CN109711552A - A kind of data processing system and electronic equipment - Google Patents
A kind of data processing system and electronic equipment Download PDFInfo
- Publication number
- CN109711552A CN109711552A CN201910235429.1A CN201910235429A CN109711552A CN 109711552 A CN109711552 A CN 109711552A CN 201910235429 A CN201910235429 A CN 201910235429A CN 109711552 A CN109711552 A CN 109711552A
- Authority
- CN
- China
- Prior art keywords
- fpga
- main
- characteristic information
- image
- feature vector
- Prior art date
- Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
- Pending
Links
Landscapes
- Image Analysis (AREA)
Abstract
The invention discloses a kind of data processing system and electronic equipment, the system include: main on-site programmable gate array FPGA and at least one from FPGA, wherein the main FPGA and each from passing through chip2chip bus interconnection between FPGA；The main FPGA, the characteristic information in image for extracting input；From FPGA, for obtaining the character pair vector in the characteristic information that the main FPGA is extracted, analyzing the described eigenvector of acquisition according to itself demand, bandwidth, computing capability and the efficiency of data processing can be effectively improved, the delay of system is lowered.
Description
Technical field
The present invention relates to artificial intelligence field more particularly to a kind of data processing system and electronic equipments.
Background technique
With the development of the smart machines such as industrial robot, Intelligent mobile equipment, these equipment are usually needed to received
Various data carry out extremely complex processing, and nerve net Luoque with its powerful computing capability and machine learning ability preferably
The complex process for carrying out data, becomes currently a popular one of the mode of data processing, still, existing neural network accelerator
Bandwidth or computing capability it is limited, and then increase the delay of neural network, and reduce neural network to data processing
Efficiency.As for other data processing methods to data carry out complex process efficiency it is lower, delay it is bigger.
Summary of the invention
The embodiment of the invention provides a kind of data processing system and electronic equipments, to solve at data in the prior art
It manages low efficiency, postpone big problem.
The embodiment of the invention provides a kind of data processing system, the system comprises: main field programmable gate array
FPGA and at least one from FPGA, wherein the main FPGA and each being interconnected between FPGA；
The main FPGA, the characteristic information in image for extracting input；
From FPGA, for according to itself demand, obtain the character pair in the characteristic information that the main FPGA is extracted to
Amount, analyzes the described eigenvector of acquisition.
Further, the main FPGA and each between FPGA pass through chip2chip bus interconnection.
Further, it is described from FPGA include first from FPGA；
Described first from FPGA, for obtaining first of corresponding target detection in the characteristic information that the main FPGA is extracted
Feature vector is split identification to described image according to the first eigenvector, determines the target in described image.
Further, it is described from FPGA include second from FPGA；
Described second from FPGA, for obtaining second of corresponding semantic segmentation in the characteristic information that the main FPGA is extracted
Described image is divided into the region with different semantic informations, and mark according to the second feature vector by feature vector
Each corresponding semantic label in region.
Further, it is described from FPGA include third from FPGA；
The third is from FPGA, for obtaining the third of corresponding motion detection in the characteristic information that the main FPGA is extracted
Feature vector determines goal object in described image and the goal object described according to the third feature vector
Position in image.
Further, it is described from FPGA include the 4th from FPGA；
Described 4th from FPGA, for obtaining the 4th of corresponding image enhancement in the characteristic information that the main FPGA is extracted the
Feature vector, according to the fourth feature vector and the pre-set process demand to described image, in described image
Data carry out corresponding conversion process or increase some other data information, the feature to require emphasis in prominent described image
And/or inhibit described image in be not required to it is however emphasized that feature.
Further, the backbone network of neural network is deployed in the main FPGA, wherein the training of the backbone network
Process includes: each training sample for obtaining training sample and concentrating, and is provided with corresponding label in training sample in advance, will be every
A training sample is input in neural network, right according to the output of the neural network and the corresponding label of each training sample
The neural network is trained, and the composition backbone network of characteristic information identification is realized in the neural network that training is completed.
The embodiment of the invention provides a kind of electronic equipment, the electronic equipment includes any of the above-described data processing
System.
The embodiment of the invention provides a kind of data processing system and electronic equipment, which includes: main field-programmable
Gate array FPGA and at least one from FPGA, wherein the main FPGA and each mutual from chip2chip bus is passed through between FPGA
Even；The main FPGA, the characteristic information in image for extracting input；From FPGA, for obtaining institute according to itself demand
The character pair vector in the characteristic information that main FPGA is extracted is stated, the described eigenvector of acquisition is analyzed.
Due to being carried out from data at least one from the FPGA array of FPGA in the embodiment of the present invention using including main FPGA
Reason, therefore bandwidth, computing capability and the efficiency of data processing can be effectively improved, lower the delay of system.
Detailed description of the invention
To describe the technical solutions in the embodiments of the present invention more clearly, make required in being described below to embodiment
Attached drawing is briefly introduced, it should be apparent that, drawings in the following description are only some embodiments of the invention, for this
For the those of ordinary skill in field, without creative efforts, it can also be obtained according to these attached drawings other
Attached drawing.
Fig. 1 is a kind of structural schematic diagram for data processing system that the embodiment of the present invention 1 provides；
Fig. 2 is a kind of operation principle schematic diagram for data processing system that the embodiment of the present invention 2 provides.
Specific embodiment
The present invention will be describe below in further detail with reference to the accompanying drawings, it is clear that described embodiment is only this
Invention a part of the embodiment, instead of all the embodiments.Based on the embodiments of the present invention, those of ordinary skill in the art exist
All other embodiment obtained under the premise of creative work is not made, shall fall within the protection scope of the present invention.
Embodiment 1:
Fig. 1 is a kind of structural schematic diagram of data processing system provided in an embodiment of the present invention, the system comprises: main scene can
Program gate array (Field-Programmable Gate Array, FPGA) and at least one from FPGA, wherein the master
It FPGA and is each interconnected between FPGA；
The main FPGA, the characteristic information in image for extracting input；
From FPGA, for according to itself demand, obtain the character pair in the characteristic information that the main FPGA is extracted to
Amount, analyzes the described eigenvector of acquisition.
In order to improve data-handling efficiency, several FPGA are formed FPGA array structure, the array by the embodiment of the present invention
In structure include main FPGA and at least one from FPGA, main FPGA and from the connection type between FPGA without limitation, specifically
Ground, main FPGA and from chip2chip bus interconnection can be passed through between FPGA.
Wherein, each FPGA in FPGA array can realize the parallel place of data with the data of calculation processing different function
Reason.Specifically, main FPGA can extract the characteristic information in input picture, bear according to pre-set information extraction rule
Blame different function slave FPGA then can according to realize itself function demand, obtain this feature information in corresponding feature to
Amount, then analyzes this feature vector got, gets corresponding results.
Specifically, being deployed with the backbone network of deep neural network in the main FPGA, which is responsible for feature and mentions
It takes, each functional module for carrying out subsequent analysis to the feature that backbone network extracts can be disposed from FPGA, and can root
The deployment from FPGA is flexibly carried out according to demand, that is to say, that the structure extension of the FPGA array is flexible, and can be adapted for
Several scenes based on deep neural network, and because the structure for using FPGA array proposes whole computing capability
Height, bandwidth improves, and reduces system delay.
Due to being carried out from data at least one from the FPGA array of FPGA in the embodiment of the present invention using including main FPGA
Reason, therefore bandwidth, computing capability and the efficiency of data processing can be effectively improved, lower the delay of system.
Embodiment 2:
On the basis of the above embodiments, in order to more rationally using each from FPGA, it is described from FPGA include first from FPGA；
Described first from FPGA, for obtaining first of corresponding target detection in the characteristic information that the main FPGA is extracted
Feature vector is split identification to described image according to the first eigenvector, determines the target in described image.
Data processing system provided in an embodiment of the present invention can be applied in automatic driving vehicle, should include the from FPGA
One from FPGA, and first is deployed with the functional module for carrying out target detection from FPGA.
First from the characteristic information of the available main FPGA of the FPGA related target detection extracted, and to this feature information into
Row processing analysis, to determine the target in input picture based on the analysis results；Also the spy that available main FPGA is extracted
The first eigenvector of corresponding target detection, is split knowledge to described image according to the first eigenvector in reference breath
Not, the target in described image is determined.In order to improve the accuracy of target detection, the embodiment of the present invention preferably obtains main FPGA and mentions
The first eigenvector of corresponding target detection in the characteristic information taken, according to the first eigenvector to described image into
Row segmentation identification, determines the target in described image.
Above-mentioned data processing system can be applied in safety monitoring or medical image, also should may include second from FPGA
From FPGA, second is deployed with the functional module for carrying out semantic segmentation from FPGA.
Second from the characteristic information of the available main FPGA of the FPGA related semantic segmentation extracted, and to this feature information into
Row processing analysis, thus based on the analysis results by the image segmentation of input at the region with different semantic informations, and mark
Each corresponding semantic label in region；Also the second of corresponding semantic segmentation is special in the characteristic information that available main FPGA is extracted
It levies vector and each area by the image segmentation at the region with different semantic informations, and is marked according to the second feature vector
The corresponding semantic label in domain.In order to improve the accuracy of semantic segmentation, the embodiment of the present invention preferably obtains the spy that main FPGA is extracted
The second feature vector of corresponding semantic segmentation in reference breath, according to the second feature vector, by the image segmentation at not
With the region of semantic information, and mark the corresponding semantic label in each region.
Above-mentioned data processing system can be applied in safety monitoring, should include third from FPGA from FPGA, third from
The functional module for carrying out motion detection is deployed in FPGA.
Third from the characteristic information of the available main FPGA of the FPGA related motion detection extracted, and to this feature information into
Row processing analysis, to determine goal object in input picture and the goal object based on the analysis results in the images
Position；Also the third feature vector of corresponding motion detection in the characteristic information that available main FPGA is extracted, according to the third
Feature vector determines the position of goal object and the goal object in the images in input picture.In order to improve movement
The accuracy of detection, the third that the embodiment of the present invention preferably obtains corresponding motion detection in the characteristic information that main FPGA is extracted are special
Vector is levied, according to the third feature vector, determines goal object in input picture and the goal object in the images
Position.
Above-mentioned data processing system can be applied in medical image, should include the 4th from FPGA from FPGA, the 4th from
The functional module for carrying out image enhancement is deployed in FPGA.
4th from the characteristic information of the available main FPGA of the FPGA related image enhancement extracted, and to this feature information into
Row processing analysis, thus based on the analysis results and the pre-set process demand to input picture, to the number in the image
According to carrying out corresponding conversion process or increase some other data informations, the feature that requires emphasis in the prominent image and/or
Inhibit described image in be not required to it is however emphasized that feature；Also corresponding image in this feature information that the available main FPGA is extracted
The fourth feature vector of enhancing, according to the fourth feature vector and the pre-set process demand to the image, to the figure
Data as in carry out corresponding conversion process or increase some other data informations, the spy to require emphasis in the prominent image
Levy and/or inhibit to be not required in the image it is however emphasized that feature.In order to improve the accuracy of image enhancement, the embodiment of the present invention is preferred
The fourth feature vector for obtaining corresponding image enhancement in this feature information that the main FPGA is extracted, according to the fourth feature to
Amount and the pre-set process demand to the image carry out corresponding conversion process or increase to the data in the image
Some other data informations, the feature to require emphasis in the prominent image and/or inhibit to be not required in the image it is however emphasized that spy
Sign.
The backbone network of neural network is deployed in main FPGA, in order to enable backbone network to extract characteristic information,
It needs to be trained neural network, the specific training process of neural network is different because of actual scene difference, but, to nerve
The specific training process of network is similar.Nerve net to be applied in automatic driving vehicle in embodiments of the present invention
The neural network of network, i.e. progress target detection and semantic segmentation illustrates the training of the neural network that is, for blitznet network
Process:
Each training sample that training sample is concentrated is obtained first, and each training sample is an image, is answered according to neural network
Scene is provided with corresponding label in training sample in advance, for example, if the neural network needs to carry out semantic knowledge
Not, then include different semantic labels in training sample, if the neural network needs to carry out the identification of target, train sample
Include target detection label in this, specifically corresponding label, same trained sample can be arranged to training sample according to demand
It may include the label for realizing a variety of detections in this.
Then, each training sample is input in neural network, according to the output of the neural network and each training
The corresponding label of sample, is trained the neural network.
Convolutional neural networks (Convolutional for the image in each training sample, in the neural network
Neural Network, CNN) extract characteristic information in the image, and by this feature information classification storage to corresponding tensor
(tensor) buffer area；It is then from the tensor about semantic segmentation that the layer (Segment) of semantic segmentation is carried out in the neural network
The feature vector of corresponding semantic segmentation is obtained in buffer area, and according to the feature vector of these semantic segmentations, by the image point
It is cut into the region with different semantic informations, and marks the corresponding semantic label in each region；Then by these semantic label groups
At the set (concatenation) of a semantic label, the semantic label set is exported；Target inspection is carried out in the neural network
The layer for surveying (Detect), then obtain the feature vector of corresponding target detection out of tensor buffer area about target detection, and
According to the feature vector (vector) of these target detections, corresponding target detection label is added in the images, and export this
A little target detection labels.
After the completion of to neural metwork training, because the neural network is to realize final target detection and semantic segmentation
, and it is known in advance which layer, which can export characteristic information, in the neural network, therefore the neural network that training is completed
The middle layer for realizing characteristic information identification extracts, and the part layer of the neural network extracted constitutes the nerve net in main FPGA
The backbone network of network.It will realize that the layer of target detection extracts in the neural network of training completion, constitute and carry out target detection
The first functional module, which can be deployed to first from FPGA, will training complete neural network in
It realizes that the layer of semantic segmentation extracts, constitutes the second functional module for carrying out semantic segmentation, it can be by second functional module
Second is deployed to from FPGA.
The CNN can be understood as the backbone network in the neural network, and the convolutional neural networks model that the CNN is used
?
Geometry Group, VGG) model, for moving-vision application efficient convolutional neural networks (MobileNet) model.
It, can in order to improve the customization of data processing system since different scenes needs the function of realizing to have differences
With calculation power and broadband needed for each function type according to needed for actual application scenarios and each function type, to from
FPGA cut, is transplanted or the operations such as logic copy, to meet the actual demand of the application scenarios.For example, data processing system
System is applied in safety monitoring, if in the data processing system in addition to include third from FPGA further include other from FPGA,
Other can be cropped from FPGA at this time, then can increase be deployed with carry out the stream of people analysis and face recognition respectively correspond
Functional module slave FPGA, to reach preferably safety monitoring.Specifically, to being cut, transplanted or logic is multiple from FPGA
The operations such as system are the prior arts, and details are not described herein.
The function type that can also need to realize according to actual scene, is designed from being customized of FPGA, for example, if
Data processing system be applied in medical image, need to carry out target detection and image enhancement, then just by first from FPGA,
4th is connected by chip2chip bus with main FPGA from FPGA.
On the basis of the above embodiments, Fig. 2 is that a kind of work of data processing system provided in an embodiment of the present invention is former
Schematic diagram is managed, main FPGA extracts the characteristic information in input picture, from FPGA1 according to the demand of target detection, obtains this feature
Corresponding feature vector 1 in information, and this feature vector 1 is analyzed；From FPGA2 according to the demand of semantic segmentation, obtain
Corresponding feature vector 2 in this feature information, and this feature vector 2 is analyzed；From FPGA3 according to the need of motion detection
It asks, obtains corresponding feature vector 3 in this feature information, and analyze this feature vector 3；Increase from FPGAn according to image
Strong demand obtains corresponding feature vector n in this feature information, and analyzes this feature vector n；Specifically, each
Illustrated that details are not described herein in the above content from analytic process of the FPGA to the corresponding feature vector of acquisition.
Due in the embodiment of the present invention first from FPGA, for obtaining corresponding target in the characteristic information that main FPGA is extracted
The first eigenvector of detection is split identification to described image according to the first eigenvector, determines the mesh in the image
Mark is more rationally utilized from FPGA in this way.
Embodiment 3:
On the basis of above-mentioned each embodiment, the embodiment of the invention provides a kind of electronic equipment.The electronic equipment includes upper
State any data processing system.
Specifically, the electronic equipment may include FPGA and at least one from FPGA, wherein the main FPGA and each from
It is interconnected between FPGA；Specifically, the main FPGA and each between FPGA pass through chip2chip bus interconnection；The main FPGA can
The characteristic information in image to extract input；The institute that the main FPGA is extracted can be obtained according to the demand of itself from FPGA
The character pair vector in characteristic information is stated, the described eigenvector of acquisition is analyzed.
Further, the slave FPGA in the electronic equipment may include first from FPGA, this is first available from FPGA
The first eigenvector of corresponding target detection in the characteristic information that above-mentioned main FPGA is extracted, according to the first eigenvector
Identification is split to described image, determines the target in the image.
Further, the slave FPGA in the electronic equipment may include second from FPGA, this is second available from FPGA
The second feature vector of corresponding semantic segmentation in the characteristic information that the main FPGA is extracted, according to the second feature vector, by this
Image segmentation marks the corresponding semantic label in each region at the region with different semantic informations.
Further, the slave FPGA in the electronic equipment may include third from FPGA, and the third is available from FPGA
The third feature vector of corresponding motion detection is determined according to the third feature vector in the characteristic information that the main FPGA is extracted
The position of goal object and the goal object in the images in the image.
Further, the slave FPGA in the electronic equipment may include the 4th available from FPGA from FPGA, the 4th
The fourth feature vector of corresponding image enhancement in the characteristic information that the main FPGA is extracted, according to the fourth feature vector, and
The pre-set process demand to the image carries out corresponding conversion process to the data in the image or increases some other
Data information, the feature to require emphasis in the prominent image and/or inhibit to be not required in described image it is however emphasized that feature.
Further, the backbone network of neural network is deployed in the main FPGA in the electronic equipment, wherein the backbone network
The training process of network includes: each training sample for obtaining training sample and concentrating, and is provided in training sample in advance corresponding
Each training sample is input in neural network by label, corresponding according to the output of the neural network and each training sample
Label, which is trained, training complete neural network in realize characteristic information identification composition backbone network
Network.
Based on the same inventive concept, a kind of electronic equipment is additionally provided in the embodiment of the present invention, due to above-mentioned electronic equipment
The principle solved the problems, such as is similar to the principle that data processing system solves the problems, such as, therefore the implementation of above-mentioned electronic equipment may refer to
The implementation of data processing system, overlaps will not be repeated.
For systems/devices embodiment, since it is substantially similar to the method embodiment, so the comparison of description is simple
Single, the relevent part can refer to the partial explaination of embodiments of method.
It should be noted that, in this document, relational terms such as first and second and the like are used merely to a reality
Body or an operation are distinguished with another entity or another operation, without necessarily requiring or implying these entities
Or there are any actual relationship or orders between operation.
It should be understood by those skilled in the art that, embodiments herein can provide as method, system or computer program
Product.Therefore, the reality of complete hardware embodiment, complete Application Example or connected applications and hardware aspect can be used in the application
Apply the form of example.Moreover, it wherein includes the computer of computer usable program code that the application, which can be used in one or more,
The computer program implemented in usable storage medium (including but not limited to magnetic disk storage, CD-ROM, optical memory etc.) produces
The form of product.
The application is referring to method, the process of equipment (system) and computer program product according to the embodiment of the present application
Figure and/or block diagram describe.It should be understood that every one stream in flowchart and/or the block diagram can be realized by computer program instructions
The combination of process and/or box in journey and/or box and flowchart and/or the block diagram.It can provide these computer programs
Instruct the processor of general purpose computer, special purpose computer, Embedded Processor or other programmable data processing devices to produce
A raw machine, so that being generated by the instruction that computer or the processor of other programmable data processing devices execute for real
The device for the function of being specified in present one or more flows of the flowchart and/or one or more blocks of the block diagram.
These computer program instructions, which may also be stored in, is able to guide computer or other programmable data processing devices with spy
Determine in the computer-readable memory that mode works, so that it includes referring to that instruction stored in the computer readable memory, which generates,
Enable the manufacture of device, the command device realize in one box of one or more flows of the flowchart and/or block diagram or
The function of being specified in multiple boxes.
These computer program instructions also can be loaded onto a computer or other programmable data processing device, so that counting
Series of operation steps are executed on calculation machine or other programmable devices to generate computer implemented processing, thus in computer or
The instruction executed on other programmable devices is provided for realizing in one or more flows of the flowchart and/or block diagram one
The step of function of being specified in a box or multiple boxes.
Although preferred embodiments of the present invention have been described, it is created once a person skilled in the art knows basic
Property concept, then additional changes and modifications may be made to these embodiments.So it includes excellent that the following claims are intended to be interpreted as
It selects embodiment and falls into all change and modification of the scope of the invention.
Obviously, various changes and modifications can be made to the invention without departing from essence of the invention by those skilled in the art
Mind and range.If in this way, these of the invention modifications and variations belong to the claims in the present invention and its equivalent technologies range it
Interior, then the present invention is also intended to include these modifications and variations.
Claims (8)
1. a kind of data processing system, which is characterized in that the system comprises: main on-site programmable gate array FPGA and at least one
It is a from FPGA, wherein the main FPGA and each being interconnected between FPGA；
The main FPGA, the characteristic information in image for extracting input；
From FPGA, for according to itself demand, obtain the character pair in the characteristic information that the main FPGA is extracted to
Amount, analyzes the described eigenvector of acquisition.
2. the system as claimed in claim 1, which is characterized in that the main FPGA and each pass through between FPGA
Chip2chip bus interconnection.
3. the system as claimed in claim 1, which is characterized in that it is described from FPGA include first from FPGA；
Described first from FPGA, for obtaining first of corresponding target detection in the characteristic information that the main FPGA is extracted
Feature vector is split identification to described image according to the first eigenvector, determines the target in described image.
4. the system as claimed in claim 1, which is characterized in that it is described from FPGA include second from FPGA；
Described second from FPGA, for obtaining second of corresponding semantic segmentation in the characteristic information that the main FPGA is extracted
Described image is divided into the region with different semantic informations, and mark according to the second feature vector by feature vector
Each corresponding semantic label in region.
5. the system as claimed in claim 1, which is characterized in that it is described from FPGA include third from FPGA；
The third is from FPGA, for obtaining the third of corresponding motion detection in the characteristic information that the main FPGA is extracted
Feature vector determines goal object in described image and the goal object described according to the third feature vector
Position in image.
6. the system as claimed in claim 1, which is characterized in that it is described from FPGA include the 4th from FPGA；
Described 4th from FPGA, for obtaining the 4th of corresponding image enhancement in the characteristic information that the main FPGA is extracted the
Feature vector, according to the fourth feature vector and the pre-set process demand to described image, in described image
Data carry out corresponding conversion process or increase some other data information, the feature to require emphasis in prominent described image
And/or inhibit described image in be not required to it is however emphasized that feature.
7. the system as claimed in claim 1, which is characterized in that the backbone network of neural network is deployed in the main FPGA,
Wherein the training process of the backbone network includes: each training sample for obtaining training sample and concentrating, in advance in training sample
In be provided with corresponding label, each training sample is input in neural network, according to the output of the neural network and every
The corresponding label of a training sample, is trained the neural network, realizes that characteristic information is known in the neural network that training is completed
Other composition backbone network.
8. a kind of electronic equipment, which is characterized in that the electronic equipment includes data processing as claimed in claim 1
System.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201910235429.1A CN109711552A (en) | 2019-03-27 | 2019-03-27 | A kind of data processing system and electronic equipment |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201910235429.1A CN109711552A (en) | 2019-03-27 | 2019-03-27 | A kind of data processing system and electronic equipment |
Publications (1)
| Publication Number | Publication Date |
|---|---|
| CN109711552A true CN109711552A (en) | 2019-05-03 |
Family
ID=66265919
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN201910235429.1A Pending CN109711552A (en) | 2019-03-27 | 2019-03-27 | A kind of data processing system and electronic equipment |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN109711552A (en) |
Cited By (1)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| WO2021190639A1 (en) * | 2020-03-26 | 2021-09-30 | Guangdong Oppo Mobile Telecommunications Corp., Ltd. | Method for video recognition and related products |
Citations (3)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN101699414A (en) * | 2009-09-30 | 2010-04-28 | 曙光信息产业(北京)有限公司 | Data processing system |
| CN102298139A (en) * | 2011-05-18 | 2011-12-28 | 中国科学院计算技术研究所 | Two-dimensional windowing method of synthetic aperture radar (SAR) imaging system based on field programmable gate array (FPGA) |
| CN107423734A (en) * | 2016-05-23 | 2017-12-01 | 中国科学院计算技术研究所 | A kind of SAR image targets in ocean quick determination method and device |
-
2019
- 2019-03-27 CN CN201910235429.1A patent/CN109711552A/en active Pending
Patent Citations (3)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN101699414A (en) * | 2009-09-30 | 2010-04-28 | 曙光信息产业(北京)有限公司 | Data processing system |
| CN102298139A (en) * | 2011-05-18 | 2011-12-28 | 中国科学院计算技术研究所 | Two-dimensional windowing method of synthetic aperture radar (SAR) imaging system based on field programmable gate array (FPGA) |
| CN107423734A (en) * | 2016-05-23 | 2017-12-01 | 中国科学院计算技术研究所 | A kind of SAR image targets in ocean quick determination method and device |
Cited By (1)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| WO2021190639A1 (en) * | 2020-03-26 | 2021-09-30 | Guangdong Oppo Mobile Telecommunications Corp., Ltd. | Method for video recognition and related products |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN107563372B (en) | License plate positioning method based on deep learning SSD frame | |
| Biswas et al. | Classification of objects in video records using neural network framework | |
| CN107273832B (en) | License plate recognition method and system based on integral channel characteristics and convolutional neural network | |
| CN110378222A (en) | A kind of vibration damper on power transmission line target detection and defect identification method and device | |
| CN115830471B (en) | Multi-scale feature fusion and alignment domain self-adaptive cloud detection method | |
| CN104166841A (en) | Rapid detection identification method for specified pedestrian or vehicle in video monitoring network | |
| CN109948637A (en) | Object test equipment, method for checking object and computer-readable medium | |
| CN114332473B (en) | Object detection method, device, computer apparatus, storage medium, and program product | |
| CN112990282B (en) | Classification method and device for fine-granularity small sample images | |
| CN110349167A (en) | A kind of image instance dividing method and device | |
| CN111553414A (en) | In-vehicle lost object detection method based on improved Faster R-CNN | |
| Sun et al. | IRDCLNet: Instance segmentation of ship images based on interference reduction and dynamic contour learning in foggy scenes | |
| CN114973199A (en) | Rail transit train obstacle detection method based on convolutional neural network | |
| CN115984537A (en) | Image processing method and device and related equipment | |
| CN112668672A (en) | TensorRT-based target detection model acceleration method and device | |
| CN114581710A (en) | Image recognition method, device, equipment, readable storage medium and program product | |
| Xu et al. | Occlusion problem-oriented adversarial faster-RCNN scheme | |
| CN116740516A (en) | Target detection method and system based on multi-scale fusion feature extraction | |
| CN112395953A (en) | Road surface foreign matter detection system | |
| Yu et al. | SignHRNet: Street-level traffic signs recognition with an attentive semi-anchoring guided high-resolution network | |
| CN117911697A (en) | Hyperspectral target tracking method, system, medium and equipment based on large model segmentation | |
| CN109711552A (en) | A kind of data processing system and electronic equipment | |
| CN116168438A (en) | Key point detection method and device and electronic equipment | |
| Huang et al. | Real-time traffic sign detection model based on multi-branch convolutional reparameterization | |
| Mohana et al. | Object detection and classification algorithms using deep learning for video surveillance applications |
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 | ||
| RJ01 | Rejection of invention patent application after publication |
Application publication date: 20190503 |
|
| RJ01 | Rejection of invention patent application after publication |