CN114693814B - Decoding method, text recognition method, device, medium and equipment for model - Google Patents
Decoding method, text recognition method, device, medium and equipment for model Download PDFInfo
- Publication number
- CN114693814B CN114693814B CN202210346298.6A CN202210346298A CN114693814B CN 114693814 B CN114693814 B CN 114693814B CN 202210346298 A CN202210346298 A CN 202210346298A CN 114693814 B CN114693814 B CN 114693814B
- Authority
- CN
- China
- Prior art keywords
- attention
- decoding
- vector
- text
- current moment
- 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.)
- Active
Links
- 238000000034 method Methods 0.000 title claims abstract description 82
- 239000013598 vector Substances 0.000 claims abstract description 172
- 238000012545 processing Methods 0.000 claims description 49
- 239000011159 matrix material Substances 0.000 claims description 31
- 238000004590 computer program Methods 0.000 claims description 16
- 238000010606 normalization Methods 0.000 claims description 9
- 230000008569 process Effects 0.000 description 27
- 238000010586 diagram Methods 0.000 description 10
- 230000006870 function Effects 0.000 description 8
- 238000004891 communication Methods 0.000 description 6
- 230000003287 optical effect Effects 0.000 description 5
- 238000013528 artificial neural network Methods 0.000 description 4
- 230000000670 limiting effect Effects 0.000 description 4
- 238000012015 optical character recognition Methods 0.000 description 4
- 238000001514 detection method Methods 0.000 description 2
- 239000013307 optical fiber Substances 0.000 description 2
- 230000036961 partial effect Effects 0.000 description 2
- 230000000644 propagated effect Effects 0.000 description 2
- 230000002829 reductive effect Effects 0.000 description 2
- 230000004044 response Effects 0.000 description 2
- 239000004065 semiconductor Substances 0.000 description 2
- 238000004458 analytical method Methods 0.000 description 1
- 238000013475 authorization Methods 0.000 description 1
- 238000004364 calculation method Methods 0.000 description 1
- 230000008859 change Effects 0.000 description 1
- 238000005516 engineering process Methods 0.000 description 1
- 239000000835 fiber Substances 0.000 description 1
- 239000004973 liquid crystal related substance Substances 0.000 description 1
- 230000000873 masking effect Effects 0.000 description 1
- 230000002441 reversible effect Effects 0.000 description 1
- 238000012549 training Methods 0.000 description 1
- 238000012546 transfer Methods 0.000 description 1
Classifications
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
- G06T9/00—Image coding
- G06T9/001—Model-based coding, e.g. wire frame
-
- 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
- G06N3/044—Recurrent networks, e.g. Hopfield networks
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
- G06N3/00—Computing arrangements based on biological models
- G06N3/02—Neural networks
- G06N3/04—Architecture, e.g. interconnection topology
- G06N3/047—Probabilistic or stochastic networks
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
- G06T9/00—Image coding
- G06T9/002—Image coding using neural networks
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
- G06T2207/00—Indexing scheme for image analysis or image enhancement
- G06T2207/20—Special algorithmic details
- G06T2207/20084—Artificial neural networks [ANN]
Landscapes
- Engineering & Computer Science (AREA)
- Physics & Mathematics (AREA)
- Theoretical Computer Science (AREA)
- General Physics & Mathematics (AREA)
- Evolutionary Computation (AREA)
- Artificial Intelligence (AREA)
- Biomedical Technology (AREA)
- Data Mining & Analysis (AREA)
- Health & Medical Sciences (AREA)
- Life Sciences & Earth Sciences (AREA)
- Multimedia (AREA)
- Biophysics (AREA)
- Computational Linguistics (AREA)
- Software Systems (AREA)
- General Health & Medical Sciences (AREA)
- Molecular Biology (AREA)
- Computing Systems (AREA)
- General Engineering & Computer Science (AREA)
- Mathematical Physics (AREA)
- Probability & Statistics with Applications (AREA)
- Image Analysis (AREA)
Abstract
The disclosure relates to a decoding method, a text recognition method, a device, a medium and equipment of a model, wherein the method comprises the following steps: acquiring a coding vector corresponding to a text image to be identified; determining a mask vector corresponding to decoding of the current moment of the attention layer in a decoder, wherein the mask vector is used for representing the positioned position information of the attention layer, the text recognition model comprises an encoder and a decoder, and the encoder is used for encoding a received text image to be recognized to obtain the encoded vector; updating the attention distribution information of the current moment of the attention layer according to the mask vector to obtain a target attention weight corresponding to the current moment; and decoding according to the target attention weight and the coding vector to obtain a recognition result of the text recognition model. Therefore, the situation of attention drift can be avoided to a certain extent, so that character missing identification of a text image and repeated positioning of the same position are avoided.
Description
Technical Field
The present disclosure relates to the field of image processing, and in particular, to a method for decoding a model, a method for identifying a text, a device, a medium, and equipment.
Background
Optical character recognition (Optical Character Recognition, OCR) refers to the process of performing an analysis recognition process on an image file to obtain text information in the image file. OCR is generally divided into two processes, namely text detection and text recognition, wherein text information is required to be obtained by recognizing text region subgraphs segmented by a text detection module in the text recognition process.
In the related art, a corresponding text recognition model can be obtained by training based on a neural network, and because of insufficient definition of an image or repeated characters in a text in the image, attention drift can occur in the decoding process based on the text recognition model, so that the current character position cannot be accurately positioned to cause positioning deviation, and the problem of repeated recognition or missing recognition of the characters occurs.
Disclosure of Invention
This summary is provided to introduce a selection of concepts in a simplified form that are further described below in the detailed description. This summary is not intended to identify key features or essential features of the claimed subject matter, nor is it intended to be used to limit the scope of the claimed subject matter.
In a first aspect, the present disclosure provides a method for decoding a text recognition model, the method comprising:
acquiring a coding vector corresponding to a text image to be identified;
Determining a mask vector corresponding to decoding of the current moment of the attention layer in a decoder, wherein the mask vector is used for representing the positioned position information of the attention layer, the text recognition model comprises an encoder and a decoder, and the encoder is used for encoding a received text image to be recognized to obtain the encoded vector;
Updating the attention distribution information of the current moment of the attention layer according to the mask vector to obtain a target attention weight corresponding to the current moment;
and decoding according to the target attention weight and the coding vector to obtain a recognition result of the text recognition model.
In a second aspect, the present disclosure provides a text recognition method, the method comprising:
Receiving a text image to be identified;
Inputting the text image to be recognized into a text recognition model to obtain a recognition result of the text image to be recognized, wherein the text recognition model comprises an encoder and a decoder, the encoder is used for encoding the text image to be recognized to obtain an encoded vector, and the decoder is used for decoding the encoded vector according to the decoding method of the text recognition model in the first aspect to obtain the recognition result.
In a third aspect, the present disclosure provides a decoding apparatus of a text recognition model, the apparatus comprising:
the acquisition module is used for acquiring the coding vector corresponding to the text image to be identified;
The determining module is used for determining a mask vector corresponding to decoding of the current moment of the attention layer in the decoder, wherein the mask vector is used for representing the position information of the attention layer, the text recognition model comprises an encoder and a decoder, and the encoder is used for encoding the received text image to be recognized to obtain the encoded vector;
the updating module is used for updating the attention distribution information of the current moment of the attention layer according to the mask vector to obtain the target attention weight corresponding to the current moment;
And the decoding module is used for decoding according to the target attention weight and the coding vector to obtain the recognition result of the text recognition model.
In a fourth aspect, the present disclosure provides a text recognition apparatus, the apparatus comprising:
the receiving module is used for receiving the text image to be identified;
The processing module is used for inputting the text image to be recognized into a text recognition model to obtain a recognition result of the text image to be recognized, wherein the text recognition model comprises an encoder and a decoder, the encoder is used for encoding the text image to be recognized to obtain an encoded vector, and the decoder is used for decoding the encoded vector according to the decoding method of the text recognition model in the first aspect to obtain the recognition result.
In a fifth aspect, the present disclosure provides a computer readable medium having stored thereon a computer program which when executed by a processing device performs the steps of the method of the first or second aspect.
In a sixth aspect, the present disclosure provides an electronic device, comprising:
A storage device having a computer program stored thereon;
processing means for executing said computer program in said storage means to carry out the steps of the method of the first or second aspect.
Therefore, through the technical scheme, in the process of decoding by the decoder of the text recognition model, when the attention layer carries out attention processing, the attention distribution information can be updated according to the positioned position information in the decoding process, so that the attention weight at the current moment is generated, the accuracy of the obtained attention weight and the matching degree of the decoding process are improved, the situation of attention drift is avoided to a certain extent, the condition of character missing recognition of a text image and repeated positioning of the same position can be avoided, the accuracy of the recognition result of the text recognition is ensured, and the user experience is improved.
Additional features and advantages of the present disclosure will be set forth in the detailed description which follows.
Drawings
The above and other features, advantages, and aspects of embodiments of the present disclosure will become more apparent by reference to the following detailed description when taken in conjunction with the accompanying drawings. The same or similar reference numbers will be used throughout the drawings to refer to the same or like elements. It should be understood that the figures are schematic and that elements and components are not necessarily drawn to scale. In the drawings:
FIG. 1 is a flow chart of a method of decoding a text recognition model provided in accordance with one embodiment of the present disclosure;
FIG. 2 is a schematic diagram of a structure of a text recognition model provided in accordance with one embodiment of the present disclosure;
FIG. 3 is a schematic diagram of determining mask vectors provided in accordance with one embodiment of the present disclosure;
FIG. 4 is a block diagram of a decoding apparatus of a text recognition model provided in accordance with one embodiment of the present disclosure;
Fig. 5 shows a schematic structural diagram of an electronic device suitable for use in implementing embodiments of the present disclosure.
Detailed Description
Embodiments of the present disclosure will be described in more detail below with reference to the accompanying drawings. While certain embodiments of the present disclosure have been shown in the accompanying drawings, it is to be understood that the present disclosure may be embodied in various forms and should not be construed as limited to the embodiments set forth herein, but are provided to provide a more thorough and complete understanding of the present disclosure. It should be understood that the drawings and embodiments of the present disclosure are for illustration purposes only and are not intended to limit the scope of the present disclosure.
It should be understood that the various steps recited in the method embodiments of the present disclosure may be performed in a different order and/or performed in parallel. Furthermore, method embodiments may include additional steps and/or omit performing the illustrated steps. The scope of the present disclosure is not limited in this respect.
The term "including" and variations thereof as used herein are intended to be open-ended, i.e., including, but not limited to. The term "based on" is based at least in part on. The term "one embodiment" means "at least one embodiment"; the term "another embodiment" means "at least one additional embodiment"; the term "some embodiments" means "at least some embodiments. Related definitions of other terms will be given in the description below.
It should be noted that the terms "first," "second," and the like in this disclosure are merely used to distinguish between different devices, modules, or units and are not used to define an order or interdependence of functions performed by the devices, modules, or units.
It should be noted that references to "one", "a plurality" and "a plurality" in this disclosure are intended to be illustrative rather than limiting, and those of ordinary skill in the art will appreciate that "one or more" is intended to be understood as "one or more" unless the context clearly indicates otherwise.
The names of messages or information interacted between the various devices in the embodiments of the present disclosure are for illustrative purposes only and are not intended to limit the scope of such messages or information.
It will be appreciated that prior to using the technical solutions disclosed in the embodiments of the present disclosure, the user should be informed and authorized of the type, usage range, usage scenario, etc. of the personal information related to the present disclosure in an appropriate manner according to the relevant legal regulations.
For example, in response to receiving an active request from a user, a prompt is sent to the user to explicitly prompt the user that the operation it is requesting to perform will require personal information to be obtained and used with the user. Thus, the user can autonomously select whether to provide personal information to software or hardware such as an electronic device, an application program, a server or a storage medium for executing the operation of the technical scheme of the present disclosure according to the prompt information.
As an alternative but non-limiting implementation, in response to receiving an active request from a user, the manner in which the prompt information is sent to the user may be, for example, a popup, in which the prompt information may be presented in a text manner. In addition, a selection control for the user to select to provide personal information to the electronic device in a 'consent' or 'disagreement' manner can be carried in the popup window.
It will be appreciated that the above-described notification and user authorization process is merely illustrative and not limiting of the implementations of the present disclosure, and that other ways of satisfying relevant legal regulations may be applied to the implementations of the present disclosure.
Meanwhile, it can be understood that the data (including but not limited to the data itself, the acquisition or the use of the data) related to the technical scheme should conform to the requirements of the corresponding laws and regulations and related regulations.
Fig. 1 is a flowchart of a method for decoding a text recognition model according to an embodiment of the present disclosure, where the method may include:
in step 11, a coding vector corresponding to the text image to be identified is obtained.
In step 12, a mask vector corresponding to decoding of the current moment of the attention layer in the decoder is determined, wherein the mask vector is used for representing the position information of the attention layer, the text recognition model comprises an encoder and a decoder, and the encoder is used for encoding the received text image to be recognized to obtain the encoded vector.
The text recognition model may be implemented by a transducer model, as shown in fig. 2, which is a schematic structural diagram of the text recognition model, where the text recognition model may include an encoder Encoder and a Decoder, where A1 is used to represent the structure of the encoder, and may be implemented by an encoder in a transducer model in the art, which is not described herein, A2 is used to represent the structure of the Decoder, where a21 is used to represent an attention layer in the Decoder, and multi-headed attention in the drawing is an exemplary illustration, and the disclosure is not limited.
For example, during decoding by the decoder, the decoder decodes only one word at each step, the input of the decoder is the output of the encoder, and the output result of the decoder at the previous time i-1, that is, the output vector in fig. 2, then the attention layer performs attention feature processing to obtain attention features, and further the feedforward neural network obtains the probability distribution of the output word at the current time i, so as to determine the output result at the current time i, and further the output result is used as the input of the decoder at the next time i+1, and the above operations are repeated until the decoding is finished, so as to obtain the recognition result.
When the attention feature processing is performed in the attention layer to obtain the attention feature, the position where attention has been paid when the attention feature is generated at the current moment, that is, the position information where the attention layer is positioned, may be determined, and in this embodiment, the position information where attention has been paid to the attention layer may be explicitly characterized by the mask vector to provide data support for the subsequent decoding process.
In step 13, the attention distribution information of the attention layer is updated according to the mask vector, and the target attention weight corresponding to the current moment is obtained.
The input matrix X formed based on the input of the attention layer (i.e., the encoding vector and the output of the decoder at the previous time) may be linearly changed to obtain matrices Q (Query), K (Key) and V (Value), and then the matrix multiplication may be performed through the matrices Q and K T (transposed matrix of K) to obtain the similarity matrix. As an example, the similarity matrix may be used as the attention distribution information. As another example, each element in the similarity matrix may be divided byTo adjust for elements in the similarity matrix, where d k is the dimension of the matrix K. Accordingly, the adjusted similarity matrix may be used as the attention distribution information.
In this embodiment, the attention distribution information is an attention distribution feature determined based on the input feature at the current moment, and the mask vector is used to represent the already located position information of the attention layer, so that for the feature corresponding to such already located position information, attention should not be paid any more in the subsequent decoding process, and in this embodiment, the attention distribution information determined by the attention layer may be updated based on the mask vector, so as to avoid attention drift caused by repositioning the feature of the already located position information.
In step 14, decoding is performed according to the target attention weight and the encoding vector, and a recognition result of the text recognition model is obtained.
After determining the target attention weight, the target attention weight may be calculated by matrix multiplication with a matrix V generated based on the output of the encoder and the encoding vector at the previous time, that is, the input features of the attention layer are weighted and summed based on the target attention weight, to obtain the output features of the attention layer, that is, the attention features. And then, inputting the attention characteristic into a feedforward neural network in a decoder to perform linear change to obtain probability distribution of an output word corresponding to the current moment i, and further obtaining a recognition result of the current moment i. And obtaining the recognition result corresponding to the text image to be recognized by repeating the process until the decoding is finished. The structure and arrangement of the feedforward neural network may be based on network structures in a transducer model commonly used in the art, which is not limited in this disclosure.
Therefore, through the technical scheme, in the process of decoding by the decoder of the text recognition model, when the attention layer carries out attention processing, the attention distribution information can be updated according to the positioned position information in the decoding process, so that the attention weight at the current moment is generated, the accuracy of the obtained attention weight and the matching degree of the decoding process are improved, the situation of attention drift is avoided to a certain extent, the condition of character missing recognition of a text image and repeated positioning of the same position can be avoided, the accuracy of the recognition result of the text recognition is ensured, and the user experience is improved.
In one possible embodiment, an exemplary implementation of determining the mask vector corresponding to decoding the current moment of the attention layer in the decoder in step 12 is as follows, and the step may include:
And if the current moment is the first moment of decoding, determining a preset vector as the mask vector. When the current time is the first time of decoding, each position in the text image is not focused, and a preset vector which does not contain shielding information can be used as a mask vector. The representation form of the preset vector can be set according to the actual application scene, and the probability that the features of each position in the text image are focused is only required to be the same.
And if the current time is not the first time of decoding, acquiring a target attention weight corresponding to the decoding of the historical time before the current time, and determining a mask vector corresponding to the current time according to the target attention weight corresponding to the decoding of the historical time.
Illustratively, the current time is not the first time of decoding, meaning that there is a partial decoding step prior to the current time, then the previous decoding step must have focused on the feature that locates the partial position of the text image. And in the attention layer, the attention weight is used for representing the attention degree of the features of each position, accordingly, in the step, the target attention weight corresponding to the decoding of the historical moment before the current moment can be obtained, the features of which positions in the text image are paid attention to at all the historical moment can be determined based on the target attention weight corresponding to the historical moment, if the target attention weights corresponding to the historical moment can be respectively normalized and then accumulated, the positions with accumulated values larger than the threshold value can be considered to be paid attention to and positioned in the previous decoding process, and then the positions can be shielded to obtain the mask vector.
Therefore, through the technical scheme, when decoding is carried out at the current moment, the mask vector corresponding to the current moment can be determined by combining the previous target attention weights at each historical moment, the determined mask vector can be matched with the decoding process, the corresponding mask vector is determined based on the self-adaption of the decoding process, and the accuracy of the mask vector is ensured.
In one possible embodiment, an exemplary implementation manner of determining the mask vector corresponding to the current moment according to the target attention weight corresponding to the historical moment decoding is as follows, and the steps may include:
And carrying out binarization processing on the target attention weight corresponding to the decoding at the last moment of the current moment to obtain a binary feature vector.
The binarization processing may be performed based on a preset binarization threshold. For example, the value of the corresponding weight of the target attention weight greater than the binarization threshold is set to 1, and the value of the weight smaller than the binarization threshold is set to 0. For example, the binarization threshold may be set according to an actual application scenario, and may be set to a value in 0 to 1, for example, set to 0.6, and for the target attention weight [0.1,0.75,0.15], binarization processing may be performed to obtain a binary feature vector [0,1,0].
And inverting each element in the binary feature vector, and performing union processing on each element and the mask vector corresponding to the previous moment to obtain the mask vector corresponding to the current moment.
As shown in fig. 3, for example, the text image includes a text SSD, and the mask vector mask corresponding to the first time (T 0) is [1, 1], that is, there is no feature to be masked when decoding at the first time. When determining the mask vector corresponding to the time T 1, the target attention weight at the time T 0 is obtained, as shown in P 0, P 0 indicates that the feature corresponding to the first position is focused on in the first decoding, and the character decoded at the time T 0 is S.
Then, for time T 1, the binarization process is performed on P 0 to obtain a binary feature vector B 0 (where white represents 1 and black represents 0) at the time immediately before time T 1 (i.e., time T 0). And each element in the binary feature vector B 0 [1, 0] is inverted to obtain a vector B 0 '[0, 1], then the mask vector corresponding to the moment T 0 can be subjected to union processing of the vector B 0', namely, the mask vector [0, 1] corresponding to the moment T 1 is obtained by the union processing of the vector B0, 1], and then the feature of the first position, namely, the feature corresponding to the position with the value 0, can be not focused on when decoding is carried out at the moment T 1 based on the mask vector, so that the shielding of the feature of the positioned position is realized. The character decoded at time T 1 is illustratively S.
When determining the mask vector corresponding to the time T 2, the target attention weight at the time T 1 is obtained as described in [0.1,0.75,0.15], corresponding to P 1 in fig. 3, and binarized to obtain a binary feature vector B 1, i.e., [0,1,0], i.e., for the time T 2, the binary feature vector [0,1,0] corresponding to the time T 1, i.e., the feature of interest at the previous time is the feature corresponding to the position with the value of 1, and if it does not need to be focused at the current time, each element in the binary feature vector may be inverted, i.e., the vector B 1' [1,0,1] is obtained. In the actual application scenario, the feature not focused on at the previous moment may be the feature that has been shielded at the previous moment or the feature that has not been shielded but not focused on at the previous moment, then a mask vector corresponding to the previous moment may be further obtained to further determine which features have been shielded at the previous moment from the feature not focused on at the previous moment, which features have not been shielded but not focused on at the previous moment, then a mask vector corresponding to the previous moment T 1 may be further obtained as [0, 1], which indicates that the first position is a position shielded at the previous moment, then the current moment should also be shielded, and the mask vectors [0, 1] corresponding to the moments B 1' [1,0,1] and T 1 are processed together to obtain a mask vector [0, 1] corresponding to the moment T 2, then the character decoded at the moment T 2 is exemplified as D based on the mask vector which can be decoded at the moment T 2. Therefore, the method can shield the positioned position information after the first S is obtained through decoding, and can identify the repeated characters based on the characteristics which are not positioned later, so that the accurate identification of the repeated characters is realized, the accuracy of the attention characteristics is ensured, and the missing identification and the repeated identification of the characters in the text image are effectively reduced.
Therefore, through the technical scheme, the positions which are focused and positioned before can be shielded and combined in sequence in the decoding process, and when the mask vector corresponding to the current moment is determined, the positions which are focused and positioned at each historical moment can be determined by combining the target attention weight corresponding to the last moment and the mask vector at the last moment, so that the accuracy of the mask vector is ensured, the data quantity required by calculating the mask vector can be reduced, and the decoding efficiency of text recognition is improved.
In a possible embodiment, the updating the attention distribution information of the current moment of the attention layer according to the mask vector, and the obtaining the target attention weight corresponding to the current moment may include:
updating an attention distribution value corresponding to a position with a value of 0 in the mask vector in the attention distribution information to a preset value to obtain updated attention distribution information;
And carrying out normalization processing on the updated attention distribution information to obtain the target attention weight.
As described above, the attention distribution information may be a matrix obtained by multiplying a matrix Q by a matrix K, which represents a weight corresponding to each element in the matrix V, and masking may be performed based on a mask vector for features that have been focused on in the previous decoding process. The preset value may be set according to an actual application scenario, and may be set to a value smaller than a minimum value in the attention distribution information, for example, may be set to minus infinity-inf. And updating the attention distribution value corresponding to the position in the attention distribution information to a preset value, namely updating the attention distribution value to a negligible value, wherein the position with the value of 0 in the mask vector is the position already positioned at the decoding historical moment, so that the feature corresponding to the position does not participate in subsequent calculation.
The normalization processing may be performed by performing softmax processing on the updated attention distribution information, and when the softmax processing is performed on the updated attention distribution information, the value of the updated attention distribution information is-inf and mapped to 0, so as to realize shielding of the features of the corresponding position. Accordingly, when attention features are calculated based on the target attention weights in the following steps, features of positions which are focused at previous historical moments are shielded, and features of other positions which are not focused are focused only, so that repeated positioning of the features in the decoding process is avoided, decoding accuracy of text recognition is guaranteed, and accuracy and effectiveness of recognition results are further guaranteed.
In a possible embodiment, the updating the attention distribution information of the current moment of the attention layer according to the mask vector, and the obtaining the target attention weight corresponding to the current moment may include:
And performing matrix dot multiplication on the matrix of the mask vector and the matrix of the attention distribution information to obtain updated attention distribution information.
The matrix point multiplication is to multiply corresponding elements in the two matrices to obtain the matrix. For example, if the mask vector is [0,1], and the attention distribution information is [1,2,3], the mask vector and the attention distribution information are subjected to matrix dot multiplication to obtain updated attention distribution information which is [0,2,3], so that the feature of the first position is shielded.
And carrying out normalization processing on the updated attention distribution information to obtain the target attention weight.
Accordingly, the normalization processing may be performed by performing softmax processing on the updated attention distribution information, so as to obtain the target attention weight, and by using an exponential form softmax function, the numerical distance with a large difference can be pulled more, so that the attention weight among the features is more highlighted. According to the technical scheme, updated attention distribution information can be directly calculated based on the matrix corresponding to the mask vector and the attention distribution information, and the data processing efficiency is improved, so that accurate and effective data support is provided for follow-up attention feature determination and decoding, repeated feature positioning in the decoding process is avoided, the decoding accuracy of text recognition is ensured, and the accuracy and the effectiveness of recognition results are further ensured.
The present disclosure also provides a text recognition method, which may include:
the text image to be recognized is received, namely, an image input by a user, and also an image obtained from other applications.
Inputting the text image to be recognized into a text recognition model to obtain a recognition result of the text image to be recognized, wherein the text recognition model comprises an encoder and a decoder, the encoder is used for encoding the text image to be recognized to obtain an encoded vector, and the decoder is used for decoding the encoded vector according to the decoding method of the text recognition model to obtain the recognition result.
Therefore, through the technical scheme, in the process of recognizing the text image to be recognized by the text recognition model, the attention weight decoded at the current moment can be determined according to the positioned position information in the decoding process, so that the situation of attention drift is avoided to a certain extent, the condition of missing recognition of characters in the text image or repeated positioning of the same position is avoided, the accuracy of the recognition result of text recognition is ensured, and the user experience is improved.
The present disclosure also provides a decoding apparatus of a text recognition model, as shown in fig. 4, the apparatus 10 includes:
the acquiring module 100 is configured to acquire a coding vector corresponding to a text image to be identified;
A determining module 200, configured to determine a mask vector corresponding to decoding of a current moment of an attention layer in a decoder, where the mask vector is used to represent location information of the attention layer located, the text recognition model includes an encoder and a decoder, and the encoder is configured to encode a received text image to be recognized, and obtain the encoded vector;
the updating module 300 is configured to update the attention distribution information of the current moment of the attention layer according to the mask vector, so as to obtain a target attention weight corresponding to the current moment;
And the decoding module 400 is configured to decode according to the target attention weight and the encoding vector, and obtain a recognition result of the text recognition model.
Optionally, the determining module includes:
The first determining submodule is used for determining a preset vector as the mask vector if the current moment is the first moment of decoding;
And the second determining submodule is used for acquiring a target attention weight corresponding to the decoding of the historical moment before the current moment if the current moment is not the first moment of decoding, and determining a mask vector corresponding to the current moment according to the target attention weight corresponding to the decoding of the historical moment.
Optionally, the second determining submodule includes:
the first processing sub-module is used for carrying out binarization processing on the target attention weight corresponding to the decoding at the last moment of the current moment to obtain a binary feature vector;
And the second processing sub-module is used for inverting each element in the binary characteristic vector, and carrying out union processing on the element and the mask vector corresponding to the last moment to obtain the mask vector corresponding to the current moment.
Optionally, the updating module includes:
a first updating sub-module, configured to update an attention distribution value corresponding to a position with a value of 0 in the mask vector in the attention distribution information to a preset value, and obtain updated attention distribution information;
and the third processing sub-module is used for carrying out normalization processing on the updated attention distribution information to obtain the target attention weight.
Optionally, the updating module includes:
the second updating sub-module is used for performing matrix dot multiplication on the matrix of the mask vector and the matrix of the attention distribution information to obtain updated attention distribution information;
and a fourth processing sub-module, configured to normalize the updated attention distribution information, and obtain the target attention weight.
The present disclosure also provides a text recognition apparatus, the apparatus comprising:
the receiving module is used for receiving the text image to be identified;
The processing module is used for inputting the text image to be identified into a text identification model to obtain an identification result of the text image to be identified, wherein the text identification model comprises an encoder and a decoder, the encoder is used for encoding the text image to be identified to obtain an encoded vector, and the decoder is used for decoding the encoded vector according to the decoding method of the text identification model to obtain the identification result.
Referring now to fig. 5, a schematic diagram of an electronic device 600 suitable for use in implementing embodiments of the present disclosure is shown. The terminal devices in the embodiments of the present disclosure may include, but are not limited to, mobile terminals such as mobile phones, notebook computers, digital broadcast receivers, PDAs (personal digital assistants), PADs (tablet computers), PMPs (portable multimedia players), in-vehicle terminals (e.g., in-vehicle navigation terminals), and the like, and stationary terminals such as digital TVs, desktop computers, and the like. The electronic device shown in fig. 5 is merely an example and should not be construed to limit the functionality and scope of use of the disclosed embodiments.
As shown in fig. 5, the electronic device 600 may include a processing means (e.g., a central processing unit, a graphics processor, etc.) 601, which may perform various appropriate actions and processes according to a program stored in a Read Only Memory (ROM) 602 or a program loaded from a storage means 608 into a Random Access Memory (RAM) 603. In the RAM 603, various programs and data required for the operation of the electronic apparatus 600 are also stored. The processing device 601, the ROM 602, and the RAM 603 are connected to each other through a bus 604. An input/output (I/O) interface 605 is also connected to bus 604.
In general, the following devices may be connected to the I/O interface 605: input devices 606 including, for example, a touch screen, touchpad, keyboard, mouse, camera, microphone, accelerometer, gyroscope, and the like; an output device 607 including, for example, a Liquid Crystal Display (LCD), a speaker, a vibrator, and the like; storage 608 including, for example, magnetic tape, hard disk, etc.; and a communication device 609. The communication means 609 may allow the electronic device 600 to communicate with other devices wirelessly or by wire to exchange data. While fig. 5 shows an electronic device 600 having various means, it is to be understood that not all of the illustrated means are required to be implemented or provided. More or fewer devices may be implemented or provided instead.
In particular, according to embodiments of the present disclosure, the processes described above with reference to 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 non-transitory computer readable medium, the computer program comprising program code for performing the method shown in the flow chart. In such an embodiment, the computer program may be downloaded and installed from a network via communication means 609, or from storage means 608, or from ROM 602. The above-described functions defined in the methods of the embodiments of the present disclosure are performed when the computer program is executed by the processing device 601.
It should be noted that the computer readable medium described in the present disclosure may be a computer readable signal medium or a computer readable storage medium, or any combination of the two. The computer readable storage medium can be, for example, but not limited to, an electronic, magnetic, optical, electromagnetic, infrared, or semiconductor system, apparatus, or device, or a combination of any 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 context of this disclosure, 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 disclosure, however, the computer-readable signal medium may include a data signal propagated in baseband or as part of a carrier wave, with the computer-readable program code embodied therein. Such a propagated data signal may take any of a variety of forms, including, but not limited to, electro-magnetic, optical, or any suitable combination of the foregoing. 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: electrical wires, fiber optic cables, RF (radio frequency), and the like, or any suitable combination of the foregoing.
In some embodiments, the clients, servers may communicate using any currently known or future developed network protocol, such as HTTP (HyperText Transfer Protocol ), and may be interconnected with any form or medium of digital data communication (e.g., a communication network). Examples of communication networks include a local area network ("LAN"), a wide area network ("WAN"), the internet (e.g., the internet), and peer-to-peer networks (e.g., ad hoc peer-to-peer networks), as well as any currently known or future developed networks.
The computer readable medium may be contained in the electronic device; or may exist alone without being incorporated into the electronic device.
The computer readable medium carries one or more programs which, when executed by the electronic device, cause the electronic device to: acquiring a coding vector corresponding to a text image to be identified; determining a mask vector corresponding to decoding of the current moment of the attention layer in a decoder, wherein the mask vector is used for representing the positioned position information of the attention layer, the text recognition model comprises an encoder and a decoder, and the encoder is used for encoding a received text image to be recognized to obtain the encoded vector; updating the attention distribution information of the current moment of the attention layer according to the mask vector to obtain a target attention weight corresponding to the current moment; and decoding according to the target attention weight and the coding vector to obtain a recognition result of the text recognition model.
Or the computer readable medium carries one or more programs which, when executed by the electronic device, cause the electronic device to: receiving a text image to be identified; inputting the text image to be recognized into a text recognition model to obtain a recognition result of the text image to be recognized, wherein the text recognition model comprises an encoder and a decoder, the encoder is used for encoding the text image to be recognized to obtain an encoded vector, and the decoder is used for decoding the encoded vector according to the decoding method of the text recognition model to obtain the recognition result.
Computer program code for carrying out operations of the present disclosure may be written in one or more programming languages, including, but not limited to, an object oriented programming language such as Java, smalltalk, C ++ and conventional procedural programming languages, such as the "C" programming language or similar programming languages. The program code may execute entirely on the user's computer, partly on the user's computer, as a stand-alone software package, partly on the user's computer and partly on a remote computer or entirely on the remote computer or server. In the case of a remote computer, the remote computer may be connected to the user's computer through any kind of network, including a Local Area Network (LAN) or a Wide Area Network (WAN), or may be connected to an external computer (for example, through the Internet using an Internet service provider).
The flowcharts 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 disclosure. 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 and/or flowchart illustration, and combinations of blocks in the block diagrams and/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 disclosure may be implemented in software or hardware. The name of the module is not limited to the module itself in some cases, and for example, the acquisition module may also be described as "a module for acquiring a code vector corresponding to a text image to be identified".
The functions described above herein may be performed, at least in part, by one or more hardware logic components. For example, without limitation, exemplary types of hardware logic components that may be used include: a Field Programmable Gate Array (FPGA), an Application Specific Integrated Circuit (ASIC), an Application Specific Standard Product (ASSP), a system on a chip (SOC), a Complex Programmable Logic Device (CPLD), and the like.
In the context of this disclosure, a machine-readable medium may be a tangible medium that can contain, or store a program for use by or in connection with an instruction execution system, apparatus, or device. The machine-readable medium may be a machine-readable signal medium or a machine-readable storage medium. The machine-readable medium may include, but is not limited to, an electronic, magnetic, optical, electromagnetic, infrared, or semiconductor system, apparatus, or device, or any suitable combination of the foregoing. More specific examples of a machine-readable storage medium would include an electrical connection based on 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.
According to one or more embodiments of the present disclosure, example 1 provides a method of decoding a text recognition model, wherein the method comprises:
acquiring a coding vector corresponding to a text image to be identified;
Determining a mask vector corresponding to decoding of the current moment of the attention layer in a decoder, wherein the mask vector is used for representing the positioned position information of the attention layer, the text recognition model comprises an encoder and a decoder, and the encoder is used for encoding a received text image to be recognized to obtain the encoded vector;
Updating the attention distribution information of the current moment of the attention layer according to the mask vector to obtain a target attention weight corresponding to the current moment;
and decoding according to the target attention weight and the coding vector to obtain a recognition result of the text recognition model.
According to one or more embodiments of the present disclosure, example 2 provides the method of example 1, wherein the determining the attention layer current time in the decoder decodes the corresponding mask vector, including:
if the current moment is the first moment of decoding, determining a preset vector as the mask vector;
And if the current time is not the first time of decoding, acquiring a target attention weight corresponding to the decoding of the historical time before the current time, and determining a mask vector corresponding to the current time according to the target attention weight corresponding to the decoding of the historical time.
According to one or more embodiments of the present disclosure, example 3 provides the method of example 2, wherein the determining the mask vector corresponding to the current time according to the target attention weight corresponding to the historical time decoding includes:
performing binarization processing on the target attention weight corresponding to the decoding at the last moment of the current moment to obtain a binary feature vector;
And inverting each element in the binary feature vector, and performing union processing on each element and the mask vector corresponding to the previous moment to obtain the mask vector corresponding to the current moment.
According to one or more embodiments of the present disclosure, example 4 provides the method of example 1, wherein the updating, according to the mask vector, the attention distribution information of the current moment of the attention layer to obtain the target attention weight corresponding to the current moment includes:
updating an attention distribution value corresponding to a position with a value of 0 in the mask vector in the attention distribution information to a preset value to obtain updated attention distribution information;
And carrying out normalization processing on the updated attention distribution information to obtain the target attention weight.
According to one or more embodiments of the present disclosure, example 5 provides the method of example 1, wherein the updating, according to the mask vector, the attention distribution information of the current moment of the attention layer to obtain the target attention weight corresponding to the current moment includes:
Performing matrix dot multiplication on the matrix of the mask vector and the matrix of the attention distribution information to obtain updated attention distribution information;
And carrying out normalization processing on the updated attention distribution information to obtain the target attention weight.
Example 6 provides a text recognition method according to one or more embodiments of the present disclosure, wherein the method comprises:
Receiving a text image to be identified;
Inputting the text image to be recognized into a text recognition model to obtain a recognition result of the text image to be recognized, wherein the text recognition model comprises an encoder and a decoder, the encoder is used for encoding the text image to be recognized to obtain an encoded vector, and the decoder is used for decoding the encoded vector according to the decoding method of the text recognition model in any one of examples 1-5 to obtain the recognition result.
Example 7 provides a decoding apparatus of a text recognition model, according to one or more embodiments of the present disclosure, wherein the apparatus comprises:
the acquisition module is used for acquiring the coding vector corresponding to the text image to be identified;
The determining module is used for determining a mask vector corresponding to decoding of the current moment of the attention layer in the decoder, wherein the mask vector is used for representing the position information of the attention layer, the text recognition model comprises an encoder and a decoder, and the encoder is used for encoding the received text image to be recognized to obtain the encoded vector;
the updating module is used for updating the attention distribution information of the current moment of the attention layer according to the mask vector to obtain the target attention weight corresponding to the current moment;
And the decoding module is used for decoding according to the target attention weight and the coding vector to obtain the recognition result of the text recognition model.
According to one or more embodiments of the present disclosure, example 8 provides a text recognition apparatus, wherein the apparatus comprises:
the receiving module is used for receiving the text image to be identified;
The processing module is configured to input the text image to be identified into a text identification model to obtain an identification result of the text image to be identified, where the text identification model includes an encoder and a decoder, the encoder is configured to encode the text image to be identified to obtain an encoded vector, and the decoder is configured to decode the encoded vector according to the decoding method of the text identification model in any one of examples 1-5 to obtain the identification result.
According to one or more embodiments of the present disclosure, example 9 provides a computer-readable medium having stored thereon a computer program, wherein the program when executed by a processing device implements the steps of the method of any of examples 1-6.
Example 10 provides an electronic device according to one or more embodiments of the present disclosure, including:
A storage device having a computer program stored thereon;
processing means for executing the computer program in the storage means to implement the steps of the method of any one of examples 1-6.
The foregoing description is only of the preferred embodiments of the present disclosure and description of the principles of the technology being employed. It will be appreciated by persons skilled in the art that the scope of the disclosure referred to in this disclosure is not limited to the specific combinations of features described above, but also covers other embodiments which may be formed by any combination of features described above or equivalents thereof without departing from the spirit of the disclosure. Such as those described above, are mutually substituted with the technical features having similar functions disclosed in the present disclosure (but not limited thereto).
Moreover, although operations are depicted in a particular order, this should not be understood as requiring that such operations be performed in the particular order shown or in sequential order. In certain circumstances, multitasking and parallel processing may be advantageous. Likewise, while several specific implementation details are included in the above discussion, these should not be construed as limiting the scope of the present disclosure. Certain features that are described in the context of separate embodiments can also be implemented in combination in a single embodiment. Conversely, various features that are described in the context of a single embodiment can also be implemented in multiple embodiments separately or in any suitable subcombination.
Although the subject matter has been described in language specific to structural features and/or methodological acts, it is to be understood that the subject matter defined in the appended claims is not necessarily limited to the specific features or acts described above. Rather, the specific features and acts described above are example forms of implementing the claims. The specific manner in which the various modules perform the operations in the apparatus of the above embodiments have been described in detail in connection with the embodiments of the method, and will not be described in detail herein.
Claims (10)
1. A method of decoding a text recognition model, the method comprising:
acquiring a coding vector corresponding to a text image to be identified;
Determining a mask vector corresponding to decoding of the current moment of the attention layer in a decoder, wherein the mask vector is used for representing the positioned position information of the attention layer, the text recognition model comprises an encoder and a decoder, and the encoder is used for encoding a received text image to be recognized to obtain the encoded vector;
Updating the attention distribution information of the current moment of the attention layer according to the mask vector to obtain a target attention weight corresponding to the current moment;
and decoding according to the target attention weight and the coding vector to obtain a recognition result of the text recognition model.
2. The method of claim 1, wherein determining the current moment of the attention layer in the decoder for decoding the corresponding mask vector comprises:
if the current moment is the first moment of decoding, determining a preset vector as the mask vector;
And if the current time is not the first time of decoding, acquiring a target attention weight corresponding to the decoding of the historical time before the current time, and determining a mask vector corresponding to the current time according to the target attention weight corresponding to the decoding of the historical time.
3. The method according to claim 2, wherein said determining a mask vector corresponding to the current time according to the target attention weight corresponding to the historical time decoding comprises:
performing binarization processing on the target attention weight corresponding to the decoding at the last moment of the current moment to obtain a binary feature vector;
And inverting each element in the binary feature vector, and performing union processing on each element and the mask vector corresponding to the previous moment to obtain the mask vector corresponding to the current moment.
4. The method according to claim 1, wherein updating the attention distribution information of the current moment of the attention layer according to the mask vector to obtain the target attention weight corresponding to the current moment comprises:
updating an attention distribution value corresponding to a position with a value of 0 in the mask vector in the attention distribution information to a preset value to obtain updated attention distribution information;
And carrying out normalization processing on the updated attention distribution information to obtain the target attention weight.
5. The method according to claim 1, wherein updating the attention distribution information of the current moment of the attention layer according to the mask vector to obtain the target attention weight corresponding to the current moment comprises:
Performing matrix dot multiplication on the matrix of the mask vector and the matrix of the attention distribution information to obtain updated attention distribution information;
And carrying out normalization processing on the updated attention distribution information to obtain the target attention weight.
6. A method of text recognition, the method comprising:
Receiving a text image to be identified;
Inputting the text image to be recognized into a text recognition model to obtain a recognition result of the text image to be recognized, wherein the text recognition model comprises an encoder and a decoder, the encoder is used for encoding the text image to be recognized to obtain an encoded vector, and the decoder is used for decoding the encoded vector according to the decoding method of the text recognition model of any one of claims 1-5 to obtain the recognition result.
7. A decoding device for a text recognition model, the device comprising:
the acquisition module is used for acquiring the coding vector corresponding to the text image to be identified;
The determining module is used for determining a mask vector corresponding to decoding of the current moment of the attention layer in the decoder, wherein the mask vector is used for representing the position information of the attention layer, the text recognition model comprises an encoder and a decoder, and the encoder is used for encoding the received text image to be recognized to obtain the encoded vector;
the updating module is used for updating the attention distribution information of the current moment of the attention layer according to the mask vector to obtain the target attention weight corresponding to the current moment;
And the decoding module is used for decoding according to the target attention weight and the coding vector to obtain the recognition result of the text recognition model.
8. A text recognition device, the device comprising:
the receiving module is used for receiving the text image to be identified;
A processing module, configured to input the text image to be identified into a text identification model to obtain an identification result of the text image to be identified, where the text identification model includes an encoder and a decoder, the encoder is configured to encode the text image to be identified to obtain an encoded vector, and the decoder is configured to decode the encoded vector according to the decoding method of the text identification model according to any one of claims 1-5 to obtain the identification result.
9. A computer readable medium on which a computer program is stored, characterized in that the program, when being executed by a processing device, carries out the steps of the method according to any one of claims 1-6.
10. An electronic device, comprising:
A storage device having a computer program stored thereon;
Processing means for executing said computer program in said storage means to carry out the steps of the method according to any one of claims 1-6.
Priority Applications (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN202210346298.6A CN114693814B (en) | 2022-03-31 | 2022-03-31 | Decoding method, text recognition method, device, medium and equipment for model |
Applications Claiming Priority (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN202210346298.6A CN114693814B (en) | 2022-03-31 | 2022-03-31 | Decoding method, text recognition method, device, medium and equipment for model |
Publications (2)
Publication Number | Publication Date |
---|---|
CN114693814A CN114693814A (en) | 2022-07-01 |
CN114693814B true CN114693814B (en) | 2024-04-30 |
Family
ID=82140566
Family Applications (1)
Application Number | Title | Priority Date | Filing Date |
---|---|---|---|
CN202210346298.6A Active CN114693814B (en) | 2022-03-31 | 2022-03-31 | Decoding method, text recognition method, device, medium and equipment for model |
Country Status (1)
Country | Link |
---|---|
CN (1) | CN114693814B (en) |
Citations (4)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
JP2005260373A (en) * | 2004-03-09 | 2005-09-22 | Ricoh Co Ltd | Image decoder, image decoding method, program and information recording medium |
CN111507328A (en) * | 2020-04-13 | 2020-08-07 | 北京爱咔咔信息技术有限公司 | Text recognition and model training method, system, equipment and readable storage medium |
CN113052156A (en) * | 2021-03-12 | 2021-06-29 | 北京百度网讯科技有限公司 | Optical character recognition method, device, electronic equipment and storage medium |
KR102272101B1 (en) * | 2021-03-24 | 2021-07-01 | 인천대학교 산학협력단 | Graphic processing apparatus capable of generating a mesh structure through learning of a pointer network for delaunay triangulation and the operating method thereof |
Family Cites Families (2)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
US10839259B2 (en) * | 2017-10-27 | 2020-11-17 | Google Llc | Attention-based image generation neural networks |
CN111832292B (en) * | 2020-06-03 | 2024-02-02 | 北京百度网讯科技有限公司 | Text recognition processing method, device, electronic equipment and storage medium |
-
2022
- 2022-03-31 CN CN202210346298.6A patent/CN114693814B/en active Active
Patent Citations (4)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
JP2005260373A (en) * | 2004-03-09 | 2005-09-22 | Ricoh Co Ltd | Image decoder, image decoding method, program and information recording medium |
CN111507328A (en) * | 2020-04-13 | 2020-08-07 | 北京爱咔咔信息技术有限公司 | Text recognition and model training method, system, equipment and readable storage medium |
CN113052156A (en) * | 2021-03-12 | 2021-06-29 | 北京百度网讯科技有限公司 | Optical character recognition method, device, electronic equipment and storage medium |
KR102272101B1 (en) * | 2021-03-24 | 2021-07-01 | 인천대학교 산학협력단 | Graphic processing apparatus capable of generating a mesh structure through learning of a pointer network for delaunay triangulation and the operating method thereof |
Also Published As
Publication number | Publication date |
---|---|
CN114693814A (en) | 2022-07-01 |
Similar Documents
Publication | Publication Date | Title |
---|---|---|
CN113436620B (en) | Training method of voice recognition model, voice recognition method, device, medium and equipment | |
CN112464921B (en) | Obstacle detection information generation method, apparatus, device and computer readable medium | |
CN113327599B (en) | Voice recognition method, device, medium and electronic equipment | |
CN116166271A (en) | Code generation method and device, storage medium and electronic equipment | |
CN114692085A (en) | Feature extraction method and device, storage medium and electronic equipment | |
WO2022012178A1 (en) | Method for generating objective function, apparatus, electronic device and computer readable medium | |
CN114067327A (en) | Text recognition method and device, readable medium and electronic equipment | |
CN114693814B (en) | Decoding method, text recognition method, device, medium and equipment for model | |
CN115984868A (en) | Text processing method, device, medium and equipment | |
CN116072108A (en) | Model generation method, voice recognition method, device, medium and equipment | |
CN116258911A (en) | Training method, device, equipment and storage medium for image classification model | |
CN116092092A (en) | Matching method, device, medium and electronic equipment | |
CN114625876B (en) | Method for generating author characteristic model, method and device for processing author information | |
CN115662517A (en) | Sequence prediction method, device, medium, and electronic apparatus | |
CN113986958B (en) | Text information conversion method and device, readable medium and electronic equipment | |
CN115269978A (en) | Video tag generation method, device, equipment and medium | |
CN111582456B (en) | Method, apparatus, device and medium for generating network model information | |
CN114445813A (en) | Character recognition method, device, equipment and medium | |
CN114429629A (en) | Image processing method and device, readable storage medium and electronic equipment | |
CN115938470B (en) | Protein characteristic pretreatment method, device, medium and equipment | |
CN116343905B (en) | Pretreatment method, pretreatment device, pretreatment medium and pretreatment equipment for protein characteristics | |
CN117349397A (en) | Address recognition method, address recognition device, medium and electronic equipment | |
CN111933122B (en) | Speech recognition method, apparatus, electronic device, and computer-readable medium | |
US20240233334A1 (en) | Multi-modal data retrieval method and apparatus, medium, and electronic device | |
CN116167433A (en) | Method for training image feature extraction model, image recognition method and medium |
Legal Events
Date | Code | Title | Description |
---|---|---|---|
PB01 | Publication | ||
PB01 | Publication | ||
SE01 | Entry into force of request for substantive examination | ||
SE01 | Entry into force of request for substantive examination | ||
GR01 | Patent grant | ||
GR01 | Patent grant |