WO2021003692A1 - Procédé de configuration d'algorithme, dispositif, système et plateforme mobile - Google Patents

Procédé de configuration d'algorithme, dispositif, système et plateforme mobile Download PDF

Info

Publication number
WO2021003692A1
WO2021003692A1 PCT/CN2019/095391 CN2019095391W WO2021003692A1 WO 2021003692 A1 WO2021003692 A1 WO 2021003692A1 CN 2019095391 W CN2019095391 W CN 2019095391W WO 2021003692 A1 WO2021003692 A1 WO 2021003692A1
Authority
WO
WIPO (PCT)
Prior art keywords
algorithm
identification
target
preset
recognition
Prior art date
Application number
PCT/CN2019/095391
Other languages
English (en)
Chinese (zh)
Inventor
万千
薛立君
Original Assignee
深圳市大疆创新科技有限公司
Priority date (The priority date 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 date listed.)
Filing date
Publication date
Application filed by 深圳市大疆创新科技有限公司 filed Critical 深圳市大疆创新科技有限公司
Priority to PCT/CN2019/095391 priority Critical patent/WO2021003692A1/fr
Priority to CN201980033933.1A priority patent/CN112400147A/zh
Publication of WO2021003692A1 publication Critical patent/WO2021003692A1/fr

Links

Images

Classifications

    • G—PHYSICS
    • G05—CONTROLLING; REGULATING
    • G05D—SYSTEMS FOR CONTROLLING OR REGULATING NON-ELECTRIC VARIABLES
    • G05D1/00—Control of position, course, altitude or attitude of land, water, air or space vehicles, e.g. using automatic pilots
    • G05D1/12—Target-seeking control

Definitions

  • the embodiments of this application relate to the field of computer technology, and in particular to an algorithm configuration method, device, system, and mobile platform.
  • the target recognition algorithm is a type of computer vision algorithm.
  • face recognition is a method that uses various sensors as input (such as a camera) to automatically recognize the face in the sensor’s field of view, and obtain its status The position and size of the screen.
  • the embodiments of the present application provide an algorithm configuration method, device, system, and movable platform to overcome the problem that when the target recognition algorithm is applied to a terminal device, it cannot run on low-performance device hardware or cannot take advantage of high-performance device hardware. .
  • an algorithm configuration method including:
  • the processing capability information select a target recognition algorithm package matching the processing capability information from a plurality of preset recognition algorithm packages; wherein, the target recognition algorithm package is used to control the local The device recognizes the preset target.
  • the embodiments of the present application provide another algorithm configuration method, including:
  • the control device obtains the processing capability information of the target device
  • the control device selects a target recognition algorithm package matching the processing capability information from a plurality of preset recognition algorithm packages;
  • the control device sends the target recognition algorithm package to the target device
  • the target device recognizes a preset target object according to the target recognition algorithm package.
  • the embodiments of the present application provide yet another algorithm configuration method, including:
  • processing capability information selecting a target recognition algorithm package matching the processing capability information from a plurality of preset recognition algorithm packages;
  • the target recognition algorithm package is sent to a target device, and the target recognition algorithm is used to instruct the target device to recognize a preset target according to the target recognition algorithm package.
  • an embodiment of the present application provides yet another algorithm configuration method, including:
  • a target recognition algorithm package sent by a control device, where the target recognition algorithm package is selected from a plurality of preset recognition algorithm packages according to the processing capability information of the target device, and the target recognition algorithm and the processing of the target device Match the capability information;
  • the preset target is recognized according to the target recognition algorithm package.
  • an embodiment of the present application provides an algorithm configuration device, including a first memory, a first processor, and a computer executable instruction stored in the first memory and running on the first processor, so The first processor implements the algorithm configuration method described in the first aspect and various possible designs of the first aspect when executing the computer-executed instruction.
  • an embodiment of the present application provides a control device, including a second memory, a second processor, and a computer-executable instruction stored in the second memory and running on the second processor, the The second processor implements the algorithm configuration method described in the third aspect and various possible designs of the third aspect when executing the computer-executed instruction.
  • an embodiment of the present application provides a target device, including a third memory, a third processor, and computer-executable instructions stored in the third memory and capable of running on the third processor.
  • the third processor implements the algorithm configuration methods described in the fourth aspect and various possible designs of the fourth aspect when executing the computer-executed instructions.
  • an embodiment of the present application provides an algorithm configuration system, including a control device and a target device; wherein,
  • the control device is used to obtain processing capability information of the target device; and according to the processing capability information, select a target recognition algorithm package matching the processing capability information from a plurality of preset recognition algorithm packages; The target recognition algorithm package is sent to the target device;
  • the target device is configured to recognize a preset target object according to the target recognition algorithm package.
  • an embodiment of the present application provides a movable platform, including:
  • the power system is provided in the body, and the power system is used to provide power to the movable platform; and the algorithm configuration device described in the fifth aspect above.
  • an embodiment of the present application provides another movable platform, including:
  • a power system arranged in the body, and the power system is used to provide power to the movable platform;
  • an embodiment of the present application provides yet another movable platform, including:
  • a power system arranged in the body, and the power system is used to provide power to the movable platform;
  • an embodiment of the present application provides yet another movable platform, including:
  • a power system arranged in the body, and the power system is used to provide power to the movable platform;
  • the algorithm configuration system described in the eighth aspect is provided in the body.
  • an embodiment of the present application provides yet another movable platform, including: a movable platform body and a control device; the movable platform body and the control device are connected wirelessly or wiredly;
  • the control device is used to obtain processing capability information of the movable platform body; according to the processing capability information, select a target recognition algorithm package matching the processing capability information from a plurality of preset recognition algorithm packages; The target recognition algorithm package is sent to the mobile platform ontology;
  • the movable platform body is used for recognizing a preset target according to the target recognition algorithm package.
  • an embodiment of the present application provides a computer-readable storage medium having computer-executable instructions stored in the computer-readable storage medium.
  • the processor executes the computer-executable instructions, the above first aspect and the first aspect are implemented.
  • the embodiments of the present application provide another computer-readable storage medium, and the computer-readable storage medium stores computer-executable instructions.
  • the processor executes the computer-executable instructions, the above third aspect and In the third aspect, various possible designs are described in the algorithm configuration method.
  • the embodiments of the present application provide yet another computer-readable storage medium, and the computer-readable storage medium stores computer-executable instructions.
  • the processor executes the computer-executable instructions, the above fourth aspect and In the fourth aspect, various possible designs are described in the algorithm configuration method.
  • the algorithm configuration method, device, system, and movable platform provided by the embodiments of the present application, the method obtains the processing capability information of the target device, and selects the processing capability from a plurality of preset identification algorithm packages according to the processing capability information
  • the target recognition algorithm package with matching information realizes the adaptive configuration of target recognition algorithm, so that it can make full use of hardware resources when running on high-performance hardware, and it can achieve the best compatibility when running on low-performance hardware.
  • the target recognition algorithm is applied to terminal equipment, there are problems that it cannot run on the hardware of low-performance equipment, or cannot take advantage of the hardware of high-performance equipment.
  • FIG. 1 is a schematic flowchart of an algorithm configuration method provided by an embodiment of this application.
  • FIG. 2 is a schematic flowchart of another algorithm configuration method provided by an embodiment of the application.
  • FIG. 3 is a schematic flowchart of yet another algorithm configuration method provided by an embodiment of the application.
  • FIG. 4 is a schematic diagram of the architecture of an algorithm configuration system provided by an embodiment of the application.
  • FIG. 5 is a schematic flowchart of yet another algorithm configuration method provided by an embodiment of this application.
  • FIG. 6 is a schematic flowchart of another algorithm configuration method provided by an embodiment of the application.
  • FIG. 7 is a schematic diagram of the hardware structure of an algorithm configuration device provided by an embodiment of the application.
  • FIG. 8 is a schematic diagram of the hardware structure of a control device provided by an embodiment of the application.
  • FIG. 9 is a schematic diagram of the hardware structure of a target device provided by an embodiment of the application.
  • FIG. 10 is a schematic structural diagram of an algorithm configuration system provided by an embodiment of this application.
  • FIG. 11 is a schematic structural diagram of a movable platform provided by an embodiment of the application.
  • FIG. 12 is a schematic structural diagram of another movable platform provided by an embodiment of this application.
  • FIG. 13 is a schematic structural diagram of yet another movable platform provided by an embodiment of the application.
  • FIG. 14 is a schematic structural diagram of another movable platform provided by an embodiment of the application.
  • FIG. 15 is a schematic structural diagram of another movable platform provided by an embodiment of the application.
  • Target recognition algorithm is a type of computer vision algorithm. Take face recognition as an example. Face recognition is a kind of face recognition that uses various sensors as input (such as a camera) to automatically recognize the face in the sensor's field of view, and get its image in the picture. Location and size of technology. However, when target recognition algorithms are applied to terminal devices, some high-precision algorithms rely on high-performance device hardware, while others sacrifice accuracy in order to run on low-performance device hardware. If the device hardware platforms for algorithm deployment are diverse, the above two algorithms have problems: the former cannot run on low-performance device hardware, while the latter cannot take advantage of high-performance device hardware.
  • this application provides an algorithm configuration method that obtains the processing capability information of the target device, and selects the processing capability information from a plurality of preset identification algorithm packages according to the processing capability information.
  • the matching target recognition algorithm package realizes the adaptive configuration of target recognition algorithm, so that it can make full use of hardware resources when running on high-performance hardware, and running on low-performance hardware can achieve the best compatibility and solve target recognition
  • algorithms are applied to terminal equipment, there are problems that they cannot run on low-performance equipment hardware, or cannot take advantage of high-performance equipment hardware.
  • FIG. 1 is a schematic flowchart of an algorithm configuration method provided by an embodiment of the present application.
  • the execution subject of the embodiment of the present application may be a control device. As shown in Figure 1, the method may include:
  • S101 Acquire processing capability information of the local device.
  • the acquiring processing capability information of the local device includes:
  • the hardware identification information may include hardware model, name, number and other information capable of identifying the hardware identity.
  • the acquiring hardware identification information of the local device includes:
  • API is some predefined functions, the purpose is to provide applications and developers with the ability to access a set of routines based on certain software or hardware without having to access the source code or understand the details of the internal working mechanism.
  • the control device can obtain the hardware identification information of the local device through the API interface, or send a hardware capability acquisition request to the back-end server, where the back-end server can pre-store the corresponding relationship between the device and its hardware identification information, and the server is receiving the hardware After the capability acquisition request, the hardware identification information of the local device is returned according to the hardware capability acquisition request.
  • the processing capability information of the local device may also be saved, or the processing capability information may be displayed, which is convenient for relevant personnel to review and view, and is suitable for practical applications.
  • S102 According to the processing capability information, select a target recognition algorithm package matching the processing capability information from a plurality of preset recognition algorithm packages; wherein the target recognition algorithm package is used to control the The local device recognizes the preset target.
  • the method further includes:
  • the preset target is recognized, and the recognition result is generated.
  • the method further includes:
  • the local device is controlled to follow the target.
  • Face recognition is a method that takes various sensors as input (such as a camera), automatically recognizes the face in the sensor’s field of view, and obtains its position in the screen. The size of the technology.
  • the control device Based on the above-mentioned target recognition algorithm, the control device recognizes the preset target and generates a recognition result. Further, according to the recognition result, the local device is controlled to follow the target.
  • the following technology is a technology that uses various sensors as input (such as a camera), automatically locks on a specified object in the sensor's field of view, and then continues to lock and follow it.
  • the selection of a target recognition algorithm package matching the processing capability information from a plurality of preset recognition algorithm packages according to the processing capability information includes:
  • the hardware identification information includes a Convolutional Neural Networks (CNN) accelerator identification
  • CNN Convolutional Neural Networks
  • the control device may first detect whether the hardware identification information includes a CNN accelerator identification, and if so, select a preset first identification algorithm as the target identification algorithm package.
  • the control device presets the hardware identification priority, for example, CNN accelerator identification>Graphics Processing Unit (GPU) identification>Central Processing Unit (CPU) identification, etc.
  • the determined identifications are sorted according to the above preset hardware identification priority, and the corresponding identification algorithm is selected as the target identification algorithm package according to the sorted priority For example, if the CNN accelerator logo is ranked first, the preset first recognition algorithm is selected as the target recognition algorithm package.
  • the CNN accelerator identifier may be the CNN accelerator model, name, number and other information.
  • the multiple preset identification algorithm packages may include the corresponding relationship between the hardware identification information and the identification algorithm. If the hardware identification information includes a CNN accelerator identification, the control device may select a preset first identification algorithm as the target identification algorithm package according to the foregoing correspondence relationship.
  • the calculation amount of the first recognition algorithm is within a range of 100 GFLOPS to 1000 GFLOPS
  • the first recognition algorithm is a recognition algorithm based on a convolutional neural network.
  • FLOPS ie, "floating-point operations per second", “peak speed per second”
  • floating-point operations per second floating-point operations per second
  • the calculation amount and other parameters of the first recognition algorithm can also be set according to actual conditions.
  • Convolutional neural network is a type of feedforward neural network (Feedforward Neural Networks) that includes convolution calculation and has a deep structure.
  • Convolutional neural network has the ability of representation learning, and can perform shift-invariant classification of input information according to its hierarchical structure, so it is also called shift-invariant artificial neural network (Shift-Invariant Artificial Neural). Networks, SIANN for short).
  • the first recognition algorithm includes one or more recognition sub-algorithms
  • selecting a preset first identification algorithm as the target identification algorithm package includes:
  • the hardware identification information includes a CNN accelerator identifier, obtain the performance parameters of the CNN accelerator in the local device according to the CNN accelerator identifier;
  • the target recognition sub-algorithm corresponding to the performance parameters of the CNN accelerator in the local device, and select the target recognition sub-algorithm as the Target recognition algorithm package.
  • the above-mentioned first recognition algorithm includes one or more recognition sub-algorithms, and the calculation amount of each recognition sub-algorithm is in the range of 100 GFLOPS to 1000 GFLOPS, and they are all recognition algorithms based on convolutional neural networks.
  • the control device may preset the corresponding relationship between the CNN accelerator performance parameters and the recognition sub-algorithm in the first recognition algorithm.
  • the specific setting rules may be determined according to actual conditions. For example, the better the CNN accelerator performance, the higher the calculation amount of the recognition algorithm.
  • the hardware identification information includes the CNN accelerator identifier
  • the performance parameters of the CNN accelerator in the target device are acquired, the target recognition sub-algorithm corresponding to the performance parameters of the CNN accelerator in the local device is determined according to the above-mentioned correspondence, and the target identifier is selected
  • the algorithm is used as a target recognition algorithm package.
  • the selection of a target recognition algorithm package matching the processing capability information from a plurality of preset recognition algorithm packages according to the processing capability information includes:
  • the hardware identification information does not include a CNN accelerator identifier, and the hardware identification information includes a GPU identifier, then a preset second identification algorithm is selected as the target identification algorithm package.
  • the processing capability information of the local device is obtained as described above, if it is detected that the hardware identification information does not include the CNN accelerator identification, then it is further detected whether the hardware identification information contains the GPU identification, and if so, the preset second The recognition algorithm is used as the target recognition algorithm package. Or, after determining the identifiers contained in the hardware identification information, and sorting the determined identifiers according to the preset hardware identifier priority, the GPU identifier is ranked first, and then the preset second identifier is selected The algorithm is used as the target recognition algorithm package.
  • the GPU identifier may be information such as GPU model, name, number, and so on.
  • GPU is called graphics processor, also known as display core, visual processor, display chip. It is a kind of microcomputer that specializes in image calculation on personal computers, workstations, game consoles and some mobile devices (such as tablet computers, smart phones, etc.). processor.
  • the calculation amount of the second recognition algorithm is in the range of 10 GFLOPS to 100 GFLOPS, and the second recognition algorithm is a recognition algorithm based on a convolutional neural network.
  • the second recognition algorithm includes one or more recognition sub-algorithms
  • selecting a preset second identification algorithm as the target identification algorithm package includes:
  • the hardware identification information does not include a CNN accelerator identification, and the hardware identification information contains a GPU identification, acquiring the performance parameters of the GPU in the local device according to the GPU identification;
  • the target recognition sub-algorithm corresponding to the performance parameters of the GPU in the local device, and select the target recognition sub-algorithm as the target recognition Algorithm package.
  • the selection of a target recognition algorithm package matching the processing capability information from a plurality of preset recognition algorithm packages according to the processing capability information includes:
  • the hardware identification information does not contain the CNN accelerator identification and the GPU identification, and the hardware identification information contains the CPU identification, then a preset third identification algorithm is selected as the target identification algorithm package.
  • the processing capability information of the local device is obtained above, if it is detected that the hardware identification information does not include the CNN accelerator identification and the GPU identification, then it is further detected whether the hardware identification information contains the CPU identification, and if so, select the preset
  • the third recognition algorithm is used as the target recognition algorithm package. Or, after determining the identifiers contained in the hardware identification information, and sorting the determined identifiers according to the preset hardware identifier priority, the CPU identifier is ranked first, and then the preset third identifier is selected
  • the algorithm is used as the target recognition algorithm package.
  • the CPU identifier may be information such as the CPU model, name, and serial number.
  • the CPU is called the central processing unit, which is a very large-scale integrated circuit, and is the core and control unit of a computer. Its function is mainly to interpret computer instructions and process data in computer software.
  • the calculation amount of the third recognition algorithm is in the range of 1 GFLOPS to 10 GFLOPS, and the third recognition algorithm is a recognition algorithm based on a convolutional neural network.
  • the third recognition algorithm includes one or more recognition sub-algorithms
  • selecting a preset third identification algorithm as the target identification algorithm package includes:
  • the hardware identification information does not contain a CNN accelerator identification and a GPU identification, and the hardware identification information contains a CPU identification, then obtain the performance parameters of the CPU in the local device according to the CPU identification;
  • the target recognition sub-algorithm corresponding to the performance parameters of the CPU in the local device, and select the target recognition sub-algorithm as the target recognition Algorithm package.
  • the method further includes:
  • the step of selecting the preset third recognition algorithm as the target recognition algorithm package is executed.
  • the aforementioned preset threshold may be set according to actual conditions, for example, the main frequency is greater than 2.0 GHz.
  • the selection of a target recognition algorithm package matching the processing capability information from a plurality of preset recognition algorithm packages according to the processing capability information includes:
  • a preset fourth identification algorithm is selected as the target identification algorithm package.
  • a preset fourth identification algorithm is selected as the target identification algorithm package .
  • the first ranked ones are not the CNN accelerator identifier, GPU identifier, and CPU identifier, Then the preset fourth recognition algorithm is selected as the target recognition algorithm package.
  • the fourth recognition algorithm is a recognition algorithm based on a correlation filter.
  • the calculation amount of the fourth recognition algorithm is lower than the calculation amount of the third recognition algorithm.
  • the correlation filter obtains the corresponding result by correlating the image to be detected with the filter, and then judges and locates according to the obtained filter output.
  • the fourth recognition algorithm includes one or more recognition sub-algorithms
  • selecting a preset fourth identification algorithm as the target identification algorithm package includes:
  • the hardware identification information does not include CNN accelerator identification, GPU identification, and CPU identification, acquiring the remaining performance parameters in the local device;
  • the target recognition sub-algorithm corresponding to the remaining performance parameters in the local device, and select the target recognition sub-algorithm as the target recognition Algorithm package, where the remaining performance parameters can be other performance parameters in the local device except CNN accelerator performance parameters, GPU performance parameters, and CPU performance parameters.
  • the selection of a target recognition algorithm package matching the processing capability information from a plurality of preset recognition algorithm packages according to the processing capability information includes :
  • the hardware identification information is a CNN accelerator identification, select a preset first identification algorithm as the target identification algorithm package;
  • the hardware identification information is a GPU identification, select a preset second identification algorithm as the target identification algorithm package;
  • the hardware identification information is a CPU identification, select a preset third identification algorithm as the target identification algorithm package;
  • the hardware identification information is not a CNN accelerator identification, a GPU identification, and a CPU identification, then a preset fourth identification algorithm is selected as the target identification algorithm package.
  • the algorithm configuration method provided in this embodiment obtains the processing capability information of the local device, and according to the processing capability information, selects a target recognition algorithm package matching the processing capability information from a plurality of preset recognition algorithm packages, thereby achieving
  • the adaptive configuration of the target recognition algorithm makes it possible to make full use of hardware resources when running on high-performance hardware. It can achieve the best compatibility when running on low-performance hardware. It solves the problem of failure when the target recognition algorithm is applied to terminal equipment. Run on low-performance equipment hardware, or fail to take advantage of high-performance equipment hardware.
  • FIG. 2 is a schematic flowchart of another algorithm configuration method provided by an embodiment of the application. Based on the embodiment in FIG. 1, this embodiment describes in detail the specific implementation process of this embodiment. As shown in Figure 2, the method includes:
  • the hardware identification information of the local device can be acquired under preset conditions, for example, the hardware identification information of the local device can be acquired within a preset time period, where the aforementioned preset time period can be set according to actual needs. In other time periods, the hardware identification information of the local device is not obtained, that is, no subsequent algorithm configuration is performed to meet the needs of various application scenarios.
  • the hardware identification information includes a CNN accelerator identifier
  • the aforementioned preset requirements can be set according to actual conditions.
  • the hardware identification information does not contain the CNN accelerator identification, but the hardware identification information contains the GPU identification, it is determined whether the performance parameters of the GPU corresponding to the GPU identification meet the preset requirements, and if so, execute the above The step of selecting a preset second recognition algorithm as the target recognition algorithm package.
  • the calculation amount of the first recognition algorithm is higher than the calculation amount of the second recognition algorithm
  • the calculation amount of the second recognition algorithm is higher than the calculation amount of the third recognition algorithm
  • the third recognition algorithm is higher than the calculation amount of the fourth recognition algorithm.
  • the first recognition algorithm, the second recognition algorithm, and the third recognition algorithm are all recognition algorithms based on convolutional neural networks
  • the fourth The recognition algorithm is a recognition algorithm based on correlation filters.
  • the convolutional neural network method is implemented with high accuracy and a large amount of calculation
  • the correlation filter method is implemented with low accuracy and a small amount of calculation.
  • S206 Recognize a preset target based on the target recognition algorithm package, and generate a recognition result.
  • the recognition of the preset target is face recognition
  • the automatic follow technology is used to follow the target.
  • face recognition is a kind of automatic detection with various sensors as input (such as a camera)
  • sensors such as a camera
  • Auto-following technology It is a technology that takes various sensors as input (such as a camera), automatically locks on a specified object in the sensor's field of view, and then continues to lock and follow it.
  • object recognition technologies can also be used to recognize preset targets, such as vehicle recognition, recognition of certain animals, etc., to meet the needs of different scenarios.
  • target recognition algorithm that can use the algorithm configuration method of this application
  • similar object detection algorithms can also use the method mentioned in this application for adaptive hardware matching.
  • the algorithm configuration method provided in this embodiment can adaptively configure the algorithm so that it can run high-precision algorithms on high-performance hardware to make full use of hardware resources, and run low-precision algorithms on low-performance hardware to achieve the best compatibility. , Make full use of hardware performance to achieve the best algorithm efficiency.
  • FIG. 3 is a schematic flowchart of yet another algorithm configuration method provided by an embodiment of this application.
  • the execution subject of this embodiment of this application may be the control device and the target device in the embodiment shown in FIG. 4.
  • FIG. 4 is a schematic diagram of the architecture of the algorithm configuration system, including a control device 401 and a target device 402.
  • the control device 401 may obtain the processing capability information of the target device 402, and may select a target recognition algorithm package matching the processing capability information from a plurality of preset recognition algorithm packages according to the processing capability information; wherein, the The target recognition algorithm package is used to control the target device 402 to recognize a preset target during execution.
  • the above-mentioned target device may be a movable platform, including an aircraft, a pan/tilt, etc.
  • the processing capability information of the target device is information that can identify the hardware processing capability of the target device, for example, the hardware identification information of the target device.
  • the control device may pre-store the corresponding relationship between the processing capability information of the device and the recognition algorithm, and determine the target recognition algorithm package corresponding to the processing capability information of the target device according to the corresponding relationship.
  • the target device may compare the preset target according to the target recognition algorithm package.
  • the preset target can be any one or more people or things that need to be identified.
  • the target recognition algorithm in the above-mentioned preset multiple recognition algorithm packages can be set according to the actual situation.
  • the target recognition algorithm can be: does not rely on prior knowledge, directly detects the target from the image sequence, and performs Target recognition, finally tracking the target of interest; or, relying on the prior knowledge of the target, first model the moving target, and then find the matching target in the image sequence in real time.
  • the method may include:
  • S301 The control device obtains the processing capability information of the target device.
  • control device acquiring the processing capability information of the target device includes:
  • the control device obtains the hardware identification information of the target device.
  • the control device selects a target recognition algorithm package matching the processing capability information from a plurality of preset recognition algorithm packages.
  • control device selects a target recognition algorithm package matching the processing capability information from a plurality of preset recognition algorithm packages according to the processing capability information, including:
  • the control device selects a preset first identification algorithm as the target identification algorithm package.
  • control device selects a target recognition algorithm package matching the processing capability information from a plurality of preset recognition algorithm packages according to the processing capability information, including:
  • the control device selects a preset second identification algorithm as the target identification algorithm package.
  • control device selects a target recognition algorithm package matching the processing capability information from a plurality of preset recognition algorithm packages according to the processing capability information, including:
  • the control device selects a preset third identification algorithm as the target identification algorithm package.
  • the method further includes:
  • the control device determines whether the main frequency of the CPU corresponding to the CPU identifier is greater than a preset threshold
  • the step of selecting the preset third recognition algorithm as the target recognition algorithm package is executed.
  • control device selects a target recognition algorithm package matching the processing capability information from a plurality of preset recognition algorithm packages according to the processing capability information, including:
  • the control device selects a preset fourth identification algorithm as the target identification algorithm package.
  • control device selects a target recognition algorithm package matching the processing capability information from a plurality of preset recognition algorithm packages according to the processing capability information, including:
  • the control device selects a preset first identification algorithm as the target identification algorithm package;
  • the control device selects a preset second identification algorithm as the target identification algorithm package;
  • the control device selects a preset third identification algorithm as the target identification algorithm package;
  • the control device selects a preset fourth identification algorithm as the target identification algorithm package.
  • the calculation amount of the first recognition algorithm is in the range of 100 GFLOPS to 1000 GFLOPS
  • the calculation amount of the second recognition algorithm is in the range of 10 GFLOPS to 100 GFLOPS
  • the calculation amount of the third recognition algorithm is in the range of 1 GFLOPS to 10 GFLOPS
  • the first recognition algorithm, the second recognition algorithm, and the third recognition algorithm are all recognition algorithms based on a convolutional neural network
  • the fourth recognition algorithm is a recognition algorithm based on a correlation filter.
  • the first recognition algorithm includes one or more recognition sub-algorithms
  • the control device selects a preset first identification algorithm as the target identification algorithm package, which includes:
  • the control device obtains the performance parameters of the CNN accelerator in the target device according to the CNN accelerator identifier;
  • the target recognition sub-algorithm corresponding to the performance parameters of the CNN accelerator in the target device, and select the target recognition sub-algorithm as the Target recognition algorithm package.
  • S303 The control device sends the target recognition algorithm package to the target device.
  • S304 The target device recognizes a preset target object according to the target recognition algorithm package.
  • the method further includes:
  • the target device generates a recognition result, and follows the target according to the recognition result.
  • control device when the control device is a remote sensing device and the target device is a movable platform, the control device obtains the corresponding recognition result and sends a follow instruction to the movable platform body to make the movable platform
  • the platform body can follow the target according to the follow instruction.
  • the method further includes:
  • the target device sends the recognition result to the control device.
  • the method further includes:
  • the control device follows the target object according to the recognition result.
  • control device is a movable platform body
  • target device is an image recognition device connected to the movable body
  • the movable platform body configures a target recognition algorithm for the image recognition device, and the image recognition device is based on the target
  • the recognition algorithm package recognizes the target object, and sends the recognition result to the movable platform body, and the movable platform body then follows the target object based on the recognition result.
  • the control device obtains the processing capability information of the target device, and according to the processing capability information, selects a target recognition algorithm package matching the processing capability information from a plurality of preset recognition algorithm packages , And send the target recognition algorithm package to the target device, and the target device recognizes the preset target according to the target recognition algorithm package, and realizes the adaptive configuration algorithm so that it can run high-precision algorithms on high-performance hardware To make full use of hardware resources, run low-precision algorithms on low-performance hardware to achieve the best compatibility, and make full use of hardware performance to achieve the best algorithm efficiency.
  • FIG. 5 is a schematic flowchart of yet another algorithm configuration method provided by an embodiment of this application.
  • the execution subject of this embodiment of this application may be a control device. It should be understood that the following related features, functions, and other parts corresponding to the description of FIG. 2 and FIG. 3 are described below for brevity, and repeated descriptions are appropriately omitted.
  • the method may include:
  • S501 Acquire processing capability information of the target device.
  • the acquiring processing capability information of the target device includes:
  • S502 According to the processing capability information, select a target recognition algorithm package matching the processing capability information from a plurality of preset recognition algorithm packages.
  • the selection of a target recognition algorithm package matching the processing capability information from a plurality of preset recognition algorithm packages according to the processing capability information includes:
  • the hardware identification information includes a CNN accelerator identification
  • a preset first identification algorithm is selected as the target identification algorithm package.
  • the selection of a target recognition algorithm package matching the processing capability information from a plurality of preset recognition algorithm packages according to the processing capability information includes:
  • the hardware identification information does not include a CNN accelerator identifier, and the hardware identification information includes a GPU identifier, then a preset second identification algorithm is selected as the target identification algorithm package.
  • the selecting a target recognition algorithm package matching the processing capability information from a plurality of preset recognition algorithm packages according to the processing capability information includes:
  • the hardware identification information does not contain the CNN accelerator identification and the GPU identification, and the hardware identification information contains the CPU identification, then a preset third identification algorithm is selected as the target identification algorithm package.
  • the method further includes:
  • the step of selecting the preset third recognition algorithm as the target recognition algorithm package is executed.
  • the selection of a target recognition algorithm package matching the processing capability information from a plurality of preset recognition algorithm packages according to the processing capability information includes:
  • a preset fourth identification algorithm is selected as the target identification algorithm package.
  • the selection of a target recognition algorithm package matching the processing capability information from a plurality of preset recognition algorithm packages according to the processing capability information includes:
  • the hardware identification information is a CNN accelerator identification, select a preset first identification algorithm as the target identification algorithm package;
  • the hardware identification information is a GPU identification, select a preset second identification algorithm as the target identification algorithm package;
  • the hardware identification information is a CPU identification, select a preset third identification algorithm as the target identification algorithm package;
  • the hardware identification information is not a CNN accelerator identification, a GPU identification, and a CPU identification, then a preset fourth identification algorithm is selected as the target identification algorithm package.
  • the calculation amount of the first recognition algorithm is in the range of 100 GFLOPS to 1000 GFLOPS
  • the calculation amount of the second recognition algorithm is in the range of 10 GFLOPS to 100 GFLOPS
  • the calculation amount of the third recognition algorithm is in the range of 1 GFLOPS to 10 GFLOPS
  • the first recognition algorithm, the second recognition algorithm, and the third recognition algorithm are all recognition algorithms based on a convolutional neural network
  • the fourth recognition algorithm is a recognition algorithm based on a correlation filter.
  • the first recognition algorithm includes one or more recognition sub-algorithms
  • selecting a preset first identification algorithm as the target identification algorithm package includes:
  • the hardware identification information includes a CNN accelerator identifier, obtain the performance parameters of the CNN accelerator in the target device according to the CNN accelerator identifier;
  • the target recognition sub-algorithm corresponding to the performance parameters of the CNN accelerator in the target device, and select the target recognition sub-algorithm as the Target recognition algorithm package.
  • S503 Send the target recognition algorithm package to a target device, where the target recognition algorithm is used to instruct the target device to recognize a preset target according to the target recognition algorithm package.
  • the method further includes:
  • the method further includes:
  • the target is followed.
  • the control device obtains the processing capability information of the target device, and according to the processing capability information, selects a target recognition algorithm package matching the processing capability information from a plurality of preset recognition algorithm packages , And send the target recognition algorithm package to the target device, and the target recognition algorithm is used to instruct the target device to recognize the preset target according to the target recognition algorithm package, so as to realize adaptively configuring the algorithm so that It runs high-precision algorithms on high-performance hardware to make full use of hardware resources, runs low-precision algorithms on low-performance hardware to achieve the best compatibility, and makes full use of hardware performance to achieve the best algorithm efficiency.
  • FIG. 6 is a schematic flowchart of another algorithm configuration method provided by an embodiment of the present application.
  • the execution subject of the embodiment of the present application may be a target device. It should be understood that the following related features, functions, and other parts corresponding to the description of FIG. 2 and FIG. 3 are described below for brevity, and repeated descriptions are appropriately omitted.
  • the method may include:
  • S601 Receive a target recognition algorithm package sent by the control device, where the target recognition algorithm package is selected from a plurality of preset recognition algorithm packages according to the processing capability information of the target device, and the target recognition algorithm and the target device Match the processing power information.
  • S602 Recognize a preset target according to the target recognition algorithm package.
  • the method further includes:
  • a recognition result is generated, and the target is followed according to the recognition result.
  • the method further includes:
  • the target device receives a target recognition algorithm package sent by the control device, and the target recognition algorithm package is selected from a plurality of preset recognition algorithm packages according to the processing capability information of the target device, so The target recognition algorithm is matched with the processing capability information of the target device, and the preset target is recognized according to the target recognition algorithm package, so as to realize the adaptive configuration algorithm, so that it can run high-precision algorithms on high-performance hardware To make full use of hardware resources, run low-precision algorithms on low-performance hardware to achieve the best compatibility, and make full use of hardware performance to achieve the best algorithm efficiency.
  • FIG. 7 is a schematic diagram of the hardware structure of an algorithm configuration device provided by an embodiment of the application.
  • the algorithm configuration device 70 of this embodiment includes: a first processor 701 and a first memory 702;
  • the memory 702 is used to store computer execution instructions
  • the processor 701 is configured to execute computer-executable instructions stored in the memory to implement the steps performed by the algorithm configuration method described in FIG. 1 and FIG. 2 in the foregoing embodiment. For details, refer to the related description in the foregoing method embodiment.
  • the memory 702 may be independent or integrated with the processor 701.
  • the algorithm configuration device further includes a bus 703 for connecting the memory 702 and the processor 701.
  • the device provided in the embodiment of the present application can be used to implement the technical solutions of the method embodiments in FIG. 1 and FIG. 2 above, and its implementation principles and technical effects are similar, and the embodiments of the present application will not be repeated here.
  • FIG. 8 is a schematic diagram of the hardware structure of a control device provided by an embodiment of the application.
  • the control device 80 of this embodiment includes: a second processor 801 and a second memory 802;
  • the memory 802 is used to store computer execution instructions
  • the processor 801 is configured to execute computer-executable instructions stored in the memory to implement each step executed by the algorithm configuration method described in FIG. 5 in the foregoing embodiment. For details, refer to the related description in the foregoing method embodiment.
  • the memory 802 may be independent or integrated with the processor 801.
  • the algorithm configuration device When the memory 802 is set independently, the algorithm configuration device also includes a bus 803 for connecting the memory 802 and the processor 801.
  • the device provided in the embodiment of the present application can be used to implement the technical solution of the method embodiment in FIG. 5, and its implementation principles and technical effects are similar, and the details of the embodiment of the present application are not repeated here.
  • FIG. 9 is a schematic diagram of the hardware structure of a target device provided by an embodiment of the application.
  • the target device 90 of this embodiment includes: a third processor 901 and a third memory 902;
  • the memory 902 is used to store computer execution instructions
  • the processor 901 is configured to execute computer-executable instructions stored in the memory to implement each step executed by the algorithm configuration method described in FIG. 6 in the foregoing embodiment. For details, refer to the related description in the foregoing method embodiment.
  • the memory 902 may be independent or integrated with the processor 901.
  • the algorithm configuration device further includes a bus 903 for connecting the memory 902 and the processor 901.
  • the device provided in the embodiment of the present application can be used to implement the technical solution of the method embodiment in FIG. 6 above, and its implementation principles and technical effects are similar, and the details of the embodiment of the present application are not repeated here.
  • FIG. 10 is a schematic structural diagram of an algorithm configuration system provided by an embodiment of this application.
  • the algorithm configuration system 100 of this embodiment includes: a control device 1001 and a target device 1002; among them,
  • the control device 1001 is configured to obtain processing capability information of a target device; and according to the processing capability information, select a target recognition algorithm package matching the processing capability information from a plurality of preset recognition algorithm packages; Sending the target recognition algorithm package to the target device;
  • the target device 1002 is configured to recognize a preset target object according to the target recognition algorithm package.
  • the target device 1002 is also used to generate a recognition result, and follow the target according to the recognition result.
  • the target device 1002 is further configured to send the identification result to the control device.
  • control device 1001 is further configured to follow the target according to the recognition result.
  • control device 1001 is also used to obtain hardware identification information of the target device.
  • control device 1001 is further configured to:
  • the hardware identification information includes a CNN accelerator identification
  • a preset first identification algorithm is selected as the target identification algorithm package.
  • control device 1001 is further configured to:
  • the hardware identification information does not include a CNN accelerator identifier, and the hardware identification information includes a GPU identifier, then a preset second identification algorithm is selected as the target identification algorithm package.
  • control device 1001 is further configured to:
  • the hardware identification information does not contain the CNN accelerator identification and the GPU identification, and the hardware identification information contains the CPU identification, then a preset third identification algorithm is selected as the target identification algorithm package.
  • control device 1001 is further configured to:
  • a preset fourth identification algorithm is selected as the target identification algorithm package.
  • control device 1001 is further configured to:
  • the step of selecting the preset third recognition algorithm as the target recognition algorithm package is executed.
  • control device 1001 is further configured to:
  • the hardware identification information is a CNN accelerator identification, select a preset first identification algorithm as the target identification algorithm package;
  • the hardware identification information is a GPU identification, select a preset second identification algorithm as the target identification algorithm package;
  • the hardware identification information is a CPU identification, select a preset third identification algorithm as the target identification algorithm package;
  • the hardware identification information is not a CNN accelerator identification, a GPU identification, and a CPU identification, then a preset fourth identification algorithm is selected as the target identification algorithm package.
  • the calculation amount of the first recognition algorithm is in the range of 100 GFLOPS to 1000 GFLOPS
  • the calculation amount of the second recognition algorithm is in the range of 10 GFLOPS to 100 GFLOPS
  • the calculation amount of the third recognition algorithm is in the range of 1 GFLOPS to 10 GFLOPS
  • the first recognition algorithm, the second recognition algorithm, and the third recognition algorithm are all recognition algorithms based on a convolutional neural network
  • the fourth recognition algorithm is a recognition algorithm based on a correlation filter.
  • the first recognition algorithm includes one or more recognition sub-algorithms
  • the control device 1001 is also used for:
  • the hardware identification information includes a CNN accelerator identifier, obtain the performance parameters of the CNN accelerator in the target device according to the CNN accelerator identifier;
  • the target recognition sub-algorithm corresponding to the performance parameters of the CNN accelerator in the target device, and select the target recognition sub-algorithm as the Target recognition algorithm package.
  • FIG. 11 is a schematic structural diagram of a movable platform provided by an embodiment of the application. As shown in FIG. 11, the movable platform 110 of this embodiment includes:
  • the power system 1102 is provided in the body 1101, and the power system 1102 is used to provide power for the movable platform; and the algorithm configuration device 70 as described above in FIG. 7.
  • the device provided by the embodiment of the present application includes the algorithm configuration device 70 described in FIG. 7 above, and its implementation principle and technical effect are as described above, and the embodiments of the present application will not be repeated here.
  • FIG. 12 is a schematic structural diagram of another movable platform provided by an embodiment of the application. As shown in FIG. 12, the movable platform 120 of this embodiment includes:
  • the power system 1202 is provided in the body 1201, and the power system 1202 is used to provide power for the movable platform;
  • control device 80 as described in FIG. 8.
  • the devices provided in the embodiments of the present application include the control device 80 described in FIG. 8, and the implementation principles and technical effects thereof are as described above, and the embodiments of the present application will not be repeated here.
  • FIG. 13 is a schematic structural diagram of still another movable platform provided by an embodiment of the application. As shown in FIG. 13, the movable platform 130 of this embodiment includes:
  • the power system 1302 is provided in the body 1301, and the power system 1302 is used to provide power for the movable platform;
  • the devices provided in the embodiments of the present application include the control device 90 described in FIG. 9 above, and the implementation principles and technical effects thereof are as described above, and the embodiments of the present application will not be repeated here.
  • Fig. 14 is a schematic structural diagram of yet another movable platform provided by an embodiment of the application. As shown in FIG. 14, the movable platform 140 of this embodiment includes:
  • the power system 1402 is provided in the body 1401, and the power system 1402 is used to provide power for the movable platform;
  • the algorithm configuration system 100 as shown in Figure 10 is located in the body.
  • the target device and the control device are both located on the fuselage, the target device can be used to identify, and the control device can be used to select algorithms and control the movement of the movable platform.
  • the control device can be used to select the algorithm, and the target device can be used to identify and control the movement of the movable platform.
  • FIG. 15 is a schematic structural diagram of another movable platform provided by an embodiment of the application.
  • the movable platform 150 of this embodiment includes: a movable platform body 1501 and a control device 1502; the movable platform body 1501 and the control device 1502 are connected wirelessly or wiredly;
  • the control device 1502 is configured to obtain the processing capability information of the movable platform body 1501; according to the processing capability information, select a target recognition algorithm package matching the processing capability information from a plurality of preset recognition algorithm packages; Send the target recognition algorithm package to the movable platform body 1501;
  • the movable platform body 1501 is used to recognize a preset target according to the target recognition algorithm package.
  • the movable platform body 1501 generates a recognition result, and follows the target according to the recognition result.
  • the movable platform body 1501 sends the recognition result to the control device 1502.
  • control device 1502 sends a follow instruction to the movable platform body 1501 according to the recognition result, so that the movable platform body 1501 can follow the target according to the follow instruction.
  • control device 1502 acquiring processing capability information of the movable platform body includes:
  • the control device obtains the hardware identification information of the movable platform body.
  • the selection of a target recognition algorithm package matching the processing capability information from a plurality of preset recognition algorithm packages according to the processing capability information includes:
  • the control device selects a preset first identification algorithm as the target identification algorithm package.
  • the selection of a target recognition algorithm package matching the processing capability information from a plurality of preset recognition algorithm packages according to the processing capability information includes:
  • the control device selects a preset second identification algorithm as the target identification algorithm package.
  • the selection of a target recognition algorithm package matching the processing capability information from a plurality of preset recognition algorithm packages according to the processing capability information includes:
  • the control device selects a preset third identification algorithm as the target identification algorithm package.
  • the selection of a target recognition algorithm package matching the processing capability information from a plurality of preset recognition algorithm packages according to the processing capability information includes:
  • the control device selects a preset fourth identification algorithm as the target identification algorithm package.
  • control device determines whether the main frequency of the CPU corresponding to the CPU identifier is greater than a preset threshold
  • the step of selecting the preset third recognition algorithm as the target recognition algorithm package is executed.
  • the selecting a target recognition algorithm package matching the processing capability information from a plurality of preset recognition algorithm packages according to the processing capability information includes:
  • the control device selects a preset first identification algorithm as the target identification algorithm package;
  • the control device selects a preset second identification algorithm as the target identification algorithm package;
  • the control device selects a preset third identification algorithm as the target identification algorithm package;
  • the control device selects a preset fourth identification algorithm as the target identification algorithm package.
  • the calculation amount of the first recognition algorithm is in the range of 100 GFLOPS to 1000 GFLOPS
  • the calculation amount of the second recognition algorithm is in the range of 10 GFLOPS to 100 GFLOPS
  • the calculation amount of the third recognition algorithm is in the range of 1 GFLOPS to 10 GFLOPS
  • the first recognition algorithm, the second recognition algorithm, and the third recognition algorithm are all recognition algorithms based on a convolutional neural network
  • the fourth recognition algorithm is a recognition algorithm based on a correlation filter.
  • the first recognition algorithm includes one or more recognition sub-algorithms
  • the control device selects a preset first identification algorithm as the target identification algorithm package, which includes:
  • the control device obtains the performance parameters of the CNN accelerator in the target device according to the CNN accelerator identifier;
  • the target recognition sub-algorithm corresponding to the performance parameters of the CNN accelerator in the target device, and select the target recognition sub-algorithm as the Target recognition algorithm package.
  • the control device obtains the processing capability information of the movable platform body, and according to the processing capability information, selects target recognition matching the processing capability information from a plurality of preset recognition algorithm packages Algorithm package, and send the target recognition algorithm package to the mobile platform body, and the mobile platform body recognizes the preset target according to the target recognition algorithm package, and realizes the adaptive configuration algorithm to make it in high performance Run high-precision algorithms on hardware to make full use of hardware resources, run low-precision algorithms on low-performance hardware to achieve the best compatibility, and make full use of hardware performance to achieve the best algorithm efficiency.
  • the embodiments of the present application also provide a computer-readable storage medium, which stores computer-executable instructions, and when the processor executes the computer-executable instructions, the algorithm described in Figure 1 and Figure 2 above is implemented. Configuration method.
  • the embodiment of the present application also provides another computer-readable storage medium, and the computer-readable storage medium stores computer-executable instructions.
  • the processor executes the computer-executable instructions, the algorithm configuration method described in FIG. .
  • the embodiments of the present application also provide yet another computer-readable storage medium, the computer-readable storage medium stores computer-executable instructions, and when the processor executes the computer-executed instructions, the algorithm configuration method described in FIG. .
  • the disclosed device and method may be implemented in other ways.
  • the device embodiments described above are only illustrative.
  • the division of the modules is only a logical function division, and there may be other divisions in actual implementation, for example, multiple modules can be combined or integrated. To another system, or some features can be ignored or not implemented.
  • the displayed or discussed mutual coupling or direct coupling or communication connection may be indirect coupling or communication connection through some interfaces, devices or modules, and may be in electrical, mechanical or other forms.
  • modules described as separate components may or may not be physically separated, and the components displayed as modules may or may not be physical units, that is, they may be located in one place, or they may be distributed on multiple network units. Some or all of the modules may be selected according to actual needs to achieve the objectives of the solutions of the embodiments.
  • the functional modules in the various embodiments of the present application may be integrated into one processing unit, or each module may exist alone physically, or two or more modules may be integrated into one unit.
  • the units formed by the above-mentioned modules can be realized in the form of hardware, or in the form of hardware plus software functional units.
  • the above-mentioned integrated modules implemented in the form of software function modules may be stored in a computer readable storage medium.
  • the above-mentioned software function module is stored in a storage medium and includes several instructions to make a computer device (which can be a personal computer, a server, or a network device, etc.) or a processor (English: processor) to execute the various embodiments of the present application Part of the method.
  • processor may be a central processing unit (Central Processing Unit, CPU for short), or other general-purpose processors, digital signal processors (Digital Signal Processor, DSP), and application specific integrated circuits (Application Specific Integrated Circuits). Referred to as ASIC) and so on.
  • the general-purpose processor may be a microprocessor or the processor may also be any conventional processor or the like. The steps of the method disclosed in combination with the invention can be directly embodied as executed by a hardware processor, or executed by a combination of hardware and software modules in the processor.
  • the memory may include a high-speed RAM memory, and may also include a non-volatile storage NVM, such as at least one disk storage, and may also be a U disk, a mobile hard disk, a read-only memory, a magnetic disk, or an optical disk.
  • NVM non-volatile storage
  • the bus may be an Industry Standard Architecture (ISA) bus, Peripheral Component (PCI) bus, or Extended Industry Standard Architecture (EISA) bus, etc.
  • ISA Industry Standard Architecture
  • PCI Peripheral Component
  • EISA Extended Industry Standard Architecture
  • the bus can be divided into address bus, data bus, control bus, etc.
  • the buses in the drawings of this application are not limited to only one bus or one type of bus.
  • the above-mentioned storage medium can be realized by any type of volatile or non-volatile storage device or their combination, such as static random access memory (SRAM), electrically erasable programmable read-only memory (EEPROM), erasable Except for programmable read only memory (EPROM), programmable read only memory (PROM), read only memory (ROM), magnetic memory, flash memory, magnetic disks or optical disks.
  • SRAM static random access memory
  • EEPROM electrically erasable programmable read-only memory
  • EPROM erasable except for programmable read only memory
  • PROM programmable read only memory
  • ROM read only memory
  • magnetic memory flash memory
  • flash memory magnetic disks or optical disks.
  • optical disks any available medium that can be accessed by a general-purpose or special-purpose computer.
  • An exemplary storage medium is coupled to the processor, so that the processor can read information from the storage medium and can write information to the storage medium.
  • the storage medium may also be an integral part of the processor.
  • the processor and the storage medium may be located in Application Specific Integrated Circuits (ASIC for short).
  • ASIC Application Specific Integrated Circuits
  • the processor and the storage medium may also exist as discrete components in the electronic device or the main control device.
  • a person of ordinary skill in the art can understand that all or part of the steps in the foregoing method embodiments can be implemented by a program instructing relevant hardware.
  • the aforementioned program can be stored in a computer readable storage medium.
  • the steps including the foregoing method embodiments are executed; and the foregoing storage medium includes: ROM, RAM, magnetic disk, or optical disk and other media that can store program codes.

Landscapes

  • Engineering & Computer Science (AREA)
  • Aviation & Aerospace Engineering (AREA)
  • Radar, Positioning & Navigation (AREA)
  • Remote Sensing (AREA)
  • Physics & Mathematics (AREA)
  • General Physics & Mathematics (AREA)
  • Automation & Control Theory (AREA)
  • Image Analysis (AREA)
  • Data Exchanges In Wide-Area Networks (AREA)

Abstract

L'invention concerne un procédé de configuration d'algorithme, un dispositif et une plateforme mobile. Le procédé comprend : l'obtention d'informations de capacité de traitement d'un dispositif local (S101) ; et la sélection, parmi une pluralité de progiciels d'algorithme de reconnaissance prédéfinis, en fonction des informations de capacité de traitement, d'un progiciel d'algorithme de reconnaissance de cible correspondant aux informations de capacité de traitement, le progiciel d'algorithme de reconnaissance de cible étant utilisé pour, lorsqu'il est exécuté, commander le dispositif local pour reconnaître un objet cible prédéfini (S102). Le procédé est utilisé pour configurer de manière adaptative un algorithme de reconnaissance de cible, de telle sorte que l'algorithme de reconnaissance de cible puisse utiliser complètement des ressources matérielles lors de son exécution sur du matériel à hautes performances, et permette d'obtenir une bonne compatibilité lorsqu'il est exécuté sur du matériel à faibles performances, ce qui permet de résoudre le problème qui est que l'algorithme de reconnaissance de cible, lorsqu'il est appliqué à un dispositif terminal, ne peut pas être exécuté sur du matériel à faibles performances du dispositif ou ne peut pas exploiter pleinement du matériel à hautes performances du dispositif.
PCT/CN2019/095391 2019-07-10 2019-07-10 Procédé de configuration d'algorithme, dispositif, système et plateforme mobile WO2021003692A1 (fr)

Priority Applications (2)

Application Number Priority Date Filing Date Title
PCT/CN2019/095391 WO2021003692A1 (fr) 2019-07-10 2019-07-10 Procédé de configuration d'algorithme, dispositif, système et plateforme mobile
CN201980033933.1A CN112400147A (zh) 2019-07-10 2019-07-10 算法配置方法、设备、***及可移动平台

Applications Claiming Priority (1)

Application Number Priority Date Filing Date Title
PCT/CN2019/095391 WO2021003692A1 (fr) 2019-07-10 2019-07-10 Procédé de configuration d'algorithme, dispositif, système et plateforme mobile

Publications (1)

Publication Number Publication Date
WO2021003692A1 true WO2021003692A1 (fr) 2021-01-14

Family

ID=74114298

Family Applications (1)

Application Number Title Priority Date Filing Date
PCT/CN2019/095391 WO2021003692A1 (fr) 2019-07-10 2019-07-10 Procédé de configuration d'algorithme, dispositif, système et plateforme mobile

Country Status (2)

Country Link
CN (1) CN112400147A (fr)
WO (1) WO2021003692A1 (fr)

Cited By (2)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN113596158A (zh) * 2021-07-29 2021-11-02 杭州海康威视***技术有限公司 一种基于场景的算法配置方法和装置
CN114924772A (zh) * 2022-07-15 2022-08-19 浙江大华技术股份有限公司 一种算法包的更新方法及设备

Families Citing this family (1)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN113657320A (zh) * 2021-08-23 2021-11-16 杭州海康威视数字技术股份有限公司 一种业务执行方法及装置

Citations (5)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN108292141A (zh) * 2016-03-01 2018-07-17 深圳市大疆创新科技有限公司 用于目标跟踪的方法和***
US20180208309A1 (en) * 2015-07-02 2018-07-26 SZ DJI Technology Co., Ltd. Unmanned aerial vehicle, control system and method thereof, and unmanned aerial vehicle landing control method
CN108596434A (zh) * 2018-03-23 2018-09-28 卫盈联信息技术（深圳）有限公司 欺诈检测和风险评估方法、***、设备及存储介质
CN109657535A (zh) * 2018-10-30 2019-04-19 银河水滴科技（北京）有限公司 图片识别方法、目标设备及云平台
CN109960965A (zh) * 2017-12-14 2019-07-02 翔升（上海）电子技术有限公司 基于无人机识别动物行为的方法、装置和***

Family Cites Families (2)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN108829518B (zh) * 2018-05-31 2020-01-03 北京百度网讯科技有限公司 用于推送信息的方法和装置
CN109471718B (zh) * 2018-10-12 2023-09-22 平安科技（深圳）有限公司 基于人脸识别的计算资源配置方法、装置、设备及介质

Patent Citations (5)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
US20180208309A1 (en) * 2015-07-02 2018-07-26 SZ DJI Technology Co., Ltd. Unmanned aerial vehicle, control system and method thereof, and unmanned aerial vehicle landing control method
CN108292141A (zh) * 2016-03-01 2018-07-17 深圳市大疆创新科技有限公司 用于目标跟踪的方法和***
CN109960965A (zh) * 2017-12-14 2019-07-02 翔升（上海）电子技术有限公司 基于无人机识别动物行为的方法、装置和***
CN108596434A (zh) * 2018-03-23 2018-09-28 卫盈联信息技术（深圳）有限公司 欺诈检测和风险评估方法、***、设备及存储介质
CN109657535A (zh) * 2018-10-30 2019-04-19 银河水滴科技（北京）有限公司 图片识别方法、目标设备及云平台

Cited By (2)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN113596158A (zh) * 2021-07-29 2021-11-02 杭州海康威视***技术有限公司 一种基于场景的算法配置方法和装置
CN114924772A (zh) * 2022-07-15 2022-08-19 浙江大华技术股份有限公司 一种算法包的更新方法及设备

Also Published As

Publication number Publication date
CN112400147A (zh) 2021-02-23

Similar Documents

Publication Publication Date Title
WO2021003692A1 (fr) Procédé de configuration d'algorithme, dispositif, système et plateforme mobile
US10917552B2 (en) Photographing method using external electronic device and electronic device supporting the same
US9733083B2 (en) Motion sensing method and user equipment thereof
US20180034224A1 (en) Electrical device
CN112528831B (zh) 多目标姿态估计方法、多目标姿态估计装置及终端设备
CN103679788B (zh) 一种移动终端中3d图像的生成方法和装置
CN111833447A (zh) 三维地图构建方法、三维地图构建装置及终端设备
CN111462174B (zh) 多目标跟踪方法、装置以及电子设备
US20160364564A1 (en) Lock screen output controlling method and electronic device for supporting the same
US10839555B2 (en) Target detection method and unmanned aerial vehicle
US9355338B2 (en) Image recognition device, image recognition method, and recording medium
CN106959927B (zh) 获取虚拟机中的逻辑分区的位置信息的方法及装置
US20190251394A1 (en) Method and apparatus for continuously displaying images on basis of similarity of images
CN108596079A (zh) 手势识别方法、装置及电子设备
JP5430636B2 (ja) データ取得装置、方法及びプログラム
US20150347853A1 (en) Method for providing service and electronic device thereof
US10706555B2 (en) Image processing method and device
EP3446240B1 (fr) Dispositif électronique et procédé de sortie de vignette correspondant à la saisie de l'utilisateur
US10091436B2 (en) Electronic device for processing image and method for controlling the same
CN110738175A (zh) 人脸图像处理方法、装置、计算机设备和存储介质
WO2018053956A1 (fr) Procédé d'interaction et dispositif portable
CN109933298B (zh) 一种根据用户位置进行显示的方法及装置
KR20180083144A (ko) 마커를 검출하기 위한 방법 및 그 전자 장치
CN113587917A (zh) 室内定位方法、装置、设备、存储介质及计算机程序产品
CN114913567A (zh) 口罩佩戴的检测方法、装置、终端设备及可读存储介质

Legal Events

Date Code Title Description
121 Ep: the epo has been informed by wipo that ep was designated in this application

Ref document number: 19937115

Country of ref document: EP

Kind code of ref document: A1

122 Ep: pct application non-entry in european phase

Ref document number: 19937115

Country of ref document: EP

Kind code of ref document: A1