CN112036387B - News picture shooting angle identification method based on gated convolutional neural network - Google Patents
News picture shooting angle identification method based on gated convolutional neural network Download PDFInfo
- Publication number
- CN112036387B CN112036387B CN202011226536.7A CN202011226536A CN112036387B CN 112036387 B CN112036387 B CN 112036387B CN 202011226536 A CN202011226536 A CN 202011226536A CN 112036387 B CN112036387 B CN 112036387B
- Authority
- CN
- China
- Prior art keywords
- neural network
- image
- model
- algorithm model
- shooting angle
- 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
Images
Classifications
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06V—IMAGE OR VIDEO RECOGNITION OR UNDERSTANDING
- G06V20/00—Scenes; Scene-specific elements
- G06V20/40—Scenes; Scene-specific elements in video content
- G06V20/41—Higher-level, semantic clustering, classification or understanding of video scenes, e.g. detection, labelling or Markovian modelling of sport events or news items
-
- 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
- G06F18/24—Classification techniques
- G06F18/241—Classification techniques relating to the classification model, e.g. parametric or non-parametric approaches
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- 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
- G06N3/08—Learning methods
- G06N3/084—Backpropagation, e.g. using gradient descent
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06V—IMAGE OR VIDEO RECOGNITION OR UNDERSTANDING
- G06V2201/00—Indexing scheme relating to image or video recognition or understanding
- G06V2201/07—Target detection
Abstract
The invention discloses a news picture shooting angle identification method based on a gated convolutional neural network, which comprises the following steps of: marking a salient target area and shooting angle classification information of an image to be identified; training a constructed significance target detection algorithm model and an image classification algorithm model by using the labeled image data set; constructing a dual-branch gate control convolution neural network algorithm model based on the trained target detection algorithm model and the image classification algorithm model, and acquiring news picture shooting angles and the like corresponding to the identification images based on the dual-branch gate control convolution neural network algorithm model; the method can identify whether the corresponding news picture is obtained by overlook, look-up or head-up shooting, consumes less resources and consumes less time in operation, and can complete practical engineering application.
Description
Technical Field
The invention relates to the technical field of computer vision, in particular to a method for identifying a news picture shooting angle based on a gated convolutional neural network.
Background
Video is an important news information transmission means, and compared with a news information transmission mode which uses newspaper, radio station, broadcast and the like as transmission carriers, the video often has the characteristic of more intuitive and more realistic description of news events. When the news video is shot, different shooting angles can be adopted for shooting according to different news types. After the news is manufactured, the shooting angle information of the news picture can be marked, and the repeated utilization rate of the news material can be improved. The traditional shooting angle identification method is mainly based on a three-dimensional reconstruction method, and the method has the problems of huge resource consumption and long time consumption, and cannot meet the actual application scene.
Disclosure of Invention
The invention aims to overcome the defects of the prior art and provides a news image shooting angle identification method based on a gated convolutional neural network, which can identify the shooting angle information of a news image, specifically can identify whether the corresponding news image is obtained by overlook, look-up or look-up shooting, has less resource consumption and short calculation time consumption, and can complete practical engineering application.
The purpose of the invention is realized by the following scheme:
a news picture shooting angle identification method based on a gated convolutional neural network comprises the following steps:
marking a salient target area and shooting angle classification information of an image to be identified; training a constructed significance target detection algorithm model and an image classification algorithm model by using the labeled image data set; and constructing a double-branch gate control convolution neural network algorithm model based on the trained saliency target detection algorithm model and the trained image classification algorithm model, and acquiring a news picture shooting angle corresponding to the identification image based on the double-branch gate control convolution neural network algorithm model.
Further, labeling the salient target area and the shooting angle classification information of the image to be recognized comprises the following steps:
s101, collecting news videos, decoding the news videos into images, and collecting the images with set quantity so as to perform subsequent labeling steps;
s102, according to a data set making standard of image saliency target detection, using the image data set obtained in the step S101 to mark saliency target information in each image;
s103, according to the data set creating standard of the image classification, the data set marked in the step S102 is marked with the lens angle type information.
Further, the salient object detection algorithm model is based on the construction process steps of a readable storage medium:
s201, constructing a deep convolutional neural network structure of a coder and a decoder on the basis of a convolutional residual neural network of 101 layers, wherein the coder and the decoder adopt symmetrical characteristic pyramids, and the network structure is used as a network structure of a training significance target detection model and is marked as the network structureNs；
S202, reverse transmission by using a batch gradient descent methodBroadcast error updateNsAnd (5) repeatedly iterating model parameters until the model converges.
Further, the image classification algorithm model is based on the construction process steps of a readable storage medium:
s301, constructing an image classification algorithm network structure based on the convolution residual error neural network of the 101 layers, wherein the network structure is used as a network structure for training an image classification algorithm and is marked as the network structureNc；
S302, updating the back propagation error by adopting a batch gradient descent methodNcAnd (5) repeatedly iterating the model parameters until the classification model converges.
Further, the method comprises the following steps of constructing a dual-branch gate-controlled convolutional neural network algorithm model based on a readable storage medium:
s401, mixingNsDeleting the layer after the last convolutional layer in the network, and reserving the remaining network structure, which is recorded asFns(ii) a Deleting the layers behind the Nc network classification layer, reserving the rest network structure and recording asFnc；
S402, constructing a double-branch gate control convolution neural network algorithm modelDPThe algorithm model adopts a double-branch network to identify the image, wherein the network structure of the branch I is usedFnsThe other branch being usedFncThe gate control module uses the spatial channel attention mechanism commonly used by computer vision, and is recorded asGA classification module is added behind the gate control moduleClsAnd the function of the method is to further optimize the output characteristics of the gating module and identify the shooting angle category of the image.
Further, acquiring a news picture shooting angle corresponding to the identification image based on the two-branch gate convolution neural network algorithm model, and the method comprises the following steps:
s501, updating the back propagation error by adopting a batch gradient descent methodDPModel parameters, training of this step only updatesGAnd ClsStopping training after the model is converged;
s502, continuing to train the model updated in the step S501, and updating the model in the step SDPThe updating mode adopts a batch gradient descent method.
The invention has the beneficial effects that:
(1) the invention provides a news picture shooting angle identification method of a gate-control convolutional neural network aiming at the defects of a three-dimensional reconstruction method, which is mainly realized by using the image significance detection and image classification technology of the current convolutional neural network in computer vision.
Drawings
In order to more clearly illustrate the embodiments of the present invention or the technical solutions in the prior art, the drawings used in the description of the embodiments or the prior art will be briefly described below, and it is obvious that the drawings in the following description are only some embodiments of the present invention, and for those skilled in the art, other drawings can be obtained according to these drawings without creative efforts.
FIG. 1 is a flow chart of the method steps of the present invention;
FIG. 2 is a schematic diagram of an algorithm network structure for constructing a two-branch gated convolutional neural network model according to the present invention.
Detailed Description
All of the features disclosed in the specification for all of the embodiments (including any accompanying claims, abstract and drawings), or all of the steps of a method or process so disclosed, may be combined and/or expanded, or substituted, in any way, except for mutually exclusive features and/or steps.
As shown in fig. 1 and 2, a method for identifying a news image shooting angle based on a gated convolutional neural network includes the steps of:
marking a salient target area and shooting angle classification information of an image to be identified; training a constructed significance target detection algorithm model and an image classification algorithm model by using the labeled image data set; and constructing a double-branch gate control convolution neural network algorithm model based on the trained saliency target detection algorithm model and the trained image classification algorithm model, and acquiring a news picture shooting angle corresponding to the identification image based on the double-branch gate control convolution neural network algorithm model.
Further, labeling the salient target area and the shooting angle classification information of the image to be recognized comprises the following steps:
s101, collecting news videos, decoding the news videos into images, and collecting the images with set quantity so as to perform subsequent labeling steps;
s102, according to a data set making standard of image saliency target detection, using the image data set obtained in the step S101 to mark saliency target information in each image;
s103, according to the data set creating standard of the image classification, the data set marked in the step S102 is marked with the lens angle type information.
Further, the salient object detection algorithm model is based on the construction process steps of a readable storage medium:
s201, constructing a deep convolutional neural network structure of a coder and a decoder on the basis of a convolutional residual neural network of 101 layers, wherein the coder and the decoder adopt symmetrical characteristic pyramids, and the network structure is used as a network structure of a training significance target detection model and is marked as the network structureNs；
S202, updating the back propagation error by using a batch gradient descent methodNsAnd (5) repeatedly iterating model parameters until the model converges.
Further, the image classification algorithm model is based on the construction process steps of a readable storage medium:
s301, constructing an image classification algorithm network structure based on the convolution residual error neural network of the 101 layers, wherein the network structure is used as a network structure for training an image classification algorithm and is marked as the network structureNc；
S302, updating the back propagation error by adopting a batch gradient descent methodNcAnd (5) repeatedly iterating the model parameters until the classification model converges.
Further, the method comprises the following steps of constructing a dual-branch gate-controlled convolutional neural network algorithm model based on a readable storage medium:
s401, mixingNsDeleting the layer after the last convolutional layer in the network, and reserving the remaining network structure, which is recorded asFns(ii) a Deleting the layers behind the Nc network classification layer, reserving the rest network structure and recording asFnc；
S402, constructing a gate-controlled convolutional neural network algorithm of double branchesDPThe algorithm adopts a double-branch network to identify the image, wherein the network structure of the branch I usesFnsThe other branch being usedFncThe gate control module uses the spatial channel attention mechanism commonly used by computer vision, and is recorded asGA classification module is added behind the gate control moduleClsAnd the function of the method is to further optimize the output characteristics of the gating module and identify the shooting angle category of the image.
Further, acquiring a news picture shooting angle corresponding to the identification image based on the two-branch gate convolution neural network algorithm model, and the method comprises the following steps:
s501, updating the back propagation error by adopting a batch gradient descent methodDPModel parameters, training of this step only updatesGAnd ClsStopping training after the model is converged;
s502, continuing to train the model updated in the step S501, and updating the model in the step SDPThe updating mode adopts a batch gradient descent method.
In another embodiment of the present invention, as shown in fig. 1, a method for identifying a shot angle of a news image based on a gated convolutional neural network according to an embodiment of the present invention includes the following steps:
the method comprises the following steps: labeling the classification information of the image saliency target area and the shooting angle;
in step one, through the discussion of professional news related practitioners and computer vision related practitioners, the saliency target of the image and the annotation standard of the image capturing angle are defined.
The first step further comprises the following substeps:
step 101: in this embodiment, a plurality of original video materials can be obtained by crawling news columns of television stations in each province and city such as beijing, anhui, guangdong and the like on the internet by using a crawler technology, videos are decoded into images by using a video decoding tool, some obvious interference images such as images synthesized by a computer are removed, and a data set to be marked is formed.
Step 102: and marking each image of the data set to be marked according to the data set making rule of the saliency target detection, and marking the shooting angle category information of each image according to the marking rule of image classification.
Step two: and designing a saliency target detection algorithm, and training a saliency target detection algorithm model by using the data set obtained in the step one.
In this embodiment, the implementation of the image saliency target detection algorithm model based on the encoder includes the following sub-steps:
step 201: constructing a convolutional neural network structure of a coder and a decoder based on a convolutional residual neural network with 101 layers, wherein the coder and the decoder are in a symmetrical network structure;
step 202: updating Ns model parameters by using a batch gradient descent method for back propagation errors, and repeatedly iterating until the model converges;
step three: design and train shooting angle classification model
In this embodiment, the image classification algorithm is mainly used to identify the shooting angle information of the image, wherein the following sub-steps are included, for example, a 101-layer convolutional neural network which is general for image classification is adopted, the classification number of the network is 3, wherein 0 represents an overhead view, 1 represents an overhead view, 2 represents an eye view, and a training mode adopts a common deep learning image classification strategy.
Step four: the training for constructing the dual-branch gated convolutional neural network algorithm comprises the following specific implementation steps:
step 401: deleting the layer after the last convolution layer in the network structure of the significance detection algorithm designed in the step two, reserving the rest network structure as one branch in the double branches, deleting the layer after the network classification layer of the image classification algorithm, reserving the rest network structure as the other branch in the double branches, and constructing the double branches.
Step 402: a gating module can be designed, which uses a spatial channel attention mechanism commonly used in computer vision, as shown in fig. 2, the module first combines features of two branches, then learns the weight of the input features through the designed spatial attention mechanism, and then performs point multiplication with the fused features to obtain features S beneficial to the task. And the S is followed by a classification layer, and the class information of the shooting angle is output.
Step five: training of dual-branch gated convolutional neural network algorithm
Step 501: initializing a double-branch gate-control convolution neural network algorithm model by using the model trained in the second step and the third step, updating the parameters of the algorithm model in the figure 2 by adopting a batch gradient descent method for back propagation errors, only updating the parameters of a gate-control convolution module in the training of the step, and stopping the training after the model is converged.
Step 502: the model updated in step 501 is trained, which needs to update the parameters of all modules in fig. 2, and the updating method adopts a batch gradient descent method.
The functionality of the present invention, if implemented in the form of software functional units and sold or used as a stand-alone product, may be stored in a computer readable storage medium. Based on such understanding, the technical solution of the present invention may be embodied in the form of a software product, which is stored in a storage medium and includes instructions for causing a computer device (which may be a personal computer, a server, or a network device) to execute all or part of the steps of the method according to the embodiments of the present invention. And the aforementioned storage medium includes: various media capable of storing program codes, such as a usb disk, a removable hard disk, a Read-Only Memory (ROM), a Random Access Memory (RAM), a magnetic disk, or an optical disk.
Other embodiments than the above examples may be devised by those skilled in the art based on the foregoing disclosure, or by adapting and using knowledge or techniques of the relevant art, and features of various embodiments may be interchanged or substituted and such modifications and variations that may be made by those skilled in the art without departing from the spirit and scope of the present invention are intended to be within the scope of the following claims.
Claims (5)
1. A news picture shooting angle identification method based on a gated convolutional neural network is characterized by comprising the following steps:
marking a salient target area of an image to be identified and shooting angle classification information, comprising the following steps: s101, collecting news videos, decoding the news videos into images, and collecting the images with set quantity so as to perform subsequent labeling steps;
s102, according to a data set making standard of image saliency target detection, using the image data set obtained in the step S101 to mark saliency target information in each image;
s103, according to the data set manufacturing standard of image classification, the data set marked in the step S102 is marked with the lens angle type information; training a constructed significance target detection algorithm model and an image classification algorithm model by using the labeled image data set; and constructing a double-branch gate control convolution neural network algorithm model based on the trained saliency target detection algorithm model and the trained image classification algorithm model, and acquiring a news picture shooting angle corresponding to the identification image based on the double-branch gate control convolution neural network algorithm model.
2. The method for identifying the shooting angle of the news picture based on the gated convolutional neural network as claimed in claim 1, wherein the salient object detection algorithm model is based on the construction process steps of a readable storage medium:
s201, constructing a deep convolutional neural network structure of a coder and a decoder on the basis of a convolutional residual neural network of 101 layers, wherein the coder and the decoder adopt symmetrical characteristic pyramids, and the network structure is used as a network structure of a training significance target detection model and is marked as the network structureNs；
S202, back propagation of errors by using a batch gradient descent methodUpdatingNsAnd (5) repeatedly iterating model parameters until the model converges.
3. The method for identifying the shooting angle of the news picture based on the gated convolutional neural network as claimed in any one of claims 1 or 2, wherein the image classification algorithm model is based on the steps of a readable storage medium construction process:
s301, constructing an image classification algorithm network structure based on the convolution residual error neural network of the 101 layers, wherein the network structure is used as a network structure for training an image classification algorithm and is marked as the network structureNc；
S302, updating the back propagation error by adopting a batch gradient descent methodNcAnd (5) repeatedly iterating the model parameters until the classification model converges.
4. The method for identifying the shooting angle of the news picture based on the gated convolutional neural network as claimed in claim 3, wherein the algorithm model of the gated convolutional neural network with two branches is based on the construction process steps of a readable storage medium:
s401, mixingNsDeleting the layer after the last convolutional layer in the network, and reserving the remaining network structure, which is recorded asFns(ii) a Deleting the layers behind the Nc network classification layer, reserving the rest network structure and recording asFnc；
S402, constructing a double-branch gate control convolution neural network algorithm modelDPThe algorithm model adopts a double-branch network to identify the image, wherein the network structure of the branch I is usedFnsThe other branch being usedFncThe gate control module uses the spatial channel attention mechanism commonly used by computer vision, and is recorded asGA classification module is added behind the gate control moduleClsAnd the function of the method is to further optimize the output characteristics of the gating module and identify the shooting angle category of the image.
5. The method for identifying the shooting angle of the news picture based on the gated convolutional neural network as claimed in claim 1, wherein the method for acquiring the shooting angle of the news picture corresponding to the identification image based on the two-branch gated convolutional neural network algorithm model comprises the following steps:
s501, updating the back propagation error by adopting a batch gradient descent methodDPModel parameters, training of this step only updatesGAnd ClsStopping training after the model is converged;
s502, continuing to train the model updated in the step S501, and updating the model in the step SDPThe updating mode adopts a batch gradient descent method.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202011226536.7A CN112036387B (en) | 2020-11-06 | 2020-11-06 | News picture shooting angle identification method based on gated convolutional neural network |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202011226536.7A CN112036387B (en) | 2020-11-06 | 2020-11-06 | News picture shooting angle identification method based on gated convolutional neural network |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN112036387A CN112036387A (en) | 2020-12-04 |
| CN112036387B true CN112036387B (en) | 2021-02-09 |
Family
ID=73572748
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN202011226536.7A Active CN112036387B (en) | 2020-11-06 | 2020-11-06 | News picture shooting angle identification method based on gated convolutional neural network |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN112036387B (en) |
Families Citing this family (2)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN113255628B (en) * | 2021-07-15 | 2021-10-15 | 成都索贝数码科技股份有限公司 | Scene identification recognition method for news scene |
| CN113793286B (en) * | 2021-11-18 | 2022-05-10 | 成都索贝数码科技股份有限公司 | Media image watermark removing method based on multi-order attention neural network |
Citations (4)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN103425757A (en) * | 2013-07-31 | 2013-12-04 | 复旦大学 | Cross-medial personage news searching method and system capable of fusing multi-mode information |
| CN109801621A (en) * | 2019-03-15 | 2019-05-24 | 三峡大学 | A kind of audio recognition method based on residual error gating cycle unit |
| CN110580450A (en) * | 2019-08-12 | 2019-12-17 | 西安理工大学 | traffic sign identification method based on convolutional neural network |
| CN111242110A (en) * | 2020-04-28 | 2020-06-05 | 成都索贝数码科技股份有限公司 | Training method of self-adaptive conditional random field algorithm for automatically breaking news items |
Family Cites Families (5)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| KR101763132B1 (en) * | 2008-08-19 | 2017-07-31 | 디지맥 코포레이션 | Methods and systems for content processing |
| EP2684145A4 (en) * | 2011-03-07 | 2014-09-03 | Kba2 Inc | Systems and methods for analytic data gathering from image providers at an event or geographic location |
| CN110889428A (en) * | 2019-10-21 | 2020-03-17 | 浙江大搜车软件技术有限公司 | Image recognition method and device, computer equipment and storage medium |
| CN111552300A (en) * | 2020-06-09 | 2020-08-18 | 南开大学 | Crop picking system based on instance segmentation and path planning |
| CN111860472A (en) * | 2020-09-24 | 2020-10-30 | 成都索贝数码科技股份有限公司 | Television station caption detection method, system, computer equipment and storage medium |
-
2020
- 2020-11-06 CN CN202011226536.7A patent/CN112036387B/en active Active
Patent Citations (4)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN103425757A (en) * | 2013-07-31 | 2013-12-04 | 复旦大学 | Cross-medial personage news searching method and system capable of fusing multi-mode information |
| CN109801621A (en) * | 2019-03-15 | 2019-05-24 | 三峡大学 | A kind of audio recognition method based on residual error gating cycle unit |
| CN110580450A (en) * | 2019-08-12 | 2019-12-17 | 西安理工大学 | traffic sign identification method based on convolutional neural network |
| CN111242110A (en) * | 2020-04-28 | 2020-06-05 | 成都索贝数码科技股份有限公司 | Training method of self-adaptive conditional random field algorithm for automatically breaking news items |
Also Published As
| Publication number | Publication date |
|---|---|
| CN112036387A (en) | 2020-12-04 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN112036387B (en) | News picture shooting angle identification method based on gated convolutional neural network | |
| CN108734210B (en) | Object detection method based on cross-modal multi-scale feature fusion | |
| JP6402408B2 (en) | Tag processing method and tag processing apparatus | |
| CN105825191A (en) | Face multi-attribute information-based gender recognition method and system and shooting terminal | |
| CN107680053A (en) | A kind of fuzzy core Optimized Iterative initial value method of estimation based on deep learning classification | |
| WO2022105608A1 (en) | Rapid face density prediction and face detection method and apparatus, electronic device, and storage medium | |
| CN109961095B (en) | Image labeling system and method based on unsupervised deep learning | |
| JP7457125B2 (en) | Translation methods, devices, electronic equipment and computer programs | |
| CN113962358B (en) | Information diffusion prediction method based on time sequence hypergraph attention neural network | |
| CN112488131B (en) | Image text cross-modal retrieval method based on self-supervision countermeasure | |
| CN110796251A (en) | Image compression optimization method based on convolutional neural network | |
| CN112445876A (en) | Entity alignment method and system fusing structure, attribute and relationship information | |
| CN113627479B (en) | Graph data anomaly detection method based on semi-supervised learning | |
| WO2023077819A1 (en) | Data processing system, method and apparatus, and device, storage medium, computer program and computer program product | |
| Zhang et al. | Efficient inductive vision transformer for oriented object detection in remote sensing imagery | |
| CN115239638A (en) | Industrial defect detection method, device and equipment and readable storage medium | |
| CN110166850A (en) | The method and system of multiple CNN neural network forecast panoramic video viewing location | |
| Sharjeel et al. | Real time drone detection by moving camera using COROLA and CNN algorithm | |
| CN114842180A (en) | Point cloud completion method, device, equipment and medium | |
| CN108647295B (en) | Image labeling method based on depth collaborative hash | |
| US11804043B2 (en) | Detecting objects in a video using attention models | |
| CN113010780A (en) | Model training and click rate estimation method and device | |
| Lu et al. | Label-efficient video object segmentation with motion clues | |
| CN116977712A (en) | Knowledge distillation-based road scene segmentation method, system, equipment and medium | |
| CN113297899B (en) | Video hash algorithm based on deep learning |
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 |