CN110135325A - Crowd's number method of counting and system based on dimension self-adaption network - Google Patents
Crowd's number method of counting and system based on dimension self-adaption network Download PDFInfo
- Publication number
- CN110135325A CN110135325A CN201910390191.XA CN201910390191A CN110135325A CN 110135325 A CN110135325 A CN 110135325A CN 201910390191 A CN201910390191 A CN 201910390191A CN 110135325 A CN110135325 A CN 110135325A
- Authority
- CN
- China
- Prior art keywords
- crowd
- image
- network
- counting
- adaption
- Prior art date
- Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
- Granted
Links
- 238000000034 method Methods 0.000 title claims abstract description 29
- 238000012549 training Methods 0.000 claims abstract description 17
- 238000012360 testing method Methods 0.000 claims description 13
- 238000013527 convolutional neural network Methods 0.000 claims description 11
- 238000000605 extraction Methods 0.000 claims description 10
- 238000012545 processing Methods 0.000 claims description 8
- 238000010276 construction Methods 0.000 claims description 7
- 230000007246 mechanism Effects 0.000 claims description 6
- 238000004590 computer program Methods 0.000 claims description 5
- 230000003044 adaptive effect Effects 0.000 claims 1
- 230000006870 function Effects 0.000 description 3
- 238000012544 monitoring process Methods 0.000 description 3
- 230000001413 cellular effect Effects 0.000 description 2
- 238000013461 design Methods 0.000 description 2
- 230000009286 beneficial effect Effects 0.000 description 1
- 230000002860 competitive effect Effects 0.000 description 1
- 230000010485 coping Effects 0.000 description 1
- 238000013135 deep learning Methods 0.000 description 1
- 238000000151 deposition Methods 0.000 description 1
- 238000001514 detection method Methods 0.000 description 1
- 238000010586 diagram Methods 0.000 description 1
- 230000000694 effects Effects 0.000 description 1
- 238000011156 evaluation Methods 0.000 description 1
- 239000000945 filler Substances 0.000 description 1
- 238000003780 insertion Methods 0.000 description 1
- 230000037431 insertion Effects 0.000 description 1
- 230000001788 irregular Effects 0.000 description 1
- 238000010801 machine learning Methods 0.000 description 1
- 230000035800 maturation Effects 0.000 description 1
- 239000000203 mixture Substances 0.000 description 1
- 238000012986 modification Methods 0.000 description 1
- 230000004048 modification Effects 0.000 description 1
- 230000001537 neural effect Effects 0.000 description 1
- 238000003909 pattern recognition Methods 0.000 description 1
- 230000008569 process Effects 0.000 description 1
- 230000009467 reduction Effects 0.000 description 1
- 230000011218 segmentation Effects 0.000 description 1
Classifications
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
- G06T3/00—Geometric image transformations in the plane of the image
- G06T3/40—Scaling of whole images or parts thereof, e.g. expanding or contracting
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
- G06T7/00—Image analysis
- G06T7/70—Determining position or orientation of objects or cameras
- G06T7/77—Determining position or orientation of objects or cameras using statistical methods
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06V—IMAGE OR VIDEO RECOGNITION OR UNDERSTANDING
- G06V20/00—Scenes; Scene-specific elements
- G06V20/50—Context or environment of the image
- G06V20/52—Surveillance or monitoring of activities, e.g. for recognising suspicious objects
- G06V20/53—Recognition of crowd images, e.g. recognition of crowd congestion
-
- 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
- G06T2207/20081—Training; Learning
-
- 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]
-
- 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
- G06T2207/30—Subject of image; Context of image processing
- G06T2207/30196—Human being; Person
-
- 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
- G06T2207/30—Subject of image; Context of image processing
- G06T2207/30242—Counting objects in image
Landscapes
- Engineering & Computer Science (AREA)
- Physics & Mathematics (AREA)
- General Physics & Mathematics (AREA)
- Theoretical Computer Science (AREA)
- Multimedia (AREA)
- Life Sciences & Earth Sciences (AREA)
- Bioinformatics & Cheminformatics (AREA)
- Bioinformatics & Computational Biology (AREA)
- Evolutionary Biology (AREA)
- Probability & Statistics with Applications (AREA)
- Computer Vision & Pattern Recognition (AREA)
- Image Analysis (AREA)
Abstract
The invention discloses a kind of crowd's number method of counting and system based on dimension self-adaption network, comprising: obtains the original image comprising crowd and scales it；Density map is generated according to the image after scaling and respectively to image and the random interception image block of density map；Dimension self-adaption network is trained using density segment corresponding to image block and each image block；The dimension self-adaption network completed using training is summed it up for each image output density figure, and by all pixels in density map, finally obtains all numbers in original image.The present invention effectively improves the accuracy of crowd's counting and the robustness to number of people size difference and complex background.
Description
Technical field
The invention belongs to field of image processing more particularly to a kind of crowd's number counting sides based on dimension self-adaption network
Method and system.
Background technique
Only there is provided background technical informations related to the present invention for the statement of this part, it is not necessary to so constitute first skill
Art.
Crowd counts (Crowd Counting) and refers to for the crowd in video or image, counts individual goal number.
In recent years, it is counted and is widely studied and applied in field of intelligent monitoring, such as based on the crowd of pattern-recognition and machine learning: machine
The flow of the people monitoring on the ground such as field station and megastore's regionality Crowds Distribute etc..Number by monitoring certain place can be management
Mechanism provides real-time density information, effectively control flow of the people, so that it is supplied to the accurate number of manager and its distributed intelligence, it can
Prevent potential collision hazard caused by due to crowd density is excessive.However, crowd is irregular to be distributed and answers due to number of people size difference
The problems such as miscellaneous background, crowd count and are still faced with very big challenge.
Inventors have found that existing people counting method focuses on the single number of output largely to indicate number, it can not
Show the detailed information such as Crowds Distribute, therefore practical application has little significance.Since 2015, output density figure has been gradually appeared
And the people counting method of number is obtained based on density map, but the ability for coping with multiscale target and complex background is weaker, it calculates
It is time-consuming also relatively long.
Summary of the invention
To solve the above-mentioned problems, the present invention propose a kind of crowd's number method of counting based on dimension self-adaption network and
System devises for number of people size difference and complex background in image and is intensively connected by basic network, scaling up unit, unit
The dimension self-adaption network for connecing mode and channel attention unit composition, exports its corresponding density map and people to each image
Number.
In some embodiments, it adopts the following technical scheme that
Crowd's number method of counting based on dimension self-adaption network, comprising:
The original image comprising crowd is obtained, processing is zoomed in and out to original image, and raw according to the number label of sample
At corresponding density map；The number label of the sample refers to the position of the number of people center marked in original image in the picture
It sets；
The image block of interception setting quantity from the image after scaling, the density image of interception setting quantity from density map
Block；
Based on expansion convolutional neural networks and channel attention mechanism construction dimension self-adaption crowd's counting and network；
Utilize described image block and density image block training dimension self-adaption crowd counting and network；
The dimension self-adaption crowd's counting and network completed using training, calculates the density map of every width testing image, and will be close
All pixels in degree figure are added up to obtain the number in testing image.
In other embodiments, it adopts the following technical scheme that
Crowd's number number system based on dimension self-adaption network, comprising:
For obtaining the original image comprising crowd, the module of processing is zoomed in and out to original image；
The module of corresponding density map is generated for the number label according to sample；The number label of the sample refers to
The position of the number of people center marked in original image in the picture；
For the module of the image block of interception setting quantity from the image after scaling, for intercepting setting from density map
The module of the density image block of quantity；
For based on expansion convolutional neural networks and channel attention mechanism construction dimension self-adaption crowd's counting and network
Module；
For the module using described image block and density image block training dimension self-adaption crowd counting and network；
Dimension self-adaption crowd's counting and network for being completed using training, calculates the density map of every width testing image, and
All pixels in density map are added up to obtain the module of the number in testing image.
In other embodiments, it adopts the following technical scheme that
A kind of terminal device, including server, the server include memory, processor and store on a memory simultaneously
The computer program that can be run on a processor, the processor realize any one of claim 1-4 institute when executing described program
The crowd's number method of counting based on dimension self-adaption network stated.
In other embodiments, it adopts the following technical scheme that
A kind of computer readable storage medium, is stored thereon with computer program, execution when which is executed by processor
The described in any item crowd's number method of counting based on dimension self-adaption network of claim 1-4.
Compared with prior art, the beneficial effects of the present invention are:
(1) the features such as present invention is complicated for background in image, number of people size difference devises dimension self-adaption network.Its
In the extractable tool of scaling up unit that is made of conventional roll integral branch and empty convolution branch there are two types of the spies of different feeling open country
Sign；Multiple scaling up units are attached using intensive connection mode, further increase receptive field range, while making to feel
It is distributed in a certain range by open country more dense, therefore can effectively deal with number of people size difference problem in image or video；Channel note
Meaning power unit enhances the feature channel with suitable receptive field to be directed to number of people size different in input picture selectively,
Negative effect caused by the competition of different characteristic interchannel has been effectively relieved.
(2) learn characteristics of image automatically by depth convolutional neural networks, so as to avoid design feature extractor with right
The complex task of image progress manual feature extraction.
(3) output is density map, it is possible to provide the detailed information such as Crowds Distribute, and can be direct according to the pixel of density map adduction
Obtain number.
Detailed description of the invention
The accompanying drawings constituting a part of this application is used to provide further understanding of the present application, and the application's shows
Meaning property embodiment and its explanation are not constituted an undue limitation on the present application for explaining the application.
Fig. 1 is the structure chart of dimension self-adaption network；
Fig. 2 is scaling up cellular construction figure；
Fig. 3 is expansion convolution sum tradition convolution comparison；
Fig. 4 is channel attention cellular construction figure.
Specific embodiment
It is noted that following detailed description is all illustrative, it is intended to provide further instruction to the application.Unless another
It indicates, all technical and scientific terms that the present invention uses have logical with the application person of an ordinary skill in the technical field
The identical meanings understood.
It should be noted that term used herein above is merely to describe specific embodiment, and be not intended to restricted root
According to the illustrative embodiments of the application.As used herein, unless the context clearly indicates otherwise, otherwise singular
Also it is intended to include plural form, additionally, it should be understood that, when in the present specification using term "comprising" and/or " packet
Include " when, indicate existing characteristics, step, operation, device, component and/or their combination.
Embodiment one
In view of deep learning in the extensive use (tracking, detection, positioning etc.) of field of machine vision and convolutional Neural net
Powerful performance of the network in terms of image procossing, feature learning.
In one or more embodiments, a kind of number method of counting based on dimension self-adaption network is disclosed, it should
Method is directed to the problems such as number of people size difference and complicated background, utilizes expansion convolutional neural networks and traditional convolutional neural networks
In conjunction with, obtain have multiple receptive fields feature output density figure go forward side by side pedestrian group count:
Crowd's counting is carried out using traditional convolutional neural networks and expansion convolutional neural networks；There are more senses to obtain
By wild feature to cope with number of people size difference and complex background, multiple units are taken into intensive linking scheme；It is multiple to reduce
The competitiveness of feature interchannel devises channel attention unit.
Number method of counting based on dimension self-adaption network, specifically includes:
Step 1: obtaining the original image comprising crowd, scaling processing is carried out to original image.Concrete operations are that will acquire
Original image length and the wide multiple respectively reset to original image size immediate 16.It is convenient in this way as training sample
Dimensionality reduction in segmentation and network training.
Corresponding density map is generated according to the number label of sample；The number label of sample refers to marking in original image
Number of people center position in the picture；
The generation method of density map D is shown in formula (1).
Wherein J is the number of crowd position in image, xiIndicate the coordinate of number of people specific location.G () and σiThen distinguish table
Show Gaussian kernel and its variance.When data are dense population, σiBy being led at a distance from two nearest targets according to target
Cross the calculating of k nearest neighbor algorithm.When data are sparse crowd, by σiDirectly select a fixed value.
Step 2: intercepting the image block of 1/4 size of length and width of 9 original images to each image at random, phase is carried out to density map
Same operation；
Step 3: training dimension self-adaption network using image block and corresponding density image block；
Step 4: the dimension self-adaption network completed using training calculates the density map of each image, and in density map
All pixels added up to obtain the number in image.Testing image is input to trained dimension self-adaption crowd
Counting and network exports corresponding density map.Pixel all in density map is summed up, the number in testing image is obtained.
Wherein, dimension self-adaption network training module can be further subdivided into 4 units, be basic network, scale respectively
The intensive connection mode of expanding unit, unit and channel attention unit.Wherein basic network is traditional convolutional layer, for extracting
Primary features.Scaling up unit integrates branch by conventional roll and expansion convolution branch forms, and can extract has different feeling wild big
Small characteristic pattern.The intensive connection mode of unit is next for the density map that each scaling up unit exports is further input into its
A and its each scaling up unit later.The purpose of channel attention unit is that have suitable sense for different image selections
By the density map channel of wild feature, to reduce the competitiveness of multiple feature interchannels.
Fig. 1 is dimension self-adaption network structure.The network carries out shallow-layer feature extraction by three convolutional layers first, then
The scaling up unit (Scale expansion unit, SEU) that the feature of extraction continues to be transmitted to 3 stackings is subjected to more rulers
Spend feature extraction.Each scaling up unit is made of simple convolution branch, while expanding feature receptive field quantity
Effectively control parameter amount.In addition, using intensive connection mode (Dense connectivity between scaling up unit
Pattern, DCP), so as to obtain the feature with distribution comparatively dense and the larger receptive field of range.But there is different size
There is competition in the feature interchannel of receptive field, in order to choose the feature with corresponding size receptive field for different inputs,
Attention unit (Residual channel-wise attention in channel is inserted after each scaling up unit
Unit, RCAU) different weights is distributed to feature channel to reinforce having the feature of suitable receptive field for different inputs
Channel.
Scaling up unit is made of Liang Ge branch, is conventional roll integral branch and empty convolution branch respectively.It is rolled up in cavity
In integral branch, two layers of empty convolution operation is used.By being inserted into cavity in convolution kernel, it is made to obtain biggish receptive field.
As shown in Figure 2, original image range corresponding to each pixel is propped up in conventional roll integral in the extracted characteristic pattern of scaling up unit
In be 3 × 3 pixels (white area), and corresponding original image range size is then 9 × 9 pixels in empty convolution branch
(bright gray parts).Therefore, conventional roll integral branch and empty convolution branch are respectively provided with 3 × 3 and 9 × 9 receptive field size.Two
A extracted feature of branch is merged, and is then delivered to next scaling up unit or feature extraction layer further progress is special
Sign is extracted, and in this manner, receptive field quantity is increased one times by single scaling up unit.In addition, it should be noted that
Traditional convolution sum cavity convolution operation in scaling up unit increases filler to guarantee output and input feature vector figure
Size is consistent, therefore the extracted feature of Liang Ge branch can be stacked directly, without carrying out the adjustment of characteristic pattern size.
Fig. 3 is traditional convolution sum cavity convolution operation contrast schematic diagram that convolution kernel is all 3 × 3.Bright gray parts in figure
For the original image to convolution.Wherein dark gray pixel is convolution kernel, is 3 × 3, and empty convolution passes through insertion blank, convolution
Template size becomes 5 × 5.It is also different to eventually pass through the characteristic pattern size that convolution operation obtains.
When carrying out crowd's counting, the receptive field of extracted feature is most important to the density map that can export high accuracy.
If receptive field is too small, it is only capable of a part of coverage goal, easily leads to leakage meter；And receptive field too greatly then easily extract it is incoherent
Background information, this is unfavorable for being counted.By intensively being connected to multiple scaling up units, counting and network be can extract
The feature of receptive field with dense distribution in a big way, but this task is counted for crowd, and not all, of the features is equal
It is necessary to not conform to wherein having for input picture because there are competitive relations for the feature interchannel with different feeling open country
The feature of suitable receptive field should be inhibited as far as possible to obtain more accurate density map.Pay attention to for this purpose, the present embodiment devises channel
Power unit adaptively reinforces or the weight of attenuation of correlation channel characteristics, structure are shown in Fig. 4；Channel attention unit is for input
Feature, learn the importance in its different channel first, weight is distributed in as different channels.Its process are as follows: first using complete
Office's pond layer, calculates its average value to each characteristic pattern and obtains 1 × 1 output, therefore the characteristic pattern in N number of channel will be ultimately formed
The vector of 1 × N；Then it is further processed by two convolutional layers；The power in each channel is finally obtained by Sigmoid function
Weight.Following each characteristic pattern is added multiplied by the weight assigned by it and with former characteristic pattern respectively.
It is tested on ShanghaiTech-B data set.
4.1ShanghaiTech-B data set:
The data set contains 400 width training images and 316 width test images altogether, number target minimum 9 in image, up to
578, averagely there are 123 people in each image.Its image size is 768 × 1024 pixels.
Using consensus forecast absolute error MAE and mean square prediction error MSE two indices as evaluation criterion, MAE
Value it is lower, then this method accuracy is higher, and the value of MSE is lower, then the robustness of this method is better.
Experimental result more only comprising basic network (BN) and is separately added into scaling up unit (SEU), intensively connects mould
Formula (DCP) and the comparison of the network structure of channel attention unit (RCAU) are as follows:
Embodiment two
In one or more embodiments, a kind of number number system based on dimension self-adaption network is disclosed, is wrapped
It includes:
For obtaining the original image comprising crowd, the module of processing is zoomed in and out to original image；
The module of corresponding density map is generated for the number label according to sample；The number label of the sample refers to
The position of the number of people center marked in original image in the picture；
For the module of the image block of interception setting quantity from the image after scaling, for intercepting setting from density map
The module of the density image block of quantity；
For based on expansion convolutional neural networks and channel attention mechanism construction dimension self-adaption crowd's counting and network
Module；
For the module using described image block and density image block training dimension self-adaption crowd counting and network；
Dimension self-adaption crowd's counting and network for being completed using training, calculates the density map of every width testing image, and
All pixels in density map are added up to obtain the module of the number in testing image.
Embodiment three
In one or more embodiments, a kind of terminal device, including server are disclosed, the server includes depositing
Reservoir, processor and storage on a memory and the computer program that can run on a processor, described in the processor execution
Crowd's number method of counting based on expansion convolutional neural networks in embodiment one is realized when program.For sake of simplicity, herein not
It repeats again.
It should be understood that processor can be central processing unit CPU, and processor can also be that other are general in the present embodiment
Processor, digital signal processor DSP, application-specific integrated circuit ASIC, ready-made programmable gate array FPGA or other are programmable
Logical device, discrete gate or transistor logic, discrete hardware components etc..General processor can be microprocessor or
The processor is also possible to any conventional processor etc..
Memory may include read-only memory and random access memory, and provides instruction and data to processor, deposits
The a part of of reservoir can also include non-volatile RAM.For example, memory can be with the information of storage device type.
During realization, each step of the above method can by the integrated logic circuit of the hardware in processor or
The instruction of software form is completed.
The method and step in one can be embodied directly in hardware processor and execute completion in conjunction with the embodiments, or use processor
In hardware and software module combination execute completion.Software module can be located at random access memory, flash memory, read-only memory, can
In the storage medium of this fields such as program read-only memory or electrically erasable programmable memory, register maturation.The storage
The step of medium is located at memory, and processor reads the information in memory, completes the above method in conjunction with its hardware.To avoid weight
It is multiple, it is not detailed herein.
Those of ordinary skill in the art may be aware that each exemplary unit, that is, algorithm steps described in conjunction with the present embodiment
Suddenly, it can be realized with the combination of electronic hardware or computer software and electronic hardware.These functions actually with hardware still
Software mode executes, specific application and design constraint depending on technical solution.Professional technician can be to each
Specific application is to use different methods to achieve the described function, but this realization is it is not considered that exceed the model of the application
It encloses.
Above-mentioned, although the foregoing specific embodiments of the present invention is described with reference to the accompanying drawings, not protects model to the present invention
The limitation enclosed, those skilled in the art should understand that, based on the technical solutions of the present invention, those skilled in the art are not
Need to make the creative labor the various modifications or changes that can be made still within protection scope of the present invention.
Claims (7)
1. crowd's number method of counting based on dimension self-adaption network characterized by comprising
The original image comprising crowd is obtained, processing is zoomed in and out to original image, and according to the generation pair of the number label of sample
The density map answered；The number label of the sample refers to the position of the number of people center marked in original image in the picture；
The image block of interception setting quantity from the image after scaling, the density image block of interception setting quantity from density map；
Based on expansion convolutional neural networks and channel attention mechanism construction dimension self-adaption crowd's counting and network；
Utilize described image block and density image block training dimension self-adaption crowd counting and network；
The dimension self-adaption crowd's counting and network completed using training, calculates the density map of every width testing image, and by density map
In all pixels added up to obtain the number in testing image.
2. crowd's number method of counting based on dimension self-adaption network as described in claim 1, which is characterized in that described
Corresponding density map is generated according to the number label of sample, specifically:
Wherein, J is the number of crowd position in image, xiIndicate the coordinate of number of people specific location；G () and σiThen respectively indicate
Gaussian kernel and its variance.
3. crowd's number method of counting based on dimension self-adaption network as described in claim 1, which is characterized in that the ruler
It spends adaptive crowd's counting and network and shallow-layer feature extraction is carried out by convolutional layer first, then continue the feature of extraction to be transmitted to heap
Folded scaling up unit carries out Multi resolution feature extraction, uses intensive connection mode between the scaling up unit；Every
Channel attention unit is inserted after a scaling up unit distributes different weights for different inputs to feature channel
To reinforce the feature channel with suitable receptive field.
4. crowd's number method of counting based on dimension self-adaption network as claimed in claim 3, which is characterized in that the ruler
Degree expanding unit includes conventional roll integral branch and empty convolution branch；In empty convolution branch, two layers of empty convolution is used
Operation, the extracted feature of Liang Ge branch is merged, be then delivered to next scaling up unit or feature extraction layer into
Row feature extraction.
5. crowd's number number system based on dimension self-adaption network characterized by comprising
For obtaining the original image comprising crowd, the module of processing is zoomed in and out to original image；
The module of corresponding density map is generated for the number label according to sample；The number label of the sample refers to original
The position of the number of people center marked in image in the picture；
For the module of the image block of interception setting quantity from the image after scaling, for the interception setting quantity from density map
Density image block module；
For the module based on expansion convolutional neural networks and channel attention mechanism construction dimension self-adaption crowd's counting and network；
For the module using described image block and density image block training dimension self-adaption crowd counting and network；
Dimension self-adaption crowd's counting and network for being completed using training, calculates the density map of every width testing image, and will be close
All pixels in degree figure are added up to obtain the module of the number in testing image.
6. a kind of terminal device, which is characterized in that including server, the server includes memory, processor and is stored in
On memory and the computer program that can run on a processor, the processor realize claim 1- when executing described program
4 described in any item crowd's number method of counting based on dimension self-adaption network.
7. a kind of computer readable storage medium, is stored thereon with computer program, which is characterized in that the program is held by processor
Perform claim requires the described in any item crowd's number method of counting based on dimension self-adaption network of 1-4 when row.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201910390191.XA CN110135325B (en) | 2019-05-10 | 2019-05-10 | Method and system for counting people of crowd based on scale adaptive network |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201910390191.XA CN110135325B (en) | 2019-05-10 | 2019-05-10 | Method and system for counting people of crowd based on scale adaptive network |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN110135325A true CN110135325A (en) | 2019-08-16 |
| CN110135325B CN110135325B (en) | 2020-12-08 |
Family
ID=67573402
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN201910390191.XA Active CN110135325B (en) | 2019-05-10 | 2019-05-10 | Method and system for counting people of crowd based on scale adaptive network |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN110135325B (en) |
Cited By (16)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN111583655A (en) * | 2020-05-29 | 2020-08-25 | 苏州大学 | Traffic flow detection method, device, equipment and medium |
| CN111626184A (en) * | 2020-05-25 | 2020-09-04 | 齐鲁工业大学 | Crowd density estimation method and system |
| CN111652152A (en) * | 2020-06-04 | 2020-09-11 | 上海眼控科技股份有限公司 | Crowd density detection method and device, computer equipment and storage medium |
| CN111723693A (en) * | 2020-06-03 | 2020-09-29 | 云南大学 | Crowd counting method based on small sample learning |
| CN111738136A (en) * | 2020-06-19 | 2020-10-02 | 新希望六和股份有限公司 | Method and device for determining number of microscopic objects, computer equipment and storage medium |
| CN111738922A (en) * | 2020-06-19 | 2020-10-02 | 新希望六和股份有限公司 | Method and device for training density network model, computer equipment and storage medium |
| CN111753671A (en) * | 2020-06-02 | 2020-10-09 | 华东师范大学 | Crowd counting method for real scene |
| CN111950458A (en) * | 2020-08-12 | 2020-11-17 | 每步科技(上海)有限公司 | Natatorium monitoring system and method and intelligent robot |
| CN112084959A (en) * | 2020-09-11 | 2020-12-15 | 腾讯科技（深圳）有限公司 | Crowd image processing method and device |
| CN112541469A (en) * | 2020-12-22 | 2021-03-23 | 山东师范大学 | Crowd counting method and system based on self-adaptive classification |
| CN112541891A (en) * | 2020-12-08 | 2021-03-23 | 山东师范大学 | Crowd counting method and system based on void convolution high-resolution network |
| CN112580545A (en) * | 2020-12-24 | 2021-03-30 | 山东师范大学 | Crowd counting method and system based on multi-scale self-adaptive context network |
| CN112734736A (en) * | 2021-01-15 | 2021-04-30 | 山东农业大学 | Accurate measuring and calculating method and device for swarm vigor based on computer vision |
| CN112767316A (en) * | 2020-12-31 | 2021-05-07 | 山东师范大学 | Crowd counting method and system based on multi-scale interactive network |
| WO2021103187A1 (en) * | 2019-11-27 | 2021-06-03 | 深圳市商汤科技有限公司 | Image processing method and apparatus, processor, electronic device, and storage medium |
| CN113052827A (en) * | 2021-03-30 | 2021-06-29 | 齐鲁工业大学 | Crowd counting method and system based on multi-branch expansion convolutional neural network |
Citations (5)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN104992223A (en) * | 2015-06-12 | 2015-10-21 | 安徽大学 | Dense population estimation method based on deep learning |
| CN106407946A (en) * | 2016-09-29 | 2017-02-15 | 北京市商汤科技开发有限公司 | Cross-line counting method, deep neural network training method, devices and electronic apparatus |
| CN107330364A (en) * | 2017-05-27 | 2017-11-07 | 上海交通大学 | A kind of people counting method and system based on cGAN networks |
| CN108364023A (en) * | 2018-02-11 | 2018-08-03 | 北京达佳互联信息技术有限公司 | Image-recognizing method based on attention model and system |
| CN109584161A (en) * | 2018-11-29 | 2019-04-05 | 四川大学 | The Remote sensed image super-resolution reconstruction method of convolutional neural networks based on channel attention |
-
2019
- 2019-05-10 CN CN201910390191.XA patent/CN110135325B/en active Active
Patent Citations (5)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN104992223A (en) * | 2015-06-12 | 2015-10-21 | 安徽大学 | Dense population estimation method based on deep learning |
| CN106407946A (en) * | 2016-09-29 | 2017-02-15 | 北京市商汤科技开发有限公司 | Cross-line counting method, deep neural network training method, devices and electronic apparatus |
| CN107330364A (en) * | 2017-05-27 | 2017-11-07 | 上海交通大学 | A kind of people counting method and system based on cGAN networks |
| CN108364023A (en) * | 2018-02-11 | 2018-08-03 | 北京达佳互联信息技术有限公司 | Image-recognizing method based on attention model and system |
| CN109584161A (en) * | 2018-11-29 | 2019-04-05 | 四川大学 | The Remote sensed image super-resolution reconstruction method of convolutional neural networks based on channel attention |
Cited By (24)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| TWI752466B (en) * | 2019-11-27 | 2022-01-11 | 大陸商深圳市商湯科技有限公司 | Image processing method, processor, electronic device, and storage medium |
| WO2021103187A1 (en) * | 2019-11-27 | 2021-06-03 | 深圳市商汤科技有限公司 | Image processing method and apparatus, processor, electronic device, and storage medium |
| CN111626184A (en) * | 2020-05-25 | 2020-09-04 | 齐鲁工业大学 | Crowd density estimation method and system |
| CN111626184B (en) * | 2020-05-25 | 2022-04-15 | 齐鲁工业大学 | Crowd density estimation method and system |
| CN111583655A (en) * | 2020-05-29 | 2020-08-25 | 苏州大学 | Traffic flow detection method, device, equipment and medium |
| CN111583655B (en) * | 2020-05-29 | 2021-12-24 | 苏州大学 | Traffic flow detection method, device, equipment and medium |
| CN111753671A (en) * | 2020-06-02 | 2020-10-09 | 华东师范大学 | Crowd counting method for real scene |
| CN111723693B (en) * | 2020-06-03 | 2022-05-27 | 云南大学 | Crowd counting method based on small sample learning |
| CN111723693A (en) * | 2020-06-03 | 2020-09-29 | 云南大学 | Crowd counting method based on small sample learning |
| CN111652152A (en) * | 2020-06-04 | 2020-09-11 | 上海眼控科技股份有限公司 | Crowd density detection method and device, computer equipment and storage medium |
| CN111738922A (en) * | 2020-06-19 | 2020-10-02 | 新希望六和股份有限公司 | Method and device for training density network model, computer equipment and storage medium |
| CN111738136A (en) * | 2020-06-19 | 2020-10-02 | 新希望六和股份有限公司 | Method and device for determining number of microscopic objects, computer equipment and storage medium |
| CN111950458A (en) * | 2020-08-12 | 2020-11-17 | 每步科技(上海)有限公司 | Natatorium monitoring system and method and intelligent robot |
| CN112084959A (en) * | 2020-09-11 | 2020-12-15 | 腾讯科技（深圳）有限公司 | Crowd image processing method and device |
| CN112084959B (en) * | 2020-09-11 | 2024-04-16 | 腾讯科技（深圳）有限公司 | Crowd image processing method and device |
| CN112541891A (en) * | 2020-12-08 | 2021-03-23 | 山东师范大学 | Crowd counting method and system based on void convolution high-resolution network |
| CN112541469A (en) * | 2020-12-22 | 2021-03-23 | 山东师范大学 | Crowd counting method and system based on self-adaptive classification |
| CN112541469B (en) * | 2020-12-22 | 2023-09-08 | 山东师范大学 | Crowd counting method and system based on self-adaptive classification |
| CN112580545A (en) * | 2020-12-24 | 2021-03-30 | 山东师范大学 | Crowd counting method and system based on multi-scale self-adaptive context network |
| CN112580545B (en) * | 2020-12-24 | 2022-07-29 | 山东师范大学 | Crowd counting method and system based on multi-scale self-adaptive context network |
| CN112767316A (en) * | 2020-12-31 | 2021-05-07 | 山东师范大学 | Crowd counting method and system based on multi-scale interactive network |
| CN112734736A (en) * | 2021-01-15 | 2021-04-30 | 山东农业大学 | Accurate measuring and calculating method and device for swarm vigor based on computer vision |
| CN113052827A (en) * | 2021-03-30 | 2021-06-29 | 齐鲁工业大学 | Crowd counting method and system based on multi-branch expansion convolutional neural network |
| CN113052827B (en) * | 2021-03-30 | 2022-12-27 | 齐鲁工业大学 | Crowd counting method and system based on multi-branch expansion convolutional neural network |
Also Published As
| Publication number | Publication date |
|---|---|
| CN110135325B (en) | 2020-12-08 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN110135325A (en) | Crowd's number method of counting and system based on dimension self-adaption network | |
| CN111723860B (en) | Target detection method and device | |
| CN106228185B (en) | A kind of general image classifying and identifying system neural network based and method | |
| WO2023015743A1 (en) | Lesion detection model training method, and method for recognizing lesion in image | |
| CN111833237B (en) | Image registration method based on convolutional neural network and local homography transformation | |
| CN108830145A (en) | A kind of demographic method and storage medium based on deep neural network | |
| CN108960404B (en) | Image-based crowd counting method and device | |
| Ou et al. | A novel rank learning based no-reference image quality assessment method | |
| CN108388905A (en) | A kind of Illuminant estimation method based on convolutional neural networks and neighbourhood context | |
| CN109711401A (en) | A kind of Method for text detection in natural scene image based on Faster Rcnn | |
| CN109146925A (en) | Conspicuousness object detection method under a kind of dynamic scene | |
| CN110781976A (en) | Extension method of training image, training method and related device | |
| CN112016454A (en) | Face alignment detection method | |
| Li et al. | Blind image quality assessment based on joint log-contrast statistics | |
| CN106898011A (en) | A kind of method that convolutional neural networks convolution nuclear volume is determined based on rim detection | |
| CN110197206A (en) | The method and device of image procossing | |
| CN116958962A (en) | Method for detecting pre-fruit-thinning pomegranate fruits based on improved YOLOv8s | |
| CN114663769B (en) | Fruit identification method based on YOLO v5 | |
| CN109902641B (en) | Semantic alignment-based face key point detection method, system and device | |
| CN110111382A (en) | Irregular area area computation method, device, computer equipment and storage medium | |
| CN109784379A (en) | The update method and device in textile picture feature library | |
| CN109741380A (en) | Textile picture fast matching method and device | |
| CN113326925A (en) | Density graph regression-based flower amount statistical method, system, equipment and medium for convolutional neural network | |
| CN113838076A (en) | Method and device for labeling object contour in target image and storage medium | |
| CN113486879B (en) | Image area suggestion frame detection method, device, equipment and storage 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 |