CN114596546A - Vehicle weight recognition method and device, computer and readable storage medium - Google Patents
Vehicle weight recognition method and device, computer and readable storage medium Download PDFInfo
- Publication number
- CN114596546A CN114596546A CN202210033507.1A CN202210033507A CN114596546A CN 114596546 A CN114596546 A CN 114596546A CN 202210033507 A CN202210033507 A CN 202210033507A CN 114596546 A CN114596546 A CN 114596546A
- Authority
- CN
- China
- Prior art keywords
- vehicle
- image
- training
- model
- knowledge distillation
- Prior art date
- Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
- Pending
Links
- 238000000034 method Methods 0.000 title claims abstract description 39
- 238000013140 knowledge distillation Methods 0.000 claims abstract description 55
- 230000008569 process Effects 0.000 claims abstract description 10
- 238000004364 calculation method Methods 0.000 claims abstract description 9
- 238000004590 computer program Methods 0.000 claims description 9
- 238000010586 diagram Methods 0.000 claims description 8
- 238000010276 construction Methods 0.000 claims description 5
- 238000000605 extraction Methods 0.000 claims description 4
- 238000005070 sampling Methods 0.000 claims description 3
- 238000005516 engineering process Methods 0.000 description 3
- 230000006870 function Effects 0.000 description 3
- 238000013528 artificial neural network Methods 0.000 description 2
- 238000004821 distillation Methods 0.000 description 2
- 230000009286 beneficial effect Effects 0.000 description 1
- 238000004422 calculation algorithm Methods 0.000 description 1
- 230000006835 compression Effects 0.000 description 1
- 238000007906 compression Methods 0.000 description 1
- 238000012986 modification Methods 0.000 description 1
- 230000004048 modification Effects 0.000 description 1
- 238000010606 normalization Methods 0.000 description 1
- 238000011176 pooling Methods 0.000 description 1
- 239000007787 solid Substances 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
- G06F18/21—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
- 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
- G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
- G06N3/00—Computing arrangements based on biological models
- G06N3/02—Neural networks
- G06N3/04—Architecture, e.g. interconnection topology
- G06N3/045—Combinations of networks
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
- G06N3/00—Computing arrangements based on biological models
- G06N3/02—Neural networks
- G06N3/04—Architecture, e.g. interconnection topology
- G06N3/047—Probabilistic or stochastic networks
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
- G06N3/00—Computing arrangements based on biological models
- G06N3/02—Neural networks
- G06N3/08—Learning methods
-
- Y—GENERAL TAGGING OF NEW TECHNOLOGICAL DEVELOPMENTS; GENERAL TAGGING OF CROSS-SECTIONAL TECHNOLOGIES SPANNING OVER SEVERAL SECTIONS OF THE IPC; TECHNICAL SUBJECTS COVERED BY FORMER USPC CROSS-REFERENCE ART COLLECTIONS [XRACs] AND DIGESTS
- Y02—TECHNOLOGIES OR APPLICATIONS FOR MITIGATION OR ADAPTATION AGAINST CLIMATE CHANGE
- Y02T—CLIMATE CHANGE MITIGATION TECHNOLOGIES RELATED TO TRANSPORTATION
- Y02T10/00—Road transport of goods or passengers
- Y02T10/10—Internal combustion engine [ICE] based vehicles
- Y02T10/40—Engine management systems
Landscapes
- Engineering & Computer Science (AREA)
- Theoretical Computer Science (AREA)
- Physics & Mathematics (AREA)
- Data Mining & Analysis (AREA)
- Evolutionary Computation (AREA)
- Life Sciences & Earth Sciences (AREA)
- Artificial Intelligence (AREA)
- General Physics & Mathematics (AREA)
- General Engineering & Computer Science (AREA)
- Computing Systems (AREA)
- Software Systems (AREA)
- Molecular Biology (AREA)
- Computational Linguistics (AREA)
- Biophysics (AREA)
- Biomedical Technology (AREA)
- Mathematical Physics (AREA)
- General Health & Medical Sciences (AREA)
- Health & Medical Sciences (AREA)
- Bioinformatics & Cheminformatics (AREA)
- Bioinformatics & Computational Biology (AREA)
- Computer Vision & Pattern Recognition (AREA)
- Evolutionary Biology (AREA)
- Probability & Statistics with Applications (AREA)
- Image Analysis (AREA)
Abstract
The invention discloses a vehicle weight recognition method, a vehicle weight recognition device, a computer and a readable storage medium, and relates to the field of transportation, wherein the method comprises the following steps: acquiring a plurality of vehicle images of the same vehicle to serve as a training image set; constructing a vehicle identification model of a double-flow main network and a knowledge distillation branch network with partial parameter sharing; inputting the training image set into the vehicle recognition model for training to obtain a trained vehicle recognition model; inputting the vehicle image to be recognized into the trained vehicle recognition model to obtain the vehicle image feature to be recognized; respectively carrying out similarity calculation on the vehicle image characteristics to be recognized and the image characteristics of a plurality of vehicles in a preset vehicle image library; and outputting the vehicle image pointed by the vehicle image feature with the highest similarity as a vehicle weight recognition result. The vehicle weight recognition method and the vehicle weight recognition system can improve the accuracy of classification in the vehicle weight recognition process and the efficiency of vehicle weight recognition.
Description
Technical Field
The invention relates to the field of transportation, in particular to a vehicle weight identification method and device, a computer and a readable storage medium.
Background
The vehicle weight identification technology refers to an identification technology for searching the same vehicle by giving a picture of a vehicle to be searched and matching the vehicle in images acquired by different cameras. The difficult aspects of the vehicle weight recognition technology are mainly as follows: factors such as appearance and shooting angle may cause a large difference in shot images obtained by shooting the same vehicle, and the acquired images of different vehicles have a higher similarity, so that the matching of the images of the same vehicle is difficult, and the recognition accuracy and efficiency are relatively low.
Disclosure of Invention
The invention provides a vehicle re-identification method, a vehicle re-identification device, a computer and a readable storage medium, aiming at the problems of relatively low identification accuracy and efficiency of the existing vehicle re-identification.
The technical scheme provided by the invention for the technical problem is as follows:
in a first aspect, the present invention provides a vehicle weight recognition method, including:
acquiring a plurality of vehicle images of the same vehicle as a training image set;
constructing a vehicle identification model of a double-flow main network and a knowledge distillation branch network with partial parameter sharing; the knowledge distillation branch comprises a knowledge distillation teacher model and a knowledge distillation student model;
inputting the training image set into the vehicle recognition model for training to obtain a trained vehicle recognition model, wherein in the training process, each knowledge distillation branch positioned on the shallow layer is used as a student to be guided and trained by the deepest knowledge distillation branch used as a teacher;
inputting the vehicle image to be recognized into the trained vehicle recognition model to obtain the vehicle image feature to be recognized;
respectively carrying out similarity calculation on the vehicle image features to be identified and the image features of a plurality of vehicles in a preset vehicle image library;
and outputting the vehicle image pointed by the vehicle image feature with the highest similarity as a vehicle weight recognition result.
Preferably, the loss of said intellectual distillation branch portion satisfies the following relation:
Ldist＝αLsoft+βLhard+γLfea，
wherein alpha, beta and gamma are equilibrium parameters of knowledge distillation, LsoftRepresenting a loss of KL divergence between the deep classifier and each shallow classifier; l ishardRepresents the cross-entropy loss of the real label; l isfeaIndicating the L2 loss between features from the deep classifier and each shallow classifier pooled.
Preferably, said LsoftSatisfies the following relation:
wherein KL (p)0||pj) Presentation computation deepest teacher classifier p0And each shallow student classifier pjKL divergence of (1); n represents the number of pictures in the mini-batch during training; m represents the total number of shallow student classifiers; p is a radical of formula0(i) And pj(i) Represents the deepest layer of the ith picture in the mini-batch.
Preferably, said LhardSatisfies the following relation:
Lhard＝∑M∑Nqj(i)log(pj(i))，
wherein p isj(i) Representing all identity prediction logic distribution, q, of the ith picture in the mini-batch in a training set corresponding to the operation of each shallow student network and softmaxj(i) Representing the distribution of the corresponding real labels of the image.
Preferably, said LfeaSatisfies the following relation:
wherein f is0(i) Representing the pooled feature of the network diagram of the teacher with the deepest image in the mini-batch, fj(i) Showing the pooled features of the feature maps of the shallow student networks,
indicating a loss of L2.
Preferably, the acquiring a plurality of vehicle images of the same vehicle as the training image set includes:
and acquiring a plurality of vehicle images of the same vehicle on line by adopting a random batch sampling strategy to serve as the training image set.
Preferably, the dual-stream backbone network adopts a resnet50 network, and comprises a feature extraction part and a feature embedding part.
In a second aspect, the present invention provides a vehicle weight recognition apparatus, the apparatus comprising:
the acquisition module is used for acquiring a plurality of vehicle images of the same vehicle to serve as a training image set;
the model construction module is used for constructing a vehicle identification model with a double-flow main network and a knowledge distillation branch network which share part of parameters; the knowledge distillation branch comprises a knowledge distillation teacher model and a knowledge distillation student model;
the model training module is used for inputting the training image set into the vehicle recognition model for training to obtain a trained vehicle recognition model, wherein in the training process, each knowledge distillation branch positioned on the shallow layer is used as a student to be guided and trained by the knowledge distillation branch positioned on the deepest layer as a teacher;
the input module is used for inputting the vehicle image to be recognized into the trained vehicle recognition model to obtain the vehicle image feature to be recognized;
the calculation module is used for respectively calculating the similarity of the image features of the vehicle to be identified and the image features of a plurality of vehicles in a preset vehicle image library;
and the output module is used for outputting the vehicle image pointed by the vehicle image feature with the highest similarity as a vehicle weight recognition result.
In a third aspect, the present invention also provides a computer comprising a processor for implementing the steps in the vehicle weight recognition method as described above when executing a computer program stored in a memory.
In a fourth aspect, the present invention also provides a readable storage medium, on which a computer program is stored, which, when being executed by a processor, carries out the steps of the vehicle weight recognition method as described above.
The technical scheme provided by the embodiment of the invention has the following beneficial effects:
according to the vehicle weight recognition method provided by the invention, the network model is used as a teacher of the vehicle and is continuously pushed forward by constructing the double-flow main network and the knowledge distillation branch network for training, so that the network model can extract the characteristic representation with higher identification capability, the obtained vehicle recognition model has higher classification performance, and the classification accuracy and the vehicle weight recognition efficiency in the vehicle weight recognition process are improved.
Drawings
In order to more clearly illustrate the technical solutions in the embodiments of the present invention, the drawings needed to be used in the description of the embodiments will be briefly introduced below, and it is obvious that the drawings in the following description are only some embodiments of the present invention, and it is obvious for those skilled in the art to obtain other drawings based on these drawings without creative efforts.
FIG. 1 is a flow chart of a vehicle weight recognition method according to an embodiment of the present invention;
FIG. 2 is a diagram of a knowledge distillation network model of a vehicle identification model provided by the present invention;
fig. 3 is a functional block diagram of a vehicle weight recognition apparatus according to an embodiment of the present invention.
Detailed Description
In order to make the objects, technical solutions and advantages of the present invention more apparent, embodiments of the present invention will be described in detail with reference to the accompanying drawings.
Referring to fig. 1, a flowchart of a vehicle weight recognition method according to an embodiment of the present invention is shown. The vehicle weight recognition method is mainly applied to a traffic recognition system, adopts a knowledge distillation mode, utilizes a network model as a teacher of the network model, and continuously pushes forward, so that the network model can extract characteristic representation with higher identification capability, and the classification performance of the same vehicle image is improved.
As shown in fig. 1, the vehicle weight recognition method includes the steps of:
s101: multiple vehicle images of the same vehicle are acquired as a training image set.
In this step, a random batch sampling strategy may be adopted to obtain a plurality of vehicle images of the same vehicle on line as the training image set, that is, a training set may be pre-established, where the training set includes images of different vehicles at different shooting angles and appearances, and each batch of images corresponding to different vehicles is adopted as a training image set corresponding to the same vehicle in a random manner.
S102: constructing a vehicle identification model of a double-flow main network and a knowledge distillation branch network with partial parameter sharing; the knowledge distillation branch comprises a knowledge distillation teacher model and a knowledge distillation student model.
In this step, the dual-flow backbone network may adopt a resnet50 network, and may specifically include 5 block portions, namely, a block1 serving as a feature extraction portion and blocks 2 to 5 serving as feature embedding portions.
And the Knowledge Distillation (Knowledge Distillation) branch is used for realizing the compression of the network model by finishing the training of a teacher-student network.
In this embodiment, the knowledge distillation teacher model and the knowledge distillation student model belong to different regions in a knowledge distillation model, and the deepest classifier is used to guide the learning of the shallow classifier in multiple dimensions, thereby realizing the knowledge self-distillation.
S103: and inputting the training image set into the vehicle recognition model for training to obtain a trained vehicle recognition model, wherein in the training process, each knowledge distillation branch positioned on the shallow layer is used as a student to be guided and trained by the deepest knowledge distillation branch used as a teacher.
In this step, the loss of the knowledge distillation branch part can satisfy the following relational expression:
Ldist＝αLsoft+βLhard+γLfea，
wherein alpha, beta and gamma are equilibrium parameters of knowledge distillation, LsoftRepresenting the loss of KL divergence (relative entropy) between the deep classifier and each shallow classifier; l ishardRepresents the cross-entropy loss of the real label; l isfeaIndicating the L2 loss between the deep classifier and the features pooled from each shallow classifier.
The KL divergence is used for measuring softmax output of a teacher network and a shallow layer student network, and by introducing the KL divergence, knowledge learned by a deep layer network can be used for guiding the shallow layer network, so that the two networks are distributed closely, and the training of the teacher-student network is correspondingly realized.
Said LsoftSatisfies the following relation:
wherein KL (p)0||pj) Presentation computation deepest teacher classifier p0And each shallow student classifier pjKL divergence of (1); n represents the number of pictures in the mini-batch during training; m represents the total number of shallow student classifiers; p is a radical of0(i) And pj(i) Represents the deepest layer of the ith picture in the mini-batch.
Here, the value of M in the vehicle weight recognition method provided by the present invention is equal to 3; and j is (1,2,3,) which respectively represents each shallow student classifier from shallow to deep according to the sequence from small to large, wherein j is 0 and represents the teacher classifier at the deepest layer.
In this embodiment, L ishardThe difference between the true labels used to measure the training dataset and the soft output of each shallow classifier satisfies the following relationship:
Lhard＝∑M∑Nqj(i)log(pj(i))，
wherein p isj(i) Representing all identity prediction logic distribution, q, of the ith picture in the mini-batch in a training set corresponding to the operation of each shallow student network and softmaxj(i) Representing the distribution of the corresponding real labels of the image.
In this embodiment, L isfeaFor calculating the L2 loss between the extracted features of the deepest neural network and the extracted features of each shallow network, the guidance of low-level features with high-level features is realized by introducing the hidden knowledge in the characteristic diagram of the deepest neural network into the shallow network. Specifically, the LfeaSatisfies the following relation:
wherein f is0(i) Representing the pooled feature of the network diagram of the teacher with the deepest image in the mini-batch, fj(i) Showing the pooled features of the feature maps of the shallow student networks,
indicating a loss of L2.
S104: and inputting the vehicle image to be recognized into the trained vehicle recognition model to obtain the vehicle image feature to be recognized.
S105: and respectively carrying out similarity calculation on the vehicle image features to be recognized and the image features of a plurality of vehicles in a preset vehicle image library. Here, the preset vehicle image database is an image database created by collecting images of various types of vehicles in advance.
In this step, when similarity calculation is performed, a feature spatial distance algorithm may be used and sorting may be performed according to the spatial distance, so as to obtain each feature to be sorted and each corresponding vehicle image.
S106: and outputting the vehicle image pointed by the vehicle image feature with the highest similarity as a vehicle weight recognition result. Here, the vehicle image with the highest rank order is taken as the vehicle image with the highest similarity.
According to the vehicle weight recognition method provided by the invention, the network model is used as a teacher of the vehicle and is continuously pushed forward by constructing the double-flow main network and the knowledge distillation branch network for training, so that the network model can extract the characteristic representation with higher identification capability, the obtained vehicle recognition model has higher classification performance, and the classification accuracy and the vehicle weight recognition efficiency in the vehicle weight recognition process are improved.
Referring to fig. 2, a knowledge distillation network model diagram of a vehicle identification model provided by the present invention is shown. Inputting the images into a Block1 of a main network for feature extraction, then embedding features through a Block 2-Block 5, and respectively inputting the images into a deep network for training a knowledge distillation teacher model and a shallow network for training a knowledge distillation student model, specifically:
in the deep network, the features output by Block5 are input into a BN (batch normalization) layer for batch standardization, the processed features are further input into an FC (full Connected) layer for processing to obtain a classification result, and then softmax processing is further performed to complete classification and output the classification result. Warp beamThe output classification result can be used as a soft label (soft label) to determine KL divergence loss between the deep classifier and each shallow classifier, and can be counted as Lsoft. On the other hand, the difference between the real label of the training data set and the classification result output by softmax of each shallow classifier can also be determined as the Lid (Local Intrinsic dimension) value, which can be counted as Lhard。
In shallow networks, features input into each shallow network from Block2 to Block5 are pooled, and the L2 loss between the previously input features into the deep network and the pooled features can be counted as Lfea. The corresponding shallow layer characteristics obtained by the pooling operation can be further input into the BN layer for batch standardization processing, and in the same way, the characteristics obtained by processing are further input into the FC layer for processing to obtain a classification result.
Referring to fig. 3, a functional block diagram of a vehicle weight recognition apparatus according to an embodiment of the present invention is shown. The vehicle weight recognition device 100 comprises an acquisition module 11, a model construction module 12, a model training module 13, an input module 14, a calculation module 15 and an output module 16, wherein the construction including a double-flow backbone network and a knowledge distillation branch network is realized through the cooperation among all functional modules for training, so that the network model serves as a teacher of the network model and is continuously pushed forward, the network model can extract characteristic representation with identification capability, the obtained vehicle recognition model has strong classification performance, and the classification accuracy and the vehicle weight recognition efficiency in the vehicle weight recognition process are improved.
As shown in fig. 3, in the vehicle weight recognition apparatus 100 provided by the present invention, the specific functions of the respective functional modules are as follows:
the acquiring module 11 is configured to acquire a plurality of vehicle images of the same vehicle as a training image set.
The model construction module 12 is used for constructing a vehicle identification model with a double-flow trunk network and a knowledge distillation branch network which share part of parameters; the knowledge distillation branch comprises a knowledge distillation teacher model and a knowledge distillation student model.
The model training module 13 is configured to input the training image set into the vehicle recognition model for training to obtain a trained vehicle recognition model, wherein in the training process, each knowledge distillation branch located in a shallow layer is used as a student to be guided and trained by a knowledge distillation branch located in a deepest layer, which is used as a teacher.
The input module 14 is configured to input a vehicle image to be recognized into the trained vehicle recognition model to obtain a vehicle image feature to be recognized;
the calculating module 15 is configured to perform similarity calculation on the image features of the vehicle to be identified and the image features of multiple vehicles in a preset vehicle image library respectively.
The output module 16 is configured to output a vehicle image pointed to by the vehicle image feature with the highest similarity as a vehicle weight recognition result.
It will be appreciated that in a particular application, the invention provides a computer comprising a processor for implementing the steps of the vehicle weight identification method described above when executing a computer program stored in a memory.
The Processor may be a Central Processing Unit (CPU), other general purpose Processor, a Digital Signal Processor (DSP), an Application Specific Integrated Circuit (ASIC), a Field-Programmable Gate Array (FPGA) or other Programmable logic device, discrete Gate or transistor logic device, discrete hardware component, etc. The general purpose processor may be a microprocessor or the processor may be any conventional processor or the like which is the control center for the computer device and which connects the various parts of the overall computer device using various interfaces and lines.
The memory may be used to store the computer programs and/or modules, and the processor may implement various functions of the computer device by running or executing the computer programs and/or modules stored in the memory and invoking data stored in the memory. The memory may mainly include a storage program area and a storage data area, wherein the storage program area may store an operating system, an application program required for at least one function, and the like; the storage data area may store data created according to the use of the mobile phone, and the like. In addition, the memory may include high speed random access memory, and may also include non-volatile memory, such as a hard disk, a memory, a plug-in hard disk, a Smart Media Card (SMC), a Secure Digital (SD) Card, a Flash memory Card (Flash Card), at least one magnetic disk storage device, a Flash memory device, or other volatile solid state storage device.
Furthermore, the present invention also provides a readable storage medium, on which a computer program is stored, which computer program, when being executed by a processor, realizes the steps of the aforementioned vehicle identification method.
The above description is only for the purpose of illustrating the preferred embodiments of the present invention and is not to be construed as limiting the invention, and any modifications, equivalents, improvements and the like that fall within the spirit and principle of the present invention are intended to be included therein.
Claims (10)
1. A vehicle weight recognition method, characterized in that the method comprises:
acquiring a plurality of vehicle images of the same vehicle to serve as a training image set;
constructing a vehicle identification model of a double-flow main network and a knowledge distillation branch network with partial parameter sharing; the knowledge distillation branch comprises a knowledge distillation teacher model and a knowledge distillation student model;
inputting the training image set into the vehicle recognition model for training to obtain a trained vehicle recognition model, wherein in the training process, each knowledge distillation branch positioned on the shallow layer is used as a student to be guided and trained by the deepest knowledge distillation branch used as a teacher;
inputting the vehicle image to be recognized into the trained vehicle recognition model to obtain the vehicle image feature to be recognized;
respectively carrying out similarity calculation on the vehicle image features to be identified and the image features of a plurality of vehicles in a preset vehicle image library;
and outputting the vehicle image pointed by the vehicle image feature with the highest similarity as a vehicle weight recognition result.
2. The vehicle weight recognition method according to claim 1, wherein the loss of the knowledge distillation branch portion satisfies the following relation:
Ldist＝αLsoft+βLhard+γLfea，
wherein alpha, beta and gamma are equilibrium parameters of knowledge distillation, LsoftRepresenting a loss of KL divergence between the deep classifier and each shallow classifier; l ishardRepresents the cross-entropy loss of the real label; l isfeaIndicating the L2 loss between features from the deep classifier and each shallow classifier pooled.
3. The vehicle weight recognition method according to claim 2, wherein L issoftSatisfies the following relation:
wherein KL (p)0||pj) Presentation computation deepest teacher classifier p0And each shallow student classifier pjKL divergence of (1); n represents the number of pictures in the mini-batch during training; m represents the total number of shallow student classifiers; p is a radical of0(i) And pj(i) Represents the deepest layer of the ith picture in the mini-batch.
4. The vehicle weight recognition method according to claim 2, wherein L ishardSatisfies the following relation:
wherein p isj(i) Representing all identity prediction logic distribution, q, of the ith picture in the mini-batch in a training set corresponding to the operation of each shallow student network and softmaxj(i) Representing the distribution of the corresponding real labels of the image.
5. The vehicle weight recognition method according to claim 2, wherein L isfeaSatisfies the following relation:
wherein f is0(i) Representing the pooled feature of the network diagram of the teacher with the deepest image in the mini-batch, fj(i) Showing the pooled features of the feature maps of the shallow student networks,
indicating a loss of L2.
6. The vehicle re-recognition method of claim 1, wherein the obtaining of multiple vehicle images of the same vehicle as a training image set comprises:
and acquiring a plurality of vehicle images of the same vehicle on line by adopting a random batch sampling strategy to serve as the training image set.
7. The vehicle weight recognition method according to any one of claims 1 to 6, wherein the dual-flow backbone network employs a resnet50 network, including a feature extraction part and a feature embedding part.
8. A vehicle weight recognition apparatus, characterized in that the apparatus comprises:
the acquisition module is used for acquiring a plurality of vehicle images of the same vehicle to serve as a training image set;
the model construction module is used for constructing a vehicle identification model with a double-flow main network and a knowledge distillation branch network which share part of parameters; the knowledge distillation branch comprises a knowledge distillation teacher model and a knowledge distillation student model;
the model training module is used for inputting the training image set into the vehicle recognition model for training to obtain a trained vehicle recognition model, wherein in the training process, each knowledge distillation branch positioned on a shallow layer is used as a student to be guided and trained by a knowledge distillation branch at the deepest layer used as a teacher;
the input module is used for inputting the vehicle image to be recognized into the trained vehicle recognition model to obtain the vehicle image feature to be recognized;
the calculation module is used for respectively calculating the similarity of the image features of the vehicle to be identified and the image features of a plurality of vehicles in a preset vehicle image library;
and the output module is used for outputting the vehicle image pointed by the vehicle image feature with the highest similarity as a vehicle weight recognition result.
9. A computer, characterized in that the computer comprises a processor for implementing the steps in the vehicle weight recognition method according to any one of claims 1-7 when executing a computer program stored in a memory.
10. A readable storage medium, on which a computer program is stored, which, when being executed by a processor, carries out the steps of the method for identifying vehicle weight according to any one of claims 1 to 7.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202210033507.1A CN114596546A (en) | 2022-01-12 | 2022-01-12 | Vehicle weight recognition method and device, computer and readable storage medium |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202210033507.1A CN114596546A (en) | 2022-01-12 | 2022-01-12 | Vehicle weight recognition method and device, computer and readable storage medium |
Publications (1)
| Publication Number | Publication Date |
|---|---|
| CN114596546A true CN114596546A (en) | 2022-06-07 |
Family
ID=81803773
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN202210033507.1A Pending CN114596546A (en) | 2022-01-12 | 2022-01-12 | Vehicle weight recognition method and device, computer and readable storage medium |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN114596546A (en) |
Cited By (2)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN115393660A (en) * | 2022-10-28 | 2022-11-25 | 松立控股集团股份有限公司 | Parking lot fire detection method based on weak supervision collaborative sparse relationship ranking mechanism |
| CN115457006A (en) * | 2022-09-23 | 2022-12-09 | 华能澜沧江水电股份有限公司 | Unmanned aerial vehicle inspection defect classification method and device based on similarity consistency self-distillation |
Citations (5)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN113221935A (en) * | 2021-02-02 | 2021-08-06 | 清华大学 | Image identification method and system based on environment perception deep convolutional neural network |
| CN113283362A (en) * | 2021-06-04 | 2021-08-20 | 中国矿业大学 | Cross-modal pedestrian re-identification method |
| CN113326768A (en) * | 2021-05-28 | 2021-08-31 | 浙江商汤科技开发有限公司 | Training method, image feature extraction method, image recognition method and device |
| CN113361464A (en) * | 2021-06-30 | 2021-09-07 | 重庆交通大学 | Vehicle weight recognition method based on multi-granularity feature segmentation |
| CN113486723A (en) * | 2021-06-10 | 2021-10-08 | 华侨大学 | Efficient vehicle re-identification method and device |
-
2022
- 2022-01-12 CN CN202210033507.1A patent/CN114596546A/en active Pending
Patent Citations (5)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN113221935A (en) * | 2021-02-02 | 2021-08-06 | 清华大学 | Image identification method and system based on environment perception deep convolutional neural network |
| CN113326768A (en) * | 2021-05-28 | 2021-08-31 | 浙江商汤科技开发有限公司 | Training method, image feature extraction method, image recognition method and device |
| CN113283362A (en) * | 2021-06-04 | 2021-08-20 | 中国矿业大学 | Cross-modal pedestrian re-identification method |
| CN113486723A (en) * | 2021-06-10 | 2021-10-08 | 华侨大学 | Efficient vehicle re-identification method and device |
| CN113361464A (en) * | 2021-06-30 | 2021-09-07 | 重庆交通大学 | Vehicle weight recognition method based on multi-granularity feature segmentation |
Non-Patent Citations (1)
| Title |
|---|
| 窦鑫泽;盛浩;吕凯;刘洋;张洋;吴玉彬;柯韦;: "基于高置信局部特征的车辆重识别优化算法", 北京航空航天大学学报, no. 09, 15 September 2020 (2020-09-15) * |
Cited By (4)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN115457006A (en) * | 2022-09-23 | 2022-12-09 | 华能澜沧江水电股份有限公司 | Unmanned aerial vehicle inspection defect classification method and device based on similarity consistency self-distillation |
| CN115457006B (en) * | 2022-09-23 | 2023-08-22 | 华能澜沧江水电股份有限公司 | Unmanned aerial vehicle inspection defect classification method and device based on similarity consistency self-distillation |
| CN115393660A (en) * | 2022-10-28 | 2022-11-25 | 松立控股集团股份有限公司 | Parking lot fire detection method based on weak supervision collaborative sparse relationship ranking mechanism |
| CN115393660B (en) * | 2022-10-28 | 2023-02-24 | 松立控股集团股份有限公司 | Parking lot fire detection method based on weak supervision collaborative sparse relationship ranking mechanism |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| EP3731154A1 (en) | Training a convolutional neural network for image retrieval with a listwise ranking loss function | |
| US20230376527A1 (en) | Generating congruous metadata for multimedia | |
| US20080063264A1 (en) | Method for classifying data using an analytic manifold | |
| CN112184508A (en) | Student model training method and device for image processing | |
| CN114596546A (en) | Vehicle weight recognition method and device, computer and readable storage medium | |
| CN111344697A (en) | Complexity-based progressive training for machine vision models | |
| CN111126396A (en) | Image recognition method and device, computer equipment and storage medium | |
| CN111914908B (en) | Image recognition model training method, image recognition method and related equipment | |
| CN113298197B (en) | Data clustering method, device, equipment and readable storage medium | |
| CN111027576A (en) | Cooperative significance detection method based on cooperative significance generation type countermeasure network | |
| CN115147632A (en) | Image category automatic labeling method and device based on density peak value clustering algorithm | |
| CN114358205A (en) | Model training method, model training device, terminal device, and storage medium | |
| CN117152459A (en) | Image detection method, device, computer readable medium and electronic equipment | |
| CN112287923A (en) | Card information identification method, device, equipment and storage medium | |
| CN112613474A (en) | Pedestrian re-identification method and device | |
| CN116108217B (en) | Fee evasion vehicle similar picture retrieval method based on depth hash coding and multitask prediction | |
| CN116958724A (en) | Training method and related device for product classification model | |
| CN113255828B (en) | Feature retrieval method, device, equipment and computer storage medium | |
| CN115063617A (en) | Advertisement image recognition method and advertisement image recognition system | |
| CN114462479A (en) | Model training method, model searching method, model, device and medium | |
| CN114220078A (en) | Target re-identification method and device and computer readable storage medium | |
| CN114596435A (en) | Semantic segmentation label generation method, device, equipment and storage medium | |
| CN113869398A (en) | Unbalanced text classification method, device, equipment and storage medium | |
| CN116052220B (en) | Pedestrian re-identification method, device, equipment and medium | |
| CN113505642B (en) | Method, device, equipment and storage medium for improving target re-identification generalization |
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 |