th step, utilize k the disaggregated model set up in the 3rd step, predict the label without label scene image; Through k disaggregated model, each pair will obtain k tag along sort without label scene image; For i-th disaggregated model, if the value of tag along sort is 1, represent that this scene image comprises the picture material of the i-th class; If the value of tag along sort is 0, represent that this scene image does not comprise the picture material of the i-th class.
4. the scene image mask method based on Active Learning and many labels multi-instance learning according to claim 3, it is characterized in that, when calculating the confidence level of each disaggregated model, the criterion of disaggregated model to prediction label confidence level is: first utilizing has the training of label many examples bag
kindividual sorter, recycling obtains
kindividual sorter is classified to without label many examples bag, obtains its prediction label, assuming that
xrepresent instance space,
yrepresent tally set space, given
n l individual have label many examples bag
with
n u individual without label many examples bag
; Target is that study obtains objective function
f mIML : 2 x →
2 y ; Wherein,
a corresponding example collection,
,
for
x i corresponding one group of tag set
y i1 ,
y i2 ...,
y il ,
y ik =0,1} (
k=1,2 ...,
l), here,
n i represent many examples bag
x i in containing the number of example,
lrepresent the label number in many examples bag; On this basis,
kthe confidence level of individual disaggregated model
c k can be defined as:
In above formula,
i [] be an indicator function (indicatorfunction), satisfied [] specified criteria then its value is 1, otherwise value is 0;
y l ik represent the
kin individual sorter
ithe individual label having label many examples bag,
y u ik represent the
kin individual sorter
ithe individual label without label many examples bag;
indicate and wrap in without the many examples of label
kthe mean value of the positive label predicted in individual sorter,
be shown with the many examples of label and wrap in
kthe mean value of positive label in individual sorter, confidence level
c k less, illustrate without exemplar proportion positive in label many examples bag and have the ratio shared by label many examples bag more not close, namely confidence level is lower, otherwise then confidence level is higher.
5. the scene image mask method based on Active Learning and many labels multi-instance learning according to claim 4, is characterized in that, from without what select label scene image that the maximum image of uncertainty adopts being in the 6th step
minimum classification distance selection strategy,this strategy represents as follows:
, in Active Learning, many examples bag of most worthy is exactly the most uncertain sample of sorter, the many examples bag therefore calculated according to selection strategy and the distance of separation vessel lineoid, and the minimum many examples bag of chosen distance joins training set and trains;
Wherein,
In above formula,
f k (X ij )represent many examples bag
x i in
jindividual example is
kthe classification function output valve of individual SVM classifier,
represent example
x ij for
kthe lineoid distance of individual SVM classifier,
represent many examples bag X
imiddle distance
kindividual SVM classifier lineoid example farthest.
Priority Applications (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN201510473322.2A CN105117429B (en) | 2015-08-05 | 2015-08-05 | Scene image mask method based on Active Learning and multi-tag multi-instance learning |
Applications Claiming Priority (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN201510473322.2A CN105117429B (en) | 2015-08-05 | 2015-08-05 | Scene image mask method based on Active Learning and multi-tag multi-instance learning |
Publications (2)
Publication Number | Publication Date |
---|---|
CN105117429A true CN105117429A (en) | 2015-12-02 |
CN105117429B CN105117429B (en) | 2018-11-23 |
Family
ID=54665419
Family Applications (1)
Application Number | Title | Priority Date | Filing Date |
---|---|---|---|
CN201510473322.2A Expired - Fee Related CN105117429B (en) | 2015-08-05 | 2015-08-05 | Scene image mask method based on Active Learning and multi-tag multi-instance learning |
Country Status (1)
Country | Link |
---|---|
CN (1) | CN105117429B (en) |
Cited By (31)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN105701509A (en) * | 2016-01-13 | 2016-06-22 | 清华大学 | Image classification method based on cross-type migration active learning |
CN106127247A (en) * | 2016-06-21 | 2016-11-16 | 广东工业大学 | Image classification method based on multitask many examples support vector machine |
CN106250924A (en) * | 2016-07-27 | 2016-12-21 | 南京大学 | A kind of newly-increased category detection method based on multi-instance learning |
CN106897424A (en) * | 2017-02-24 | 2017-06-27 | 北京时间股份有限公司 | Information labeling system and method |
CN107392125A (en) * | 2017-07-11 | 2017-11-24 | 中国科学院上海高等研究院 | Training method/system, computer-readable recording medium and the terminal of model of mind |
CN107577994A (en) * | 2017-08-17 | 2018-01-12 | 南京邮电大学 | A kind of pedestrian based on deep learning, the identification of vehicle auxiliary product and search method |
CN107832780A (en) * | 2017-10-17 | 2018-03-23 | 北京木业邦科技有限公司 | Low confidence sample processing method and system are sorted based on artificial intelligence plank |
CN108009589A (en) * | 2017-12-12 | 2018-05-08 | 腾讯科技(深圳)有限公司 | Sample data processing method, device and computer-readable recording medium |
CN108334943A (en) * | 2018-01-03 | 2018-07-27 | 浙江大学 | The semi-supervised soft-measuring modeling method of industrial process based on Active Learning neural network model |
CN108427970A (en) * | 2018-03-29 | 2018-08-21 | 厦门美图之家科技有限公司 | Picture mask method and device |
CN108959304A (en) * | 2017-05-22 | 2018-12-07 | 阿里巴巴集团控股有限公司 | A kind of Tag Estimation method and device |
CN109242013A (en) * | 2018-08-28 | 2019-01-18 | 北京九狐时代智能科技有限公司 | A kind of data mask method, device, electronic equipment and storage medium |
WO2019021088A1 (en) * | 2017-07-24 | 2019-01-31 | International Business Machines Corporation | Navigating video scenes using cognitive insights |
CN109800776A (en) * | 2017-11-17 | 2019-05-24 | 中兴通讯股份有限公司 | Material mask method, device, terminal and computer readable storage medium |
CN109886211A (en) * | 2019-02-25 | 2019-06-14 | 北京达佳互联信息技术有限公司 | Data mask method, device, electronic equipment and storage medium |
CN109977994A (en) * | 2019-02-02 | 2019-07-05 | 浙江工业大学 | A kind of presentation graphics choosing method based on more example Active Learnings |
CN110175657A (en) * | 2019-06-05 | 2019-08-27 | 广东工业大学 | A kind of image multi-tag labeling method, device, equipment and readable storage medium storing program for executing |
CN110288007A (en) * | 2019-06-05 | 2019-09-27 | 北京三快在线科技有限公司 | The method, apparatus and electronic equipment of data mark |
CN110378396A (en) * | 2019-06-26 | 2019-10-25 | 北京百度网讯科技有限公司 | Sample data mask method, device, computer equipment and storage medium |
CN110414622A (en) * | 2019-08-06 | 2019-11-05 | 广东工业大学 | Classifier training method and device based on semi-supervised learning |
CN110458180A (en) * | 2019-04-28 | 2019-11-15 | 广东工业大学 | A kind of classifier training method based on small sample |
CN111340131A (en) * | 2020-03-09 | 2020-06-26 | 北京字节跳动网络技术有限公司 | Image annotation method and device, readable medium and electronic equipment |
CN111353549A (en) * | 2020-03-10 | 2020-06-30 | 创新奇智(重庆)科技有限公司 | Image tag verification method and device, electronic device and storage medium |
CN111368917A (en) * | 2020-03-04 | 2020-07-03 | 西安邮电大学 | Multi-example ensemble learning method for criminal investigation image classification |
CN111461265A (en) * | 2020-05-27 | 2020-07-28 | 东北大学 | Scene image labeling method based on coarse-fine granularity multi-image multi-label learning |
CN111476285A (en) * | 2020-04-01 | 2020-07-31 | 深圳力维智联技术有限公司 | Training method of image classification model, image classification method and storage medium |
CN111582329A (en) * | 2020-04-22 | 2020-08-25 | 西安交通大学 | Natural scene text character detection and labeling method based on multi-example learning |
CN111797660A (en) * | 2019-04-09 | 2020-10-20 | Oppo广东移动通信有限公司 | Image labeling method and device, storage medium and electronic equipment |
CN113177576A (en) * | 2021-03-31 | 2021-07-27 | 中国科学院大学 | Multi-example active learning method for target detection |
US11409589B1 (en) | 2019-10-23 | 2022-08-09 | Relativity Oda Llc | Methods and systems for determining stopping point |
CN115035406A (en) * | 2022-06-08 | 2022-09-09 | 中国科学院空间应用工程与技术中心 | Method and system for labeling remote sensing scene data set, storage medium and electronic equipment |
Citations (4)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN101853400A (en) * | 2010-05-20 | 2010-10-06 | 武汉大学 | Multiclass image classification method based on active learning and semi-supervised learning |
CN103116893A (en) * | 2013-03-15 | 2013-05-22 | 南京大学 | Digital image labeling method based on multi-exampling multi-marking learning |
CN103258214A (en) * | 2013-04-26 | 2013-08-21 | 南京信息工程大学 | Remote sensing image classification method based on image block active learning |
CN104182767A (en) * | 2014-09-05 | 2014-12-03 | 西安电子科技大学 | Active learning and neighborhood information combined hyperspectral image classification method |
-
2015
- 2015-08-05 CN CN201510473322.2A patent/CN105117429B/en not_active Expired - Fee Related
Patent Citations (4)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN101853400A (en) * | 2010-05-20 | 2010-10-06 | 武汉大学 | Multiclass image classification method based on active learning and semi-supervised learning |
CN103116893A (en) * | 2013-03-15 | 2013-05-22 | 南京大学 | Digital image labeling method based on multi-exampling multi-marking learning |
CN103258214A (en) * | 2013-04-26 | 2013-08-21 | 南京信息工程大学 | Remote sensing image classification method based on image block active learning |
CN104182767A (en) * | 2014-09-05 | 2014-12-03 | 西安电子科技大学 | Active learning and neighborhood information combined hyperspectral image classification method |
Non-Patent Citations (3)
Title |
---|
XIN LI等: "Active Learning with Multi-Label SVM Classification", 《PROCEEDINGS OF THE TWENTY-THIRD INTERNATIONAL JOINT CONFERENCE ON ARTIFICIAL INTELLIGENCE》 * |
ZHI-HUA ZHOU等: "Multi-instance multi-label learning", 《ARTIFICIAL INTELLIGENCE》 * |
徐美香等: "主动学习的多标签图像在线分类", 《中国图象图形学报》 * |
Cited By (50)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN105701509A (en) * | 2016-01-13 | 2016-06-22 | 清华大学 | Image classification method based on cross-type migration active learning |
CN105701509B (en) * | 2016-01-13 | 2019-03-12 | 清华大学 | A kind of image classification method based on across classification migration Active Learning |
CN106127247A (en) * | 2016-06-21 | 2016-11-16 | 广东工业大学 | Image classification method based on multitask many examples support vector machine |
CN106127247B (en) * | 2016-06-21 | 2019-07-09 | 广东工业大学 | Image classification method based on the more example support vector machines of multitask |
CN106250924A (en) * | 2016-07-27 | 2016-12-21 | 南京大学 | A kind of newly-increased category detection method based on multi-instance learning |
CN106250924B (en) * | 2016-07-27 | 2019-07-16 | 南京大学 | A kind of newly-increased category detection method based on multi-instance learning |
CN106897424A (en) * | 2017-02-24 | 2017-06-27 | 北京时间股份有限公司 | Information labeling system and method |
CN108959304A (en) * | 2017-05-22 | 2018-12-07 | 阿里巴巴集团控股有限公司 | A kind of Tag Estimation method and device |
CN108959304B (en) * | 2017-05-22 | 2022-03-25 | 阿里巴巴集团控股有限公司 | Label prediction method and device |
CN107392125A (en) * | 2017-07-11 | 2017-11-24 | 中国科学院上海高等研究院 | Training method/system, computer-readable recording medium and the terminal of model of mind |
WO2019021088A1 (en) * | 2017-07-24 | 2019-01-31 | International Business Machines Corporation | Navigating video scenes using cognitive insights |
US10970334B2 (en) | 2017-07-24 | 2021-04-06 | International Business Machines Corporation | Navigating video scenes using cognitive insights |
CN107577994A (en) * | 2017-08-17 | 2018-01-12 | 南京邮电大学 | A kind of pedestrian based on deep learning, the identification of vehicle auxiliary product and search method |
CN107832780B (en) * | 2017-10-17 | 2020-04-10 | 北京木业邦科技有限公司 | Artificial intelligence-based wood board sorting low-confidence sample processing method and system |
CN107832780A (en) * | 2017-10-17 | 2018-03-23 | 北京木业邦科技有限公司 | Low confidence sample processing method and system are sorted based on artificial intelligence plank |
CN109800776A (en) * | 2017-11-17 | 2019-05-24 | 中兴通讯股份有限公司 | Material mask method, device, terminal and computer readable storage medium |
CN108009589A (en) * | 2017-12-12 | 2018-05-08 | 腾讯科技(深圳)有限公司 | Sample data processing method, device and computer-readable recording medium |
CN108334943A (en) * | 2018-01-03 | 2018-07-27 | 浙江大学 | The semi-supervised soft-measuring modeling method of industrial process based on Active Learning neural network model |
CN108427970A (en) * | 2018-03-29 | 2018-08-21 | 厦门美图之家科技有限公司 | Picture mask method and device |
CN109242013A (en) * | 2018-08-28 | 2019-01-18 | 北京九狐时代智能科技有限公司 | A kind of data mask method, device, electronic equipment and storage medium |
CN109242013B (en) * | 2018-08-28 | 2021-06-08 | 北京九狐时代智能科技有限公司 | Data labeling method and device, electronic equipment and storage medium |
CN109977994A (en) * | 2019-02-02 | 2019-07-05 | 浙江工业大学 | A kind of presentation graphics choosing method based on more example Active Learnings |
CN109977994B (en) * | 2019-02-02 | 2021-04-09 | 浙江工业大学 | Representative image selection method based on multi-example active learning |
CN109886211B (en) * | 2019-02-25 | 2022-03-01 | 北京达佳互联信息技术有限公司 | Data labeling method and device, electronic equipment and storage medium |
CN109886211A (en) * | 2019-02-25 | 2019-06-14 | 北京达佳互联信息技术有限公司 | Data mask method, device, electronic equipment and storage medium |
CN111797660A (en) * | 2019-04-09 | 2020-10-20 | Oppo广东移动通信有限公司 | Image labeling method and device, storage medium and electronic equipment |
CN110458180A (en) * | 2019-04-28 | 2019-11-15 | 广东工业大学 | A kind of classifier training method based on small sample |
CN110458180B (en) * | 2019-04-28 | 2023-09-19 | 广东工业大学 | Classifier training method based on small samples |
CN110175657B (en) * | 2019-06-05 | 2021-10-01 | 广东工业大学 | Image multi-label marking method, device, equipment and readable storage medium |
CN110288007B (en) * | 2019-06-05 | 2021-02-02 | 北京三快在线科技有限公司 | Data labeling method and device and electronic equipment |
CN110175657A (en) * | 2019-06-05 | 2019-08-27 | 广东工业大学 | A kind of image multi-tag labeling method, device, equipment and readable storage medium storing program for executing |
CN110288007A (en) * | 2019-06-05 | 2019-09-27 | 北京三快在线科技有限公司 | The method, apparatus and electronic equipment of data mark |
CN110378396A (en) * | 2019-06-26 | 2019-10-25 | 北京百度网讯科技有限公司 | Sample data mask method, device, computer equipment and storage medium |
CN110414622A (en) * | 2019-08-06 | 2019-11-05 | 广东工业大学 | Classifier training method and device based on semi-supervised learning |
CN110414622B (en) * | 2019-08-06 | 2022-06-24 | 广东工业大学 | Classifier training method and device based on semi-supervised learning |
US11921568B2 (en) | 2019-10-23 | 2024-03-05 | Relativity Oda Llc | Methods and systems for determining stopping point |
US11409589B1 (en) | 2019-10-23 | 2022-08-09 | Relativity Oda Llc | Methods and systems for determining stopping point |
CN111368917A (en) * | 2020-03-04 | 2020-07-03 | 西安邮电大学 | Multi-example ensemble learning method for criminal investigation image classification |
CN111340131A (en) * | 2020-03-09 | 2020-06-26 | 北京字节跳动网络技术有限公司 | Image annotation method and device, readable medium and electronic equipment |
CN111353549B (en) * | 2020-03-10 | 2023-01-31 | 创新奇智(重庆)科技有限公司 | Image label verification method and device, electronic equipment and storage medium |
CN111353549A (en) * | 2020-03-10 | 2020-06-30 | 创新奇智(重庆)科技有限公司 | Image tag verification method and device, electronic device and storage medium |
CN111476285A (en) * | 2020-04-01 | 2020-07-31 | 深圳力维智联技术有限公司 | Training method of image classification model, image classification method and storage medium |
CN111476285B (en) * | 2020-04-01 | 2023-07-28 | 深圳力维智联技术有限公司 | Training method of image classification model, image classification method and storage medium |
CN111582329A (en) * | 2020-04-22 | 2020-08-25 | 西安交通大学 | Natural scene text character detection and labeling method based on multi-example learning |
CN111582329B (en) * | 2020-04-22 | 2023-03-28 | 西安交通大学 | Natural scene text character detection and labeling method based on multi-example learning |
CN111461265B (en) * | 2020-05-27 | 2023-07-25 | 东北大学 | Scene image labeling method based on coarse-fine granularity multi-image multi-label learning |
CN111461265A (en) * | 2020-05-27 | 2020-07-28 | 东北大学 | Scene image labeling method based on coarse-fine granularity multi-image multi-label learning |
CN113177576A (en) * | 2021-03-31 | 2021-07-27 | 中国科学院大学 | Multi-example active learning method for target detection |
CN115035406A (en) * | 2022-06-08 | 2022-09-09 | 中国科学院空间应用工程与技术中心 | Method and system for labeling remote sensing scene data set, storage medium and electronic equipment |
CN115035406B (en) * | 2022-06-08 | 2023-08-04 | 中国科学院空间应用工程与技术中心 | Remote sensing scene data set labeling method, remote sensing scene data set labeling system, storage medium and electronic equipment |
Also Published As
Publication number | Publication date |
---|---|
CN105117429B (en) | 2018-11-23 |
Similar Documents
Publication | Publication Date | Title |
---|---|---|
CN105117429A (en) | Scenario image annotation method based on active learning and multi-label multi-instance learning | |
CN102508859B (en) | Advertisement classification method and device based on webpage characteristic | |
CN112925921B (en) | DIKW atlas-based resource identification method, related device and readable medium | |
CN107169049A (en) | The label information generation method and device of application | |
CN105631479A (en) | Imbalance-learning-based depth convolution network image marking method and apparatus | |
CN106991382A (en) | A kind of remote sensing scene classification method | |
CN104217225A (en) | A visual target detection and labeling method | |
CN106815604A (en) | Method for viewing points detecting based on fusion of multi-layer information | |
CN105005794A (en) | Image pixel semantic annotation method with combination of multi-granularity context information | |
CN104376105A (en) | Feature fusing system and method for low-level visual features and text description information of images in social media | |
CN113515632B (en) | Text classification method based on graph path knowledge extraction | |
CN112052818A (en) | Unsupervised domain adaptive pedestrian detection method, unsupervised domain adaptive pedestrian detection system and storage medium | |
CN102945372B (en) | Classifying method based on multi-label constraint support vector machine | |
CN116129286A (en) | Method for classifying graphic neural network remote sensing images based on knowledge graph | |
CN106204103A (en) | The method of similar users found by a kind of moving advertising platform | |
Zhang et al. | Generative AI-enabled vehicular networks: Fundamentals, framework, and case study | |
CN113781404B (en) | Road disease detection method and system based on self-supervision pre-training | |
CN104346456B (en) | The digital picture multi-semantic meaning mask method measured based on spatial dependence | |
Alajlan et al. | Automatic lane marking prediction using convolutional neural network and S-Shaped Binary Butterfly Optimization | |
CN110377753A (en) | Relation extraction method and device based on relationship trigger word Yu GRU model | |
CN115035354A (en) | Reservoir water surface floater target detection method based on improved YOLOX | |
CN113076741A (en) | Multilingual text data analysis method | |
Lu | Design of visual communication based on deep learning approaches | |
Li et al. | Object detection based on knowledge graph network | |
Song et al. | Prior-guided multi-scale fusion transformer for face attribute recognition |
Legal Events
Date | Code | Title | Description |
---|---|---|---|
C06 | Publication | ||
PB01 | Publication | ||
C10 | Entry into substantive examination | ||
SE01 | Entry into force of request for substantive examination | ||
GR01 | Patent grant | ||
GR01 | Patent grant | ||
CF01 | Termination of patent right due to non-payment of annual fee | ||
CF01 | Termination of patent right due to non-payment of annual fee |
Granted publication date: 20181123 Termination date: 20190805 |