CN114612699A - Image data processing method and device - Google Patents
Image data processing method and device Download PDFInfo
- Publication number
- CN114612699A CN114612699A CN202210235982.7A CN202210235982A CN114612699A CN 114612699 A CN114612699 A CN 114612699A CN 202210235982 A CN202210235982 A CN 202210235982A CN 114612699 A CN114612699 A CN 114612699A
- Authority
- CN
- China
- Prior art keywords
- data
- sample
- labeling
- initial training
- result data
- 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
- 238000003672 processing method Methods 0.000 title description 5
- 238000002372 labelling Methods 0.000 claims abstract description 172
- 238000012549 training Methods 0.000 claims abstract description 132
- 238000000034 method Methods 0.000 claims abstract description 39
- 238000012545 processing Methods 0.000 claims abstract description 29
- 238000004590 computer program Methods 0.000 claims description 11
- 238000012216 screening Methods 0.000 claims description 8
- 238000012937 correction Methods 0.000 description 11
- 238000010586 diagram Methods 0.000 description 11
- 238000004891 communication Methods 0.000 description 6
- 230000006870 function Effects 0.000 description 6
- 238000003062 neural network model Methods 0.000 description 4
- 230000003287 optical effect Effects 0.000 description 4
- 238000005516 engineering process Methods 0.000 description 3
- 238000012986 modification Methods 0.000 description 3
- 230000004048 modification Effects 0.000 description 3
- 230000008901 benefit Effects 0.000 description 2
- 238000013527 convolutional neural network Methods 0.000 description 2
- 239000000835 fiber Substances 0.000 description 2
- 230000008569 process Effects 0.000 description 2
- 230000000644 propagated effect Effects 0.000 description 2
- 239000004065 semiconductor Substances 0.000 description 2
- 238000012163 sequencing technique Methods 0.000 description 2
- 238000013473 artificial intelligence Methods 0.000 description 1
- 238000004364 calculation method Methods 0.000 description 1
- 238000010276 construction Methods 0.000 description 1
- 238000013135 deep learning Methods 0.000 description 1
- 238000013136 deep learning model Methods 0.000 description 1
- 238000013461 design Methods 0.000 description 1
- 238000011161 development Methods 0.000 description 1
- 230000000694 effects Effects 0.000 description 1
- 238000000605 extraction Methods 0.000 description 1
- 230000006872 improvement Effects 0.000 description 1
- 239000004973 liquid crystal related substance Substances 0.000 description 1
- 238000013507 mapping Methods 0.000 description 1
- 239000013307 optical fiber Substances 0.000 description 1
- 238000005070 sampling Methods 0.000 description 1
- 238000006467 substitution reaction Methods 0.000 description 1
Images
Classifications
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F18/00—Pattern recognition
- G06F18/20—Analysing
- G06F18/22—Matching criteria, e.g. proximity measures
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F18/00—Pattern recognition
- G06F18/20—Analysing
- G06F18/21—Design or setup of recognition systems or techniques; Extraction of features in feature space; Blind source separation
- G06F18/214—Generating training patterns; Bootstrap methods, e.g. bagging or boosting
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- 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
Abstract
The invention discloses a method and a device for processing image data, and relates to the technical field of computers. One embodiment of the method comprises: labeling a data set to be processed by using an image labeling model obtained by training an initial training sample set to obtain a labeling result data set; determining the similarity between the labeling result data set and the initial training sample set; and updating the initial training sample set according to the similarity. According to the embodiment, wrong labels in the initial training sample set can be efficiently updated and corrected, the labeling quality of the initial training sample set is improved, and the performance of the image labeling model obtained through training is further improved.
Description
Technical Field
The present invention relates to the field of computer technologies, and in particular, to a method and an apparatus for processing image data.
Background
With the continuous development of artificial intelligence technology, deep learning technology has more and more prominent role in various industries, for example, it can be applied to the processing of image data. When the deep learning model is trained, a large amount of training data needs to be obtained, and the quality of labels in the training data influences the performance of the model.
In the prior art, the scheme of weak supervision and semi-supervision learning is mainly adopted to improve the labeling speed and efficiency of data, but due to the difficulty in efficiently updating and correcting wrong labels of original data, the labeling accuracy is low, and further the labeling accuracy of a trained model is low.
Disclosure of Invention
In view of this, embodiments of the present invention provide a method and an apparatus for processing image data, which can efficiently update and correct a wrong label in an initial training sample set, improve the labeling quality of the updated initial training sample set, and further improve the performance of an image labeling model obtained by training.
To achieve the above object, according to an aspect of an embodiment of the present invention, there is provided an image data processing method including:
labeling a data set to be processed by using an image labeling model obtained by training an initial training sample set to obtain a labeling result data set;
determining similarity between the labeling result data set and the initial training sample set;
and updating the initial training sample set according to the similarity.
Optionally, after obtaining the annotation result data set, the method further includes: and correcting the labeling result data set.
Optionally, determining the similarity between the labeling result data set and the initial training sample set includes:
and for any one piece of labeling result data in the labeling result data set, respectively calculating the similarity between the labeling result data and each piece of sample data in the initial training sample set.
Optionally, updating the initial training sample set according to the similarity includes:
aiming at any one piece of marked result data, screening out a sample data subset which has the similarity with any one piece of marked result data and meets a preset condition from the initial training sample set;
and updating the initial training sample set according to the sample data subsets corresponding to all the marked result data and the marked result data set.
Optionally, before updating the initial training sample set according to the sample data subset corresponding to all the annotation result data and the annotation result data set, the method further includes: correcting the sample data meeting the preset conditions; and/or
And carrying out duplicate removal on the sample data meeting the preset condition.
Optionally, calculating a similarity between the labeling result data and each sample data in the initial training sample set includes:
determining a sample characteristic corresponding to each sample data and an annotation result characteristic corresponding to any annotation result data;
and determining the similarity between the labeling result data and each piece of sample data according to the distance between the labeling result characteristics and each piece of sample characteristics.
Optionally, determining a sample feature corresponding to each sample data includes:
respectively determining M sub-features corresponding to the sample data by utilizing M preset models, wherein M is a positive integer greater than or equal to 1;
splicing the M sub-features to obtain a sample feature corresponding to each piece of sample data; or, the M sub-features are used as M sample features corresponding to each piece of sample data.
In another aspect of the embodiments of the present invention, an apparatus for processing image data is provided, including:
the model labeling module is used for labeling the data set to be processed by using an image labeling model obtained by training of the initial training sample set to obtain a labeling result data set;
the determining module is used for determining the similarity between the labeling result data set and the initial training sample set;
and the updating module is used for updating the initial training sample set according to the similarity.
According to still another aspect of an embodiment of the present invention, there is provided an electronic apparatus including:
one or more processors;
a storage device for storing one or more programs,
when the one or more programs are executed by the one or more processors, the one or more processors implement the method for processing image data provided by the present invention.
According to a further aspect of an embodiment of the present invention, there is provided a computer-readable medium on which a computer program is stored, the program, when executed by a processor, implementing the method of image data processing provided by the present invention.
One embodiment of the above invention has the following advantages or benefits: according to the image data processing method provided by the embodiment of the invention, an image labeling model obtained by training an initial training sample set is used for labeling a data set to be processed to obtain a labeling result data set; then, determining the similarity between the labeling result data set and the initial training sample set, and screening or searching out a sample data subset with the similarity meeting a preset condition from the initial training sample set; and updating the initial training sample set by using the labeling result data set and the initial training sample set. The method provided by the embodiment of the invention can efficiently update and correct the wrong labels in the initial training sample set, improve the labeling quality of the initial training sample set, further improve the performance of the image labeling model obtained by training, and improve the labeling accuracy of the image labeling model.
Further effects of the above-mentioned non-conventional alternatives will be described below in connection with the embodiments.
Drawings
The drawings are included to provide a better understanding of the invention and are not to be construed as unduly limiting the invention. Wherein:
fig. 1 is a schematic diagram of a main flow of a method of image data processing according to an embodiment of the present invention;
FIG. 2 is a schematic main flow diagram of another method of image data processing according to an embodiment of the invention;
FIG. 3 is a schematic flow chart of a method for processing image data according to another embodiment of the present invention;
FIG. 4 is a flow chart illustrating a method of image data processing according to an embodiment of the present invention;
fig. 5 is a schematic diagram of main blocks of an apparatus for image data processing according to an embodiment of the present invention;
FIG. 6 is an exemplary system architecture diagram in which embodiments of the present invention may be employed;
fig. 7 is a schematic block diagram of a computer system suitable for use in implementing a terminal device or server of an embodiment of the invention.
Detailed Description
Exemplary embodiments of the present invention are described below with reference to the accompanying drawings, in which various details of embodiments of the invention are included to assist understanding, and which are to be considered as merely exemplary. Accordingly, those of ordinary skill in the art will recognize that various changes and modifications of the embodiments described herein can be made without departing from the scope and spirit of the invention. Also, descriptions of well-known functions and constructions are omitted in the following description for clarity and conciseness.
Fig. 1 is a schematic diagram of a main flow of a method for processing image data according to an embodiment of the present invention, as shown in fig. 1, the method for processing image data, applied to a server, includes the following steps:
step S101: labeling a data set to be processed by using an image labeling model obtained by training an initial training sample set to obtain a labeling result data set;
step S102: determining the similarity between the labeling result data set and the initial training sample set;
step S103: and updating the initial training sample set according to the similarity.
In the embodiment of the present invention, the initial training sample set includes a plurality of pieces of sample data, and each piece of sample data includes image data and label data (i.e., a label). The initial training sample set can be a purely artificially labeled image data set, and can also be a noise data set obtained by a semi-supervised or weakly supervised learning method. The initial training sample set can have wrong labels, the labeling quality of the wrong labels is low, and further the performance of an image labeling model obtained by training the initial training sample set containing the wrong labels is poor, the labeling accuracy is low, and therefore the wrong labels in the initial training sample set need to be updated and corrected.
Training a neural network model by using an initial training sample set to obtain an image annotation model, wherein the neural network model can be a CNN (convolutional neural network) model of resnet50 (an image recognition model structure), and can also be other neural network models, such as EfficientNet, GoogleNet and other neural network models.
The image annotation model can be used for annotating the data set to be processed without the label to obtain an annotation result data set, the annotation result data set comprises a plurality of pieces of annotation result data, and each piece of annotation result data comprises image data and model annotation data obtained by the annotation of the image annotation model. The source of the data set to be processed is not particularly limited, and may be online data of some businesses, or data crawled from a third party or the internet.
In an optional implementation manner of the embodiment of the present invention, in step S101, after obtaining the annotation result data set, the method further includes: and correcting the labeling result data set. Because the model annotation data in the annotation result dataset may be inaccurate, the incorrect or inaccurate model annotation data can be corrected by correcting. The correction can be a manual correction mode or an automatic correction mode so as to realize the correction of wrong labels in the model labeling data.
Optionally, when the labeled result data set is corrected, in order to improve the efficiency and accuracy of the correction, the labeled result data set may be sampled and then corrected, for example, part of the labeled result data may be extracted from the labeled result data set for correction, so that the similarity between the labeled result data set and the initial training sample set may be determined according to part of the labeled result data in the labeled result set. For example, 50 pieces of labeling result data are extracted from a labeling result data set including 100 pieces of labeling result data to perform manual correction, and the similarity between the 50 pieces of labeling result data and the initial training sample set is determined by taking the 50 pieces of labeling result data after the manual correction as the labeling result data set.
In this embodiment of the present invention, in step S102, determining a similarity between the labeling result data set and the initial training sample set includes: and respectively calculating the similarity between the labeling result data and each sample data in the initial training sample set for any labeling result data in the labeling result data set.
And when the similarity between the labeling result data set and the initial training sample set is determined, calculating the similarity between each sample data in the initial training sample set and the labeling result data aiming at any labeling result data. For example, if the initial training sample set includes K pieces of sample data, where K is an integer greater than or equal to 1, then a result of K similarities may be calculated for any piece of labeling result data, and if the labeling result data set includes L pieces of labeling result data, then a result of L × K similarities is calculated in total. The result of calculating the obtained L × K similarities may be used as the similarity between the labeled result set and the initial training sample set.
In an optional implementation manner of the embodiment of the present invention, the similarity may be a distance, for example, the similarity may be at least one of a Cosine distance (Cosine distance, also referred to as Cosine similarity), an euclidean distance, and the like.
In the embodiment of the present invention, as shown in fig. 2, updating the initial training sample set according to the similarity includes:
step S201: for any marking result data, screening a sample data subset which has the similarity with any marking result data and meets a preset condition from the initial training sample set;
step S201: and updating the initial training sample set according to the sample data subsets corresponding to all the marked result data and the marked result data set.
In the embodiment of the invention, after the similarity between any labeling result data and each sample data in the initial training sample set is calculated for any labeling result data, whether the similarity meets the preset condition is judged, and the sample data subset is formed according to the sample data meeting the preset condition, wherein the similarity is the similarity between the standard result data and the sample data, the expression modes of different similarities are different, and the preset conditions are different. When the similarity is a distance, taking sample data of which the distance from any one of the labeled result data is smaller than a preset distance threshold (such as 0.1) as sample data meeting a preset condition; the sample data meeting the preset condition can also be: and for any marking result data, sequencing the distance between each piece of sample data and any marking result data in a descending order, and taking the sample data corresponding to the distance N before sequencing as the sample data meeting the preset condition, wherein N is a positive integer greater than 1.
For each piece of labeling result data in the labeling result data set, the labeling result data set corresponds to one sample data subset, and for the labeling result data set containing L pieces of labeling result data, L sample data subsets can be obtained, that is, the sample data subsets corresponding to all the labeling result data are obtained. The initial training sample set may be updated according to the sample data subsets and the labeled result data sets corresponding to all the labeled result data, that is, the updated initial training sample set includes the sample data subsets and the labeled result data sets corresponding to all the labeled result data. Or updating the initial training sample set according to part of sample data in the sample data subset corresponding to all the labeling result data and all or part of the labeling result data.
In the embodiment of the present invention, before updating the initial training sample set according to the sample data subset and the labeled result data set corresponding to all labeled result data, the method further includes: correcting sample data meeting preset conditions; and/or
And carrying out duplicate removal on the sample data meeting the preset condition.
In order to further improve the accuracy of labeling the labels in the sample data subsets, sample data meeting preset conditions is corrected before updating the initial training sample set according to the sample data subsets and the labeling result data sets corresponding to all the labeling result data, the corrected sample data subsets are obtained, correction of wrong labels is achieved, the initial training sample set is updated by using the corrected sample data subsets and the labeling result data sets, the labeling quality of the updated initial training sample set can be improved, and the performance of the image labeling model is further improved.
In the embodiment of the present invention, each piece of labeling result data corresponds to one sample data subset, the sample data subsets include sample data meeting preset conditions, all pieces of labeling result data correspond to multiple sample data subsets, and the multiple sample data subsets may have the same data, so that the same data in the sample data subsets corresponding to all pieces of labeling result data can be deduplicated to reduce data.
In an optional implementation manner of the embodiment of the present invention, as shown in fig. 3, the calculating a similarity between the annotation result data and each sample data in the initial training sample set includes:
step S301: determining sample characteristics corresponding to each sample datum and labeling result characteristics corresponding to any labeling result datum;
step S301: and determining the similarity between the labeling result data and each sample data according to the distance between the labeling result characteristics and each sample characteristic.
In the embodiment of the invention, when the similarity between the labeling result data and each sample data in the initial training sample set is calculated, the labeling result characteristic corresponding to the labeling result data and the sample characteristic corresponding to each sample data can be extracted, wherein the labeling result characteristic and the sample characteristic can be represented by using a characteristic vector, and the labeling result characteristic and the sample characteristic can be extracted by using an image labeling model or other extensive open source models.
And extracting the characteristic of the labeling result from the data set of the labeling result to form a characteristic set of the labeling result, extracting the characteristic of the sample from the initial training sample set to form a characteristic set of the sample, and taking the similarity between the characteristic set of the labeling result and the characteristic set of the sample as the similarity between the data of the labeling result and the initial training sample set. And aiming at any marking result characteristic, taking the distance between any marking result characteristic and each sample characteristic obtained by calculation as the similarity between any marking result data and each sample data, then determining the sample characteristic of which the distance from any marking result characteristic meets the preset condition, mapping the sample characteristic back to the initial training sample set according to the sample characteristic meeting the preset condition, and determining the corresponding sample data meeting the preset condition, namely, screening the sample data subset meeting the preset condition from the initial training sample set.
In the embodiment of the invention, the distance between the labeling result characteristic and the sample characteristic can be calculated by selecting an open source library faiss (open source aiming at a clustering and similarity search library).
In the embodiment of the present invention, determining the sample characteristics corresponding to each piece of sample data includes:
respectively determining M sub-features corresponding to the sample data by utilizing M preset models, wherein M is a positive integer greater than or equal to 1;
splicing the M sub-features to obtain a sample feature corresponding to each sample data; or, taking the M sub-features as M sample features corresponding to each piece of sample data.
In the embodiment of the present invention, the M preset models may select an image annotation model obtained by training an initial training sample set, and may also select one or more of other generalized open source models, such as a convolution network model of VGG, resnet, inceptionV3, and the like. The extraction of the labeling result characteristics and the sample characteristics can be extracted by adopting one image labeling model or a plurality of preset models.
In the embodiment of the invention, when the sample characteristics corresponding to each piece of sample data are determined, for each piece of sample data, each preset model can extract one sample sub-characteristic corresponding to the piece of sample data, and M preset models are adopted to respectively extract the sample sub-characteristics, so that M sample sub-characteristics can be obtained.
In an implementation manner of the embodiment of the present invention, after M sample sub-features corresponding to the piece of sample data are determined, each sample sub-feature is used as a sample feature corresponding to the piece of sample data, that is, M sample features may be obtained for one piece of sample data.
Similarly, when the labeling result feature corresponding to any one piece of labeling result data is determined, for each piece of labeling result data, each preset model can extract one labeling result sub-feature, M preset models are adopted to respectively extract M labeling result sub-features, and each labeling result sub-feature is used as the labeling result feature corresponding to the piece of sample data, that is, M labeling result features can be obtained for one piece of labeling result data.
In an implementation manner of the embodiment of the present invention, when determining sample features meeting preset conditions, for any one of the labeled result features, each preset model may determine N sample features meeting the preset conditions, and for M preset models, M × N sample features meeting the preset conditions may be determined, so that sample data corresponding to the M × N sample features may be used as sample data meeting the preset conditions, so as to determine a sample data subset, and then the initial training sample set is updated according to the sample data subset and the labeled result data set.
In another implementation manner of the embodiment of the present invention, after M sample sub-features corresponding to the sample data are determined, the M sample sub-features are spliced into one sample feature, which is used as a sample feature corresponding to the sample data; for example, if M is 3, for a certain sample data, the 3 sample sub-features extracted by the 3 preset models are (a, b), (c, d), and (e, f), and then the 3 sample sub-features may be concatenated into one sample feature (a, b, c, d, e, f) to obtain a sample data subset with higher similarity from the initial training sample set.
Similarly, M preset models are adopted to extract the marking result features, and for each piece of marking result data, each preset model can extract one marking result sub-feature, so that M marking result sub-features can be obtained, and the M marking result sub-features are spliced into one marking result feature to serve as the marking result feature corresponding to the piece of marking result data.
In another implementation manner of the embodiment of the present invention, when the sample features meeting the preset condition are determined, for one labeled result feature, N sample features meeting the preset condition may be determined, so that sample data corresponding to the N sample features meeting the preset condition may be used as sample data meeting the preset condition, thereby determining a sample data subset, and then updating the initial training sample set according to the sample data subset and the labeled result data set.
In the embodiment of the invention, the updated initial training sample set comprises the annotation result data set and the sample data subset, the performance of the image annotation model obtained by adopting the updated initial training sample set for training is better, and the accuracy of the annotation by adopting the image annotation model is higher.
Fig. 4 shows a method for processing image data according to an embodiment of the present invention, the method includes:
step S401: obtaining an initial training sample set;
step S402: training by using an initial training sample set to obtain an image annotation model;
step S403: acquiring a data set to be processed, and labeling the data set to be processed by using an image labeling model to obtain a labeling result data set;
step S404: manually correcting the marked result data set after sampling to obtain a corrected marked result data set;
step S405: extracting the marking result characteristics in the corrected marking result data set by using an image marking model, and extracting the sample characteristics in the initial training sample set;
step S406: calculating the cosine distance between each sample characteristic and any marking result characteristic aiming at any marking result characteristic, and determining sample data corresponding to the sample characteristic of which the cosine distance is smaller than a preset distance threshold value as the sample data meeting preset conditions;
step S407: constructing a sample data subset according to sample data which is corresponding to all the marking result characteristics and meets preset conditions;
step S408: manually correcting the sample data subset to obtain a corrected sample data subset;
step S409: and adding the corrected sample data subset and the corrected labeling result data set to obtain an updated initial training sample set, and returning to the step S401.
According to the image data processing method provided by the embodiment of the invention, an image labeling model obtained by training an initial training sample set is used for labeling a data set to be processed to obtain a labeling result data set; correcting the labeling result data set to realize correction of wrong labels in the labeling result data set; then, determining the similarity between the labeling result data set and the initial training sample set, and screening or searching out a sample data subset with the similarity meeting a preset condition from the initial training sample set; and updating the initial training sample set by using the labeling result data set and the initial training sample set. The method provided by the embodiment of the invention can efficiently update and correct the wrong labels in the initial training sample set, improve the labeling quality of the initial training sample set, further improve the performance of the image labeling model obtained by training, and improve the labeling accuracy of the image labeling model.
As shown in fig. 5, another aspect of the present invention provides an apparatus 500 for processing image data, including:
a model labeling module 501, labeling a to-be-processed data set by using an image labeling model obtained by training an initial training sample set, and obtaining a labeling result data set;
a determining module 502, configured to determine similarity between the labeling result data set and the initial training sample set;
and an updating module 503, for updating the initial training sample set according to the similarity.
In this embodiment of the present invention, the model labeling module 501 is further configured to: and after the labeling result data set is obtained, correcting the labeling result data set.
In this embodiment of the present invention, the determining module 502 is further configured to: and respectively calculating the similarity between the labeling result data and each sample data in the initial training sample set for any labeling result data in the labeling result data set.
In this embodiment of the present invention, the updating module 503 is further configured to: aiming at any marking result data, screening out a sample data subset which has the similarity with any marking result data and meets a preset condition from the initial training sample set; and updating the initial training sample set according to the sample data subsets corresponding to all the marked result data and the marked result data set.
In this embodiment of the present invention, the updating module 503 is further configured to: before updating the initial training sample set according to the sample data subsets corresponding to all the marked result data and the marked result data set, correcting the sample data meeting the preset conditions; and/or
And carrying out duplicate removal on the sample data meeting the preset condition.
In this embodiment of the present invention, the determining module 502 is further configured to: determining sample characteristics corresponding to each sample datum and labeling result characteristics corresponding to any labeling result datum; and determining the similarity between the labeling result data and each sample data according to the distance between the labeling result characteristics and each sample characteristic.
In this embodiment of the present invention, the determining module 502 is further configured to: respectively determining M sub-features corresponding to the sample data by utilizing M preset models, wherein M is a positive integer greater than or equal to 1; splicing the M sub-features to obtain a sample feature corresponding to each sample data; or, taking the M sub-features as M sample features corresponding to each piece of sample data.
In another aspect, an embodiment of the present invention provides an electronic device, including: one or more processors; a storage device for storing one or more programs which, when executed by one or more processors, cause the one or more processors to implement the method of image data processing of an embodiment of the present invention.
Embodiments of the present invention also provide a computer readable medium, on which a computer program is stored, where the computer program, when executed by a processor, implements a method for image data processing according to an embodiment of the present invention.
Fig. 6 shows an exemplary system architecture 600 of a method of image data processing or an apparatus of image data processing to which embodiments of the present invention may be applied.
As shown in fig. 6, the system architecture 600 may include terminal devices 601, 602, 603, a network 604, and a server 605. The network 604 serves as a medium for providing communication links between the terminal devices 601, 602, 603 and the server 605. Network 604 may include various types of connections, such as wire, wireless communication links, or fiber optic cables, to name a few.
A user may use the terminal devices 601, 602, 603 to interact with the server 605 via the network 604 to receive or send messages or the like. The terminal devices 601, 602, 603 may have various messaging client applications installed thereon, such as shopping applications, web browser applications, search applications, instant messaging tools, mailbox clients, social platform software, etc. (by way of example only).
The terminal devices 601, 602, 603 may be various electronic devices having a display screen and supporting web browsing, including but not limited to smart phones, tablet computers, laptop portable computers, desktop computers, and the like.
The server 605 may be a server providing various services, such as a background management server (for example only) providing support for shopping websites browsed by users using the terminal devices 601, 602, 603. The backend management server may analyze and perform other processing on the received data such as the product information query request, and feed back a processing result (for example, target push information, product information — just an example) to the terminal device.
It should be noted that the method for processing image data provided by the embodiment of the present invention is generally executed by the server 605, and accordingly, the apparatus for processing image data is generally disposed in the server 605.
It should be understood that the number of terminal devices, networks, and servers in fig. 6 is merely illustrative. There may be any number of terminal devices, networks, and servers, as desired for implementation.
Referring now to FIG. 7, shown is a block diagram of a computer system 700 suitable for use with a terminal device implementing an embodiment of the present invention. The terminal device shown in fig. 7 is only an example, and should not bring any limitation to the functions and the scope of use of the embodiments of the present invention.
As shown in fig. 7, the computer system 700 includes a Central Processing Unit (CPU)701, which can perform various appropriate actions and processes according to a program stored in a Read Only Memory (ROM)702 or a program loaded from a storage section 708 into a Random Access Memory (RAM) 703. In the RAM 703, various programs and data necessary for the operation of the system 700 are also stored. The CPU 701, the ROM 702, and the RAM 703 are connected to each other via a bus 704. An input/output (I/O) interface 705 is also connected to bus 704.
The following components are connected to the I/O interface 705: an input portion 706 including a keyboard, a mouse, and the like; an output section 707 including a display such as a Cathode Ray Tube (CRT), a Liquid Crystal Display (LCD), and the like, and a speaker; a storage section 708 including a hard disk and the like; and a communication section 709 including a network interface card such as a LAN card, a modem, or the like. The communication section 709 performs communication processing via a network such as the internet. A drive 710 is also connected to the I/O interface 705 as needed. A removable medium 711 such as a magnetic disk, an optical disk, a magneto-optical disk, a semiconductor memory, or the like is mounted on the drive 710 as necessary, so that a computer program read out therefrom is mounted into the storage section 708 as necessary.
In particular, according to the embodiments of the present disclosure, the processes described above with reference to the flowcharts may be implemented as computer software programs. For example, embodiments of the present disclosure include a computer program product comprising a computer program embodied on a computer readable medium, the computer program comprising program code for performing the method illustrated in the flow chart. In such an embodiment, the computer program can be downloaded and installed from a network through the communication section 709, and/or installed from the removable medium 711. The computer program performs the above-described functions defined in the system of the present invention when executed by the Central Processing Unit (CPU) 701.
It should be noted that the computer readable medium shown in the present invention can be a computer readable signal medium or a computer readable storage medium or any combination of the two. A computer readable storage medium may be, for example, but not limited to, an electronic, magnetic, optical, electromagnetic, infrared, or semiconductor system, apparatus, or device, or any combination of the foregoing. More specific examples of the computer readable storage medium may include, but are not limited to: an electrical connection having one or more wires, a portable computer diskette, a hard disk, a Random Access Memory (RAM), a read-only memory (ROM), an erasable programmable read-only memory (EPROM or flash memory), an optical fiber, a portable compact disc read-only memory (CD-ROM), an optical storage device, a magnetic storage device, or any suitable combination of the foregoing. In the present invention, a computer readable storage medium may be any tangible medium that can contain, or store a program for use by or in connection with an instruction execution system, apparatus, or device. In the present invention, however, a computer readable signal medium may include a propagated data signal with computer readable program code embodied therein, for example, in baseband or as part of a carrier wave. Such a propagated data signal may take many forms, including, but not limited to, electro-magnetic, optical, or any suitable combination thereof. A computer readable signal medium may also be any computer readable medium that is not a computer readable storage medium and that can communicate, propagate, or transport a program for use by or in connection with an instruction execution system, apparatus, or device. Program code embodied on a computer readable medium may be transmitted using any appropriate medium, including but not limited to: wireless, wire, fiber optic cable, RF, etc., or any suitable combination of the foregoing.
The flowchart and block diagrams in the figures illustrate the architecture, functionality, and operation of possible implementations of systems, methods and computer program products according to various embodiments of the present invention. In this regard, each block in the flowchart or block diagrams may represent a module, segment, or portion of code, which comprises one or more executable instructions for implementing the specified logical function(s). It should also be noted that, in some alternative implementations, the functions noted in the block may occur out of the order noted in the figures. For example, two blocks shown in succession may, in fact, be executed substantially concurrently, or the blocks may sometimes be executed in the reverse order, depending upon the functionality involved. It will also be noted that each block of the block diagrams or flowchart illustration, and combinations of blocks in the block diagrams or flowchart illustration, can be implemented by special purpose hardware-based systems which perform the specified functions or acts, or combinations of special purpose hardware and computer instructions.
The modules described in the embodiments of the present invention may be implemented by software or hardware. The described modules may also be provided in a processor, which may be described as: a processor includes a model labeling module, a determination module, and an update module. The names of these modules do not in some cases constitute a limitation on the module itself, for example, an update module may also be described as a "module that updates an initial training sample set according to similarity".
As another aspect, the present invention also provides a computer-readable medium that may be contained in the apparatus described in the above embodiments; or may be separate and not incorporated into the device. The computer readable medium carries one or more programs which, when executed by a device, cause the device to comprise: labeling a data set to be processed by using an image labeling model obtained by training an initial training sample set to obtain a labeling result data set; determining the similarity between the labeling result data set and the initial training sample set; and updating the initial training sample set according to the similarity.
According to the technical scheme of the embodiment of the invention, an image labeling model obtained by training an initial training sample set is used for labeling a data set to be processed to obtain a labeling result data set; correcting the labeling result data set to realize correction of wrong labels in the labeling result data set; then, determining the similarity between the labeling result data set and the initial training sample set, and screening out a sample data subset with the similarity meeting a preset condition from the initial training sample set; and updating the initial training sample set by using the labeling result data set and the initial training sample set. The method provided by the embodiment of the invention can efficiently update and correct the wrong labels in the initial training sample set, improve the labeling quality of the initial training sample set, further improve the performance of the image labeling model obtained by training, and improve the labeling accuracy of the image labeling model.
The above-described embodiments should not be construed as limiting the scope of the invention. Those skilled in the art will appreciate that various modifications, combinations, sub-combinations, and substitutions can occur, depending on design requirements and other factors. Any modification, equivalent replacement, and improvement made within the spirit and principle of the present invention should be included in the protection scope of the present invention.
Claims (10)
1. A method of image data processing, comprising:
labeling a data set to be processed by using an image labeling model obtained by training an initial training sample set to obtain a labeling result data set;
determining the similarity between the labeling result data set and the initial training sample set;
and updating the initial training sample set according to the similarity.
2. The method of claim 1, wherein after obtaining the annotated results dataset, further comprising: and correcting the labeling result data set.
3. The method of claim 1, wherein determining a similarity between the annotated result dataset and the initial training sample set comprises:
and for any one piece of labeling result data in the labeling result data set, respectively calculating the similarity between the labeling result data and each piece of sample data in the initial training sample set.
4. The method of claim 3, wherein updating the initial training sample set according to the similarity comprises:
aiming at any one piece of marked result data, screening out a sample data subset which has the similarity with any one piece of marked result data and meets a preset condition from the initial training sample set;
and updating the initial training sample set according to the sample data subsets corresponding to all the marked result data and the marked result data set.
5. The method according to claim 4, wherein before updating the initial training sample set according to the sample data subsets corresponding to all the labeled result data and the labeled result data set, further comprising: correcting the sample data meeting the preset conditions; and/or
And carrying out duplicate removal on the sample data meeting the preset condition.
6. The method of claim 3, wherein calculating the similarity between the labeling result data and each sample data in the initial training sample set comprises:
determining a sample characteristic corresponding to each piece of sample data and a labeling result characteristic corresponding to any labeling result data;
and determining the similarity between the labeling result data and each piece of sample data according to the distance between the labeling result characteristics and each piece of sample characteristics.
7. The method of claim 6, wherein determining the sample feature corresponding to each of the sample data comprises:
respectively determining M sub-features corresponding to the sample data by utilizing M preset models, wherein M is a positive integer greater than or equal to 1;
splicing the M sub-features to obtain a sample feature corresponding to each piece of sample data; or, the M sub-features are used as M sample features corresponding to each piece of sample data.
8. An apparatus for image data processing, comprising:
the model labeling module is used for labeling the data set to be processed by using an image labeling model obtained by training of the initial training sample set to obtain a labeling result data set;
the determining module is used for determining the similarity between the labeling result data set and the initial training sample set;
and the updating module is used for updating the initial training sample set according to the similarity.
9. An electronic device, comprising:
one or more processors;
a storage device to store one or more programs,
when executed by the one or more processors, cause the one or more processors to implement the method of any one of claims 1-7.
10. A computer-readable medium, on which a computer program is stored which, when being executed by a processor, carries out the method according to any one of claims 1-7.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202210235982.7A CN114612699A (en) | 2022-03-10 | 2022-03-10 | Image data processing method and device |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202210235982.7A CN114612699A (en) | 2022-03-10 | 2022-03-10 | Image data processing method and device |
Publications (1)
| Publication Number | Publication Date |
|---|---|
| CN114612699A true CN114612699A (en) | 2022-06-10 |
Family
ID=81863241
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN202210235982.7A Pending CN114612699A (en) | 2022-03-10 | 2022-03-10 | Image data processing method and device |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN114612699A (en) |
Citations (13)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| 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 |
| CN109492684A (en) * | 2018-10-31 | 2019-03-19 | 西安同瑞恒达电子科技有限公司 | Data processing method and device |
| CN110580290A (en) * | 2019-09-12 | 2019-12-17 | 北京小米智能科技有限公司 | method and device for optimizing training set for text classification |
| CN110689038A (en) * | 2019-06-25 | 2020-01-14 | 深圳市腾讯计算机***有限公司 | Training method and device of neural network model and medical image processing system |
| CN111104479A (en) * | 2019-11-13 | 2020-05-05 | 中国建设银行股份有限公司 | Data labeling method and device |
| US10719301B1 (en) * | 2018-10-26 | 2020-07-21 | Amazon Technologies, Inc. | Development environment for machine learning media models |
| CN111723870A (en) * | 2020-06-22 | 2020-09-29 | 中国平安人寿保险股份有限公司 | Data set acquisition method, device, equipment and medium based on artificial intelligence |
| CN112036491A (en) * | 2020-09-01 | 2020-12-04 | 北京推想科技有限公司 | Method and device for determining training sample and method for training deep learning model |
| CN112560993A (en) * | 2020-12-25 | 2021-03-26 | 北京百度网讯科技有限公司 | Data screening method and device, electronic equipment and storage medium |
| CN113064993A (en) * | 2021-03-23 | 2021-07-02 | 南京视察者智能科技有限公司 | Design method, optimization method and labeling method of automatic text classification labeling system based on big data |
| CN113095444A (en) * | 2021-06-07 | 2021-07-09 | 北京智芯微电子科技有限公司 | Image labeling method, device and storage medium |
| CN113673607A (en) * | 2021-08-24 | 2021-11-19 | 支付宝(杭州)信息技术有限公司 | Method and device for training image annotation model and image annotation |
| CN113780367A (en) * | 2021-08-19 | 2021-12-10 | 北京三快在线科技有限公司 | Classification model training and data classification method and device, and electronic equipment |
-
2022
- 2022-03-10 CN CN202210235982.7A patent/CN114612699A/en active Pending
Patent Citations (13)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| 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 |
| US10719301B1 (en) * | 2018-10-26 | 2020-07-21 | Amazon Technologies, Inc. | Development environment for machine learning media models |
| CN109492684A (en) * | 2018-10-31 | 2019-03-19 | 西安同瑞恒达电子科技有限公司 | Data processing method and device |
| CN110689038A (en) * | 2019-06-25 | 2020-01-14 | 深圳市腾讯计算机***有限公司 | Training method and device of neural network model and medical image processing system |
| CN110580290A (en) * | 2019-09-12 | 2019-12-17 | 北京小米智能科技有限公司 | method and device for optimizing training set for text classification |
| CN111104479A (en) * | 2019-11-13 | 2020-05-05 | 中国建设银行股份有限公司 | Data labeling method and device |
| CN111723870A (en) * | 2020-06-22 | 2020-09-29 | 中国平安人寿保险股份有限公司 | Data set acquisition method, device, equipment and medium based on artificial intelligence |
| CN112036491A (en) * | 2020-09-01 | 2020-12-04 | 北京推想科技有限公司 | Method and device for determining training sample and method for training deep learning model |
| CN112560993A (en) * | 2020-12-25 | 2021-03-26 | 北京百度网讯科技有限公司 | Data screening method and device, electronic equipment and storage medium |
| CN113064993A (en) * | 2021-03-23 | 2021-07-02 | 南京视察者智能科技有限公司 | Design method, optimization method and labeling method of automatic text classification labeling system based on big data |
| CN113095444A (en) * | 2021-06-07 | 2021-07-09 | 北京智芯微电子科技有限公司 | Image labeling method, device and storage medium |
| CN113780367A (en) * | 2021-08-19 | 2021-12-10 | 北京三快在线科技有限公司 | Classification model training and data classification method and device, and electronic equipment |
| CN113673607A (en) * | 2021-08-24 | 2021-11-19 | 支付宝(杭州)信息技术有限公司 | Method and device for training image annotation model and image annotation |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN107491547B (en) | Search method and device based on artificial intelligence | |
| CN112860866B (en) | Semantic retrieval method, device, equipment and storage medium | |
| CN108830329B (en) | Picture processing method and device | |
| CN109871311B (en) | Method and device for recommending test cases | |
| CN108959236B (en) | Medical literature classification model training method, medical literature classification method and device thereof | |
| US20210042470A1 (en) | Method and device for separating words | |
| CN111340220B (en) | Method and apparatus for training predictive models | |
| CN111104479A (en) | Data labeling method and device | |
| CN108121814B (en) | Search result ranking model generation method and device | |
| CN107908662B (en) | Method and device for realizing search system | |
| CN112988753B (en) | Data searching method and device | |
| CN110874532A (en) | Method and device for extracting keywords of feedback information | |
| CN113657113A (en) | Text processing method and device and electronic equipment | |
| CN111078849B (en) | Method and device for outputting information | |
| CN113268560A (en) | Method and device for text matching | |
| CN111368697A (en) | Information identification method and device | |
| CN111008213A (en) | Method and apparatus for generating language conversion model | |
| CN112712795B (en) | Labeling data determining method, labeling data determining device, labeling data determining medium and electronic equipment | |
| CN111435406A (en) | Method and device for correcting database statement spelling errors | |
| CN110852057A (en) | Method and device for calculating text similarity | |
| CN113590756A (en) | Information sequence generation method and device, terminal equipment and computer readable medium | |
| CN110807097A (en) | Method and device for analyzing data | |
| CN114612699A (en) | Image data processing method and device | |
| CN112148865B (en) | Information pushing method and device | |
| CN109308299B (en) | Method and apparatus for searching information |
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 |