CN110414401A - A kind of intelligent monitor system and monitoring method based on PYNQ - Google Patents
A kind of intelligent monitor system and monitoring method based on PYNQ Download PDFInfo
- Publication number
- CN110414401A CN110414401A CN201910661356.2A CN201910661356A CN110414401A CN 110414401 A CN110414401 A CN 110414401A CN 201910661356 A CN201910661356 A CN 201910661356A CN 110414401 A CN110414401 A CN 110414401A
- Authority
- CN
- China
- Prior art keywords
- pynq
- module
- monitor system
- neural networks
- intelligent monitor
- 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
Classifications
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F18/00—Pattern recognition
- G06F18/20—Analysing
- G06F18/23—Clustering techniques
- G06F18/232—Non-hierarchical techniques
- G06F18/2321—Non-hierarchical techniques using statistics or function optimisation, e.g. modelling of probability density functions
- G06F18/23213—Non-hierarchical techniques using statistics or function optimisation, e.g. modelling of probability density functions with fixed number of clusters, e.g. K-means clustering
-
- 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
- G06F18/2411—Classification techniques relating to the classification model, e.g. parametric or non-parametric approaches based on the proximity to a decision surface, e.g. support vector machines
-
- 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
-
- 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
-
- 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
-
- H—ELECTRICITY
- H04—ELECTRIC COMMUNICATION TECHNIQUE
- H04N—PICTORIAL COMMUNICATION, e.g. TELEVISION
- H04N7/00—Television systems
- H04N7/18—Closed-circuit television [CCTV] systems, i.e. systems in which the video signal is not broadcast
-
- 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
Landscapes
- Engineering & Computer Science (AREA)
- Theoretical Computer Science (AREA)
- Physics & Mathematics (AREA)
- Data Mining & Analysis (AREA)
- General Physics & Mathematics (AREA)
- General Engineering & Computer Science (AREA)
- Life Sciences & Earth Sciences (AREA)
- Artificial Intelligence (AREA)
- Evolutionary Computation (AREA)
- Multimedia (AREA)
- Biophysics (AREA)
- Biomedical Technology (AREA)
- Computing Systems (AREA)
- General Health & Medical Sciences (AREA)
- Computational Linguistics (AREA)
- Mathematical Physics (AREA)
- Software Systems (AREA)
- Molecular Biology (AREA)
- Health & Medical Sciences (AREA)
- Evolutionary Biology (AREA)
- Bioinformatics & Cheminformatics (AREA)
- Bioinformatics & Computational Biology (AREA)
- Computer Vision & Pattern Recognition (AREA)
- Signal Processing (AREA)
- Probability & Statistics with Applications (AREA)
- Image Analysis (AREA)
Abstract
The present invention relates to a kind of intelligent monitor system and monitoring method based on PYNQ passes through software-hardware synergism and realizes the classification of mesh multi-target detection.Include mainly multi-target detection module, carries out the transplanting and optimization of algorithm, general convolutional neural networks accelerator IP, the api interface based on python.The PYNQ is integrated with arm processor system and FPGA programmable logic, and software section has transplanted caffe frame, is suitable for mainstream intelligent algorithm, improves faster-RCNN algorithm and is transplanted to PYNQ platform realization target detection function.FPGA portion accelerates IP using convolutional neural networks to carry out the calculating of algorithmic derivation part.API based on Python provides convenient calling interface.The advantages of present invention has image processing speed fast, and hardware resource requirements are few, facilitates transplanting and exploitation.
Description
Technical field
The present invention relates to the target detection techniques based on embedded platform, and in particular to a kind of intelligent monitoring based on PYNQ
System and monitoring method.
Background technique
Video monitoring is a sub-industry of security industry, during 2010-2017, China's video monitoring market scale from
24200000000 yuan rise to 112,400,000,000 yuan, and average annual recombination rate is up to 24.53%.With the construction of China's Transportation facilities, and
The construction of " safe city " accelerates, it is contemplated that is expected to reach 155,800,000,000 yuan to the year two thousand twenty China video monitoring market scale, to 2023
It is expected in year break through 190,000,000,000 yuan.And intelligent will be the following long-term developing direction of video monitoring.Therefore artificial intelligence will
Play the part of more and more important role in monitoring system.Target detection is a hot topic of computer vision and Digital Image Processing
Direction and the core of intelligent monitor system are reduced the consumption to human capital by computer vision, are had important
Realistic meaning.Due to the extensive utilization of deep learning, algorithm of target detection has obtained more quickly development.
PYNQ development board is added to the support to python on the basis of original Zynq framework.Make embedded programming people
Member can be without that can give full play to Xilinx Zynq All in the case where designing programmable logic circuit
Programmable SoC(APSoC) function.PYNQ is integrated with arm processor and FPGA programmable logic device, with routine
Unlike mode, by PYNQ, user can be used Python and carry out APSoC programming, and code can be directly on PYNQ
It is developed and is tested.By PYNQ, programmable logic circuit will import as hardware library and pass through its API and be programmed,
Mode and importing and programming software library are essentially identical.Python is widely used in as a kind of simple scripting language of gracefulness
Every field, the control system based on Python exploitation will have very high portability.
Traditional monitoring system generally requires artificial intervention, for example plays back when traffic accident, theft generally require
Monitoring misses the best opportunity, has very big time delay, and often cost of human resources is relatively high.As long as can detect automatically
The information needed out can provide feedback in time, and machine replaces manually capable of also reducing cost.
Summary of the invention
For technical problem of the existing technology, the present invention provides a kind of intelligent monitor system and prison based on PYNQ
Prosecutor method.
A kind of intelligent monitor system based on PYNQ, it is described including the camera connected by USB and PYNQ processing system
PYNQ processing system includes arm processor and FPGA, it is characterised in that the PYNQ processing system include multi-target detection module,
The api interface of general convolutional neural networks accelerator IP and Python, the multi-target detection module transplanting and optimization faster-
RCNN multi-target detection algorithm optimizes the structure of AlexNet network as feedforward network, and cluster to testing result kmeans；
The faster-RCNN multi-target detection algorithm includes the extraction module of Suggestion box, svm classifier module, linear regression amendment mould
Block, convolution module, pond module, full articulamentum module.
Further, the convolution module, pond module, full articulamentum module use general convolution nerve net in FPGA
Network accelerator IP is calculated.
Further, the api interface of the Python include general convolutional neural networks accelerator IP configuration parameter,
Data are carried, calculating is executed, obtains state.
Further, the extraction module of the Suggestion box, svm classifier module, linear regression correction module and kmeans
Cluster module calculates in arm processor.
Further, the arm processor pre-process characteristic pattern is stored in DDR, by row input, using AXI-lite
Bus marco, DMA transfer image data.
Further, the general convolutional neural networks accelerator IP includes computing unit, inside the computing unit
Using ranks multiplexing, 6 grades of flowing water, the general convolutional neural networks accelerator IP may be selected to realize convolution, Chi Hua, activation primitive
Function, can with the size of customized core, step-length, mend 0.
A kind of monitoring method based on the intelligent monitor system based on PYNQ, includes the following steps:
(1) trained network weight is imported in the SD card of PYNQ, then reads weight into DDR from SD；
(2) Image Acquisition, control USB camera acquires image, and is transmitted on a frame picture to PYNQ by USB interface；At arm
Reason device is written to the characteristic pattern for the input format that image preprocessing is AlexNet network in DDR；
(3) API Configuration network parameter is called, corresponding register is controlled by AXI_lite, not according to every layer of network structure
Together, configure core size, step-length, whether mend 0 and the layer be convolutional layer or pond layer, if need the information such as activation primitive；
(4) start general convolutional neural networks accelerator IP to be calculated, FPGA carries feature by row from DDR automatically by DMA
Result is write back DDR after calculating by diagram data, and circulation executes each layer, and positive derive for completing convolution sum pond layer calculates；
(5) arm processor intervention, detection flag bit judge whether to complete to calculate；
(6) selection that Suggestion box is carried out according to anchor point box, being cut out according to Suggestion box may carry out for the Partial Feature figure of target
The pond ROI, calls general convolutional neural networks accelerator IP to carry out the calculating of full articulamentum again, and the realization of full articulamentum is to turn
Change the convolutional calculation that length and width are 1 into；
(7) classified to full articulamentum calculated result with support vector machines, corrected to obtain the bounding box of target with regression model
Coordinate, and the target that screening repeats identification is clustered with kmeans.
Further, the Suggestion box can configure three kinds of areas, three kinds of scales, the Suggestion box of totally 9 kinds of forms.
The present invention is suitable for mainstream intelligent algorithm, has image processing speed fast, hardware resource requirements are few, facilitate shifting
The advantages of planting with exploitation.
Detailed description of the invention
Fig. 1 is the structural diagram of the present invention.
Specific embodiment
With reference to the accompanying drawings of the specification, as shown in Figure 1, being described further to technical solution of the present invention, particular technique
Scheme is as follows:
The present invention provides a kind of intelligent monitor system based on PYNQ development platform, including camera and PYNQ processing system,
Camera is connected with PYNQ processing system by USB.The PYNQ processing system includes arm processor and FPGA.Camera is adopted
Collect image information, pre-processed by USB transmission to arm processor, processing result inputs FPGA again and carries out positive derivation, quickly examines
Multiple target is surveyed, and makes corresponding control.By software-hardware synergism module, detection real-time is improved.
PYNQ is added to the support to python on the basis of original Zynq framework.It is integrated with arm processor and FPGA
Programmable logic device, programmable logic circuit will import as hardware library and pass through its API and be programmed.Software and hardware is facilitated to assist
With implementation, the high positive derivation part of calculating multiplicity, which is placed in FPGA, to be handled, and the big part of computation complexity is handled with software.
The PYNQ processing system mainly includes multi-target detection module, general convolutional neural networks acceleration IP and is based on
The api interface of python.The multi-target detection module carries out the detection of multiple target by the transplanting of algorithm with optimization, selects first
With multi-target detection algorithm faster-RCNN, for the hardware resource of the platform, improve the structure of AlexNet network as
Feedforward network；Then kmeans cluster is carried out to testing result, improves accuracy.For the ease of multi-target detection algorithm
Caffe to PYNQ is transplanted in the realization of faster-RCNN.
The multi-target detection algorithm faster-RCNN includes the extraction module of Suggestion box, svm classifier module, linear time
Return correction module, convolution module, pond module, full articulamentum module, the extraction mould of the software-hardware synergism module, Suggestion box
The part such as block, svm classifier module, linear regression correction module and kmean cluster module is calculated using arm processor；It calculates simultaneously
The high positive part that derives of the big repetitive rate of row degree is calculated using convolutional neural networks accelerator IP general in FPGA.The general use volume
Product neural network accelerator IP, using AXI-lite bus marco, DMA transfer image data, may be selected to realize convolution, Chi Hua,
The function of activation primitive can be suitable for various convolutional neural networks with the size of customized core, step-length, benefit 0 etc..
General convolutional neural networks accelerator IP includes computing unit, is flowed inside computing unit using ranks multiplexing, 6 grades
Water, acceleration effect are good.Loop configuration and the general convolutional neural networks accelerator IP of calling realize the convolution mould in faster-RCNN
Block, pond module, full articulamentum module.The API based on Python, the configuration of general convolutional neural networks accelerator IP
Parameter, carry data, executes calculatings, acquisition state has all been packaged into the interface of Python, algorithm realization during easily
Insertion is called, and calling and transplanting are facilitated.
A kind of intelligent control method based on PYNQ, includes the following steps:
Trained network weight is imported in the SD card of PYNQ, then reads weight into DDR from SD.
Image Acquisition, control USB camera acquires image, and is transmitted on a frame picture to PYNQ by USB interface.arm
Processor is the input format of AlexNet network image preprocessing, and 8 triple channel 224*224 pixels are written in DDR.Cause
The data volume for being characterized figure is very big, and resource is fewer in FPGA, can not leave, therefore is stored in relatively much larger DDR, later
Enter FPGA by row write according to general convolutional neural networks accelerator IP to be calculated.
API Configuration network parameter is called, corresponding register is controlled by AXI_lite, not according to every layer of network structure
Together, configure core size, step-length, whether mend 0 and the layer be convolutional layer or pond layer, if need the information such as activation primitive,
Can be convenient indicates various network structures.Start general convolutional neural networks accelerator IP to be calculated, the end FPGA can lead to automatically
It crosses DMA and carries feature diagram data by row from DDR, result is write back DDR after calculating, does not during which need arm processor intervention, is only needed
Flag bit is detected to judge whether to complete to calculate.Circulation executes each layer, and positive derive for completing convolution sum pond layer calculates.
The selection that Suggestion box is carried out according to anchor point box, can configure three kinds of areas, three kinds of scales, the suggestion of totally 9 kinds of forms
Frame.The pond ROI may be carried out for the Partial Feature figure of target by being cut out according to Suggestion box, call general convolutional neural networks again
Accelerator IP carries out the calculating of full articulamentum, and the realization of full articulamentum is the convolutional calculation for being converted into length and width and being 1.
Classified to full articulamentum calculated result with support vector machines, is corrected to obtain the bounding box of target with regression model
Coordinate, and the target that screening repeats identification is clustered with kmeans, improve accuracy.
Corresponding control is carried out according to the result of identification.
A kind of unmanned ticket-checking system of the amusement park of the intelligent monitor system based on PYNQ, for distinguishing adult ticket, child
Ticket, family's ticket.
Step 1, parent carry small children ticket checking, provide ticket information in inlet, control camera starts to acquire image.And
It is transmitted on a frame picture to the intelligent monitor system based on PYNQ by USB interface.Arm processor is to acquired image information
It pre-processes, is converted into the characteristic pattern of 8 triple channel 224*224 pixels, and be written in DDR.
Step 2 calls API to configure every layer parameter, constructs network.Then start general convolutional neural networks accelerator IP
Start to calculate feedforward network, FPGA presses row by DMA automatically and carries feature diagram data, calculates result and write back in DDR.Work as calculating
Flag bit can be changed by finishing, and notify arm processor.
Step 3 carries out the selection of Suggestion box according to anchor point box, and cutting out in calculated result to be the part of target
Characteristic pattern carries out the pond ROI, and general convolutional neural networks accelerator IP is called to carry out the calculating of full articulamentum again.To full connection
Layer calculated result is classified with support vector machines, is corrected to obtain the bounding box coordinates of target with regression model, and use kmeans
Cluster screening repeats the target of identification.
Step 4 finally obtains the classification and bounding box of testing result multiple target, judges in entrance area, is two big
People, a child is corresponding with family's ticket information, and control motor, which opens the door, lets pass.
Claims (8)
1. a kind of intelligent monitor system based on PYNQ, it is characterised in that: handled including the camera connected by USB and PYNQ
System, the PYNQ processing system include arm processor and FPGA, it is characterised in that the PYNQ system includes multi-target detection
The api interface of module, general convolutional neural networks accelerator IP and Python, the multi-target detection module transplanting and optimization
Faster-RCNN multi-target detection algorithm optimizes the structure of AlexNet network as feedforward network, and to testing result
Kmeans cluster；The faster-RCNN multi-target detection algorithm includes the extraction module of Suggestion box, svm classifier module, linear
Return correction module, convolution module, pond module, full articulamentum module.
2. a kind of intelligent monitor system based on PYNQ according to claim 1, it is characterised in that the convolution module,
Pond module, full articulamentum module are calculated using convolutional neural networks accelerator IP general in FPGA.
3. a kind of intelligent monitor system based on PYNQ according to claim 1, it is characterised in that the Python's
Api interface includes the configuration parameter of general convolutional neural networks accelerator IP, carries data, executes calculating, acquisition state.
4. a kind of intelligent monitor system based on PYNQ according to claim 1, it is characterised in that the Suggestion box
Extraction module, svm classifier module, linear regression correction module and kmeans cluster module calculate in arm processor.
5. a kind of intelligent monitor system based on PYNQ according to claim 2, it is characterised in that the arm processor is pre-
Handle characteristic pattern is stored in DDR, by row input, using AXI-lite bus marco, DMA transfer image data.
6. a kind of intelligent monitor system based on PYNQ according to claim 1, it is characterised in that the general convolution
Neural network accelerator IP includes computing unit, using ranks multiplexing, 6 grades of flowing water, the general use volume inside the computing unit
Product neural network accelerator IP may be selected realize convolution, the function of Chi Hua, activation primitive, can with the size of customized core, step-length,
Mend 0.
7. a kind of monitoring method based on any intelligent monitor system based on PYNQ of claim 1-6, feature exist
In including the following steps:
(1) trained network weight is imported in the SD card of PYNQ, then reads weight into DDR from SD；
(2) Image Acquisition, control USB camera acquires image, and is transmitted on a frame picture to PYNQ by USB interface；At arm
Reason device is written to the characteristic pattern for the input format that image preprocessing is AlexNet network in DDR；
(3) API Configuration network parameter is called, corresponding register is controlled by AXI_lite, not according to every layer of network structure
Together, configure core size, step-length, whether mend 0 and the layer be convolutional layer or pond layer, if need the information such as activation primitive；
(4) start general convolutional neural networks accelerator IP to be calculated, FPGA carries feature by row from DDR automatically by DMA
Result is write back DDR after calculating by diagram data, and circulation executes each layer, and positive derive for completing convolution sum pond layer calculates；
(5) arm processor intervention, detection flag bit judge whether to complete to calculate；
(6) selection that Suggestion box is carried out according to anchor point box, being cut out according to Suggestion box may carry out for the Partial Feature figure of target
The pond ROI, calls general convolutional neural networks accelerator IP to carry out the calculating of full articulamentum again, and the realization of full articulamentum is to turn
Change the convolutional calculation that length and width are 1 into；
(7) classified to full articulamentum calculated result with support vector machines, corrected to obtain the bounding box of target with regression model
Coordinate, and the target that screening repeats identification is clustered with kmeans.
8. a kind of intelligent control method based on PYNQ according to claim 7, it is characterised in that affiliated Suggestion box can
To configure three kinds of areas, three kinds of scales, the Suggestion box of totally 9 kinds of forms.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201910661356.2A CN110414401B (en) | 2019-07-22 | 2019-07-22 | PYNQ-based intelligent monitoring system and monitoring method |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201910661356.2A CN110414401B (en) | 2019-07-22 | 2019-07-22 | PYNQ-based intelligent monitoring system and monitoring method |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN110414401A true CN110414401A (en) | 2019-11-05 |
| CN110414401B CN110414401B (en) | 2022-02-15 |
Family
ID=68362350
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN201910661356.2A Active CN110414401B (en) | 2019-07-22 | 2019-07-22 | PYNQ-based intelligent monitoring system and monitoring method |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN110414401B (en) |
Cited By (6)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN112580751A (en) * | 2020-12-31 | 2021-03-30 | 杭州电子科技大学 | Snore identification device based on ZYNQ and deep learning |
| CN112616043A (en) * | 2020-12-22 | 2021-04-06 | 杭州电子科技大学 | PYNQ-based neural network identification video monitoring alarm system and method |
| CN112991382A (en) * | 2019-12-02 | 2021-06-18 | 中国科学院国家空间科学中心 | PYNQ frame-based heterogeneous visual target tracking system and method |
| CN113126767A (en) * | 2021-04-25 | 2021-07-16 | 合肥工业大学 | PYNQ and multi-mode brain-computer interface-based aircraft control system and method |
| WO2021237608A1 (en) * | 2020-05-28 | 2021-12-02 | 京东方科技集团股份有限公司 | Target detection method based on heterogeneous platform, and terminal device and storage medium |
| CN116630709A (en) * | 2023-05-25 | 2023-08-22 | 中国科学院空天信息创新研究院 | Hyperspectral image classification device and method capable of configuring mixed convolutional neural network |
Citations (4)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN103158620A (en) * | 2013-03-25 | 2013-06-19 | 中国电子科技集团公司第三十八研究所 | Vehicle pedestrian detection tracking early warning system |
| CN105975931A (en) * | 2016-05-04 | 2016-09-28 | 浙江大学 | Convolutional neural network face recognition method based on multi-scale pooling |
| CN108256636A (en) * | 2018-03-16 | 2018-07-06 | 成都理工大学 | A kind of convolutional neural networks algorithm design implementation method based on Heterogeneous Computing |
| CN109167966A (en) * | 2018-09-29 | 2019-01-08 | 南京邮电大学南通研究院有限公司 | Image dynamic detection system and method based on FPGA+ARM |
-
2019
- 2019-07-22 CN CN201910661356.2A patent/CN110414401B/en active Active
Patent Citations (4)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN103158620A (en) * | 2013-03-25 | 2013-06-19 | 中国电子科技集团公司第三十八研究所 | Vehicle pedestrian detection tracking early warning system |
| CN105975931A (en) * | 2016-05-04 | 2016-09-28 | 浙江大学 | Convolutional neural network face recognition method based on multi-scale pooling |
| CN108256636A (en) * | 2018-03-16 | 2018-07-06 | 成都理工大学 | A kind of convolutional neural networks algorithm design implementation method based on Heterogeneous Computing |
| CN109167966A (en) * | 2018-09-29 | 2019-01-08 | 南京邮电大学南通研究院有限公司 | Image dynamic detection system and method based on FPGA+ARM |
Non-Patent Citations (2)
| Title |
|---|
| 寇大磊 等: "基于深度学习的目标检测框架进展研究", 《计算机工程与应用》 * |
| 董亦博: "基于PYNQ的图像视觉显著性检测系统设计", 《中国优秀硕士学位论文全文数据库 信息科技辑》 * |
Cited By (8)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN112991382A (en) * | 2019-12-02 | 2021-06-18 | 中国科学院国家空间科学中心 | PYNQ frame-based heterogeneous visual target tracking system and method |
| CN112991382B (en) * | 2019-12-02 | 2024-04-09 | 中国科学院国家空间科学中心 | Heterogeneous visual target tracking system and method based on PYNQ framework |
| WO2021237608A1 (en) * | 2020-05-28 | 2021-12-02 | 京东方科技集团股份有限公司 | Target detection method based on heterogeneous platform, and terminal device and storage medium |
| CN112616043A (en) * | 2020-12-22 | 2021-04-06 | 杭州电子科技大学 | PYNQ-based neural network identification video monitoring alarm system and method |
| CN112580751A (en) * | 2020-12-31 | 2021-03-30 | 杭州电子科技大学 | Snore identification device based on ZYNQ and deep learning |
| CN113126767A (en) * | 2021-04-25 | 2021-07-16 | 合肥工业大学 | PYNQ and multi-mode brain-computer interface-based aircraft control system and method |
| CN116630709A (en) * | 2023-05-25 | 2023-08-22 | 中国科学院空天信息创新研究院 | Hyperspectral image classification device and method capable of configuring mixed convolutional neural network |
| CN116630709B (en) * | 2023-05-25 | 2024-01-09 | 中国科学院空天信息创新研究院 | Hyperspectral image classification device and method capable of configuring mixed convolutional neural network |
Also Published As
| Publication number | Publication date |
|---|---|
| CN110414401B (en) | 2022-02-15 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN110414401A (en) | A kind of intelligent monitor system and monitoring method based on PYNQ | |
| CN108647665A (en) | Vehicle real-time detection method of taking photo by plane based on deep learning | |
| CN111340195B (en) | Training method and device for network model, image processing method and storage medium | |
| CN109493303A (en) | A kind of image defogging method based on generation confrontation network | |
| CN109816769A (en) | Scene based on depth camera ground drawing generating method, device and equipment | |
| CN111914937A (en) | Lightweight improved target detection method and detection system | |
| CN105654066A (en) | Vehicle identification method and device | |
| CN108596338A (en) | A kind of acquisition methods and its system of neural metwork training collection | |
| CN110296660B (en) | Method and device for detecting livestock body ruler | |
| CN110532883A (en) | On-line tracking is improved using off-line tracking algorithm | |
| CN109805963A (en) | The determination method and system of one Endometrium parting | |
| CN110795975B (en) | Face false detection optimization method and device | |
| US20230137337A1 (en) | Enhanced machine learning model for joint detection and multi person pose estimation | |
| CN111507325B (en) | Industrial visual OCR recognition system and method based on deep learning | |
| CN110334719A (en) | The method and system of object image are built in a kind of extraction remote sensing image | |
| CN115049966A (en) | GhostNet-based lightweight YOLO pet identification method | |
| CN116524183A (en) | Camouflage target detection method based on multitask adapter fine adjustment | |
| CN112766040A (en) | Method, device and apparatus for detecting residual bait and readable storage medium | |
| CN114299573A (en) | Video processing method and device, electronic equipment and storage medium | |
| CN110516572A (en) | A kind of method, electronic equipment and storage medium identifying competitive sports video clip | |
| CN110533688A (en) | Follow-on method for tracking target, device and computer readable storage medium | |
| CN114155551A (en) | Improved pedestrian detection method and device based on YOLOv3 under complex environment | |
| CN110175588A (en) | A kind of few sample face expression recognition method and system based on meta learning | |
| CN116524195B (en) | Semantic segmentation method, semantic segmentation device, electronic equipment and storage medium | |
| CN108710820A (en) | Infantile state recognition methods, device and server based on recognition of face |
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 |