CN112785466A - AI enabling method and device of hardware, storage medium and equipment - Google Patents
AI enabling method and device of hardware, storage medium and equipment Download PDFInfo
- Publication number
- CN112785466A CN112785466A CN202011641320.7A CN202011641320A CN112785466A CN 112785466 A CN112785466 A CN 112785466A CN 202011641320 A CN202011641320 A CN 202011641320A CN 112785466 A CN112785466 A CN 112785466A
- Authority
- CN
- China
- Prior art keywords
- hardware
- model
- energized
- enabling
- running
- 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
- 238000000034 method Methods 0.000 title claims abstract description 56
- 238000003860 storage Methods 0.000 title claims abstract description 11
- 238000012549 training Methods 0.000 claims description 39
- 238000012795 verification Methods 0.000 claims description 35
- 238000000605 extraction Methods 0.000 claims description 16
- 238000004891 communication Methods 0.000 claims description 12
- 238000013527 convolutional neural network Methods 0.000 claims description 10
- 238000004590 computer program Methods 0.000 claims description 4
- 238000004364 calculation method Methods 0.000 claims description 3
- 238000013473 artificial intelligence Methods 0.000 abstract description 89
- 230000005540 biological transmission Effects 0.000 abstract description 15
- 230000008569 process Effects 0.000 description 12
- 230000009471 action Effects 0.000 description 3
- 238000010586 diagram Methods 0.000 description 3
- 238000005516 engineering process Methods 0.000 description 3
- 230000006870 function Effects 0.000 description 3
- 238000010276 construction Methods 0.000 description 2
- 230000000694 effects Effects 0.000 description 2
- 238000012545 processing Methods 0.000 description 2
- 230000004044 response Effects 0.000 description 2
- 230000006399 behavior Effects 0.000 description 1
- 238000004422 calculation algorithm Methods 0.000 description 1
- 230000008859 change Effects 0.000 description 1
- 239000003086 colorant Substances 0.000 description 1
- 230000007547 defect Effects 0.000 description 1
- 238000011161 development Methods 0.000 description 1
- 238000011478 gradient descent method Methods 0.000 description 1
- 230000003993 interaction Effects 0.000 description 1
- 238000004519 manufacturing process Methods 0.000 description 1
- 238000012986 modification Methods 0.000 description 1
- 230000004048 modification Effects 0.000 description 1
- 238000003062 neural network model Methods 0.000 description 1
- 230000003287 optical effect Effects 0.000 description 1
- 238000005457 optimization Methods 0.000 description 1
- 238000002360 preparation method Methods 0.000 description 1
- 230000000750 progressive effect Effects 0.000 description 1
- 238000012546 transfer Methods 0.000 description 1
- 238000013526 transfer learning Methods 0.000 description 1
Images
Classifications
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06Q—INFORMATION AND COMMUNICATION TECHNOLOGY [ICT] SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES; SYSTEMS OR METHODS SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES, NOT OTHERWISE PROVIDED FOR
- G06Q50/00—Information and communication technology [ICT] specially adapted for implementation of business processes of specific business sectors, e.g. utilities or tourism
- G06Q50/10—Services
- G06Q50/20—Education
- G06Q50/205—Education administration or guidance
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
- G06N3/00—Computing arrangements based on biological models
- G06N3/02—Neural networks
- G06N3/04—Architecture, e.g. interconnection topology
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
- G06N3/00—Computing arrangements based on biological models
- G06N3/02—Neural networks
- G06N3/08—Learning methods
-
- H—ELECTRICITY
- H04—ELECTRIC COMMUNICATION TECHNIQUE
- H04L—TRANSMISSION OF DIGITAL INFORMATION, e.g. TELEGRAPHIC COMMUNICATION
- H04L67/00—Network arrangements or protocols for supporting network services or applications
- H04L67/01—Protocols
- H04L67/02—Protocols based on web technology, e.g. hypertext transfer protocol [HTTP]
- H04L67/025—Protocols based on web technology, e.g. hypertext transfer protocol [HTTP] for remote control or remote monitoring of applications
-
- H—ELECTRICITY
- H04—ELECTRIC COMMUNICATION TECHNIQUE
- H04L—TRANSMISSION OF DIGITAL INFORMATION, e.g. TELEGRAPHIC COMMUNICATION
- H04L67/00—Network arrangements or protocols for supporting network services or applications
- H04L67/14—Session management
Landscapes
- Engineering & Computer Science (AREA)
- Physics & Mathematics (AREA)
- Business, Economics & Management (AREA)
- Theoretical Computer Science (AREA)
- Health & Medical Sciences (AREA)
- General Physics & Mathematics (AREA)
- General Health & Medical Sciences (AREA)
- Computational Linguistics (AREA)
- Life Sciences & Earth Sciences (AREA)
- Evolutionary Computation (AREA)
- Biophysics (AREA)
- Molecular Biology (AREA)
- Computing Systems (AREA)
- General Engineering & Computer Science (AREA)
- Biomedical Technology (AREA)
- Mathematical Physics (AREA)
- Software Systems (AREA)
- Artificial Intelligence (AREA)
- Data Mining & Analysis (AREA)
- Educational Administration (AREA)
- Educational Technology (AREA)
- Strategic Management (AREA)
- Tourism & Hospitality (AREA)
- Signal Processing (AREA)
- Computer Networks & Wireless Communication (AREA)
- Marketing (AREA)
- Primary Health Care (AREA)
- General Business, Economics & Management (AREA)
- Human Resources & Organizations (AREA)
- Economics (AREA)
- Image Analysis (AREA)
Abstract
The application discloses an AI enabling method, a device, a storage medium and equipment of hardware, wherein the method comprises the following steps: after the running image of the hardware to be energized is obtained, the running image is sent to a pre-constructed energizing model to predict the running state of the hardware to be energized at the next moment; and then, controlling the hardware to be energized to automatically run according to the predicted running state at the next moment. Therefore, as the hardware to be energized and the equipment where the energized model is located are located in the same local area network, when the energized model is used for performing customized AI energization on the hardware to be energized, the safety of data transmission can be ensured. In the application scene of artificial intelligence innovation education, teachers and students can use a pre-trained self-defined enabling model to perform self-defined AI enabling on open source hardware in the scene, data transmission safety can be guaranteed, and learning experience of students in AI courses is improved.
Description
Technical Field
The present application relates to the field of artificial intelligence technologies, and in particular, to a hardware AI enabling method, apparatus, storage medium, and device.
Background
With the rapid development of network technology, big data, Artificial Intelligence (AI) enabling, cloud computing, internet of things platform, intelligent manufacturing technology and the like are comprehensively promoted and widely developed.
At present, for an open-source hardware device with low computing capability, because of the limitation of the computing capability of its own processor, the open-source hardware device basically has no AI capability and can only rely on the AI capability of a cloud, that is, the open-source hardware needs to transmit acquired data to the cloud device through a wide area network, and then returns a result to the open-source hardware after the computation of the cloud device by using an existing AI engine is completed. However, for the application scenario of artificial intelligence innovation education with slow network transmission speed, the simple AI capability calling cannot well stimulate the learning interest of students in AI courses, and the human-computer interaction experience is poor. The following four disadvantages exist in particular: firstly, because data are transmitted through a wide area network, a plurality of intermediate links exist, and network delay exists; secondly, a large amount of original data of the AI engine to be trained are transmitted to a cloud server through a wide area network for training, so that the flow consumption is high; thirdly, the data has security risks of being hijacked or modified in the internet transmission process; fourth, because the AI capability of the current open-source hardware is that the cloud end provides reasoning and prediction capabilities after an engineer trains a model in advance, and does not support the prediction capability of a user-defined AI engine, students can only see the calling result of the cloud end AI capability in an artificial intelligence innovation education class, and cannot well participate in and experience the implementation process of the AI capability, and cannot fully understand the knowledge point of the AI, and thus cannot achieve good classroom learning effect.
Therefore, in the application scene of artificial intelligence innovation education, the conventional AI enabling method for open source hardware has data security risk, and can not meet the classroom learning requirements of students, so that the experience of the students is poor.
Disclosure of Invention
The embodiment of the application mainly aims to provide an AI enabling method, device, storage medium and equipment for hardware, which can endow flexible self-defined AI capacity to open-source hardware in a local area network through a pre-constructed enabling model, and further can improve the learning experience and learning interest of students in AI courses.
The embodiment of the application provides an AI enabling method of hardware, which comprises the following steps:
acquiring a running image of hardware to be energized;
sending the running image to a pre-constructed enabling model to predict the running state of the hardware to be enabled at the next moment; the hardware to be energized and the equipment where the energized model is located are located in the same local area network;
and controlling the hardware to be energized to automatically operate according to the operating state of the next moment.
In a possible implementation manner, before sending the running image to a pre-constructed energized model, the method further includes:
establishing communication connection with the equipment where the enabling model is located in the same local area network in advance;
then, the sending the running image to a pre-constructed enabling model to predict the running state of the hardware to be enabled at the next moment comprises:
and sending the running image to a pre-constructed enabling model through the communication connection so as to predict the running state of the hardware to be enabled at the next moment.
In one possible implementation, constructing the enabling model includes:
acquiring a training running image of hardware;
training an initial enabling model according to the training running image of the hardware and the running state label of the hardware at the next moment corresponding to the training running image of the hardware to generate the enabling model; the hardware and the device where the initial enabling model is located are located in the same local area network.
In one possible implementation, the initial enabling model includes a feature extraction network, a fully connected layer, and an output layer.
In one possible implementation, the feature extraction network includes a convolutional neural network Resnet-18 and/or a convolutional neural network VGG 16.
In a possible implementation, the method further includes:
acquiring a verification running image of the hardware;
inputting the verification running image of the hardware into the enabling model to obtain a running state prediction result of the hardware at the next moment;
and when the running state prediction result of the hardware at the next moment is inconsistent with the running state marking result of the hardware at the next moment corresponding to the verification running image of the hardware, taking the verification running image of the hardware as the training running image of the hardware again, and updating the enabling model.
An embodiment of the present application further provides an AI enabling apparatus for hardware, including:
the first acquisition unit is used for acquiring an operation image of hardware to be energized;
the sending unit is used for sending the running image to a pre-constructed enabling model so as to predict the running state of the hardware to be enabled at the next moment; the hardware to be energized and the equipment where the edge calculation AI model is located are located in the same local area network;
and the control unit is used for controlling the hardware to be energized to automatically run according to the running state of the next moment.
In a possible implementation manner, the apparatus further includes:
the establishing unit is used for establishing communication connection with the equipment where the enabling model is located in the same local area network in advance;
the sending unit is specifically configured to:
and sending the running image to a pre-constructed enabling model through the communication connection so as to predict the running state of the hardware to be enabled at the next moment.
In a possible implementation manner, the apparatus further includes:
the second acquisition unit is used for acquiring a training running image of the hardware;
the training unit is used for training an initial energized model according to the training running image of the hardware and the running state label of the hardware at the next moment corresponding to the training running image of the hardware, and generating the energized model; the hardware and the device where the initial enabling model is located are located in the same local area network.
In one possible implementation, the initial enabling model includes a feature extraction network, a fully connected layer, and an output layer.
In one possible implementation, the feature extraction network includes a convolutional neural network Resnet-18 and/or a convolutional neural network VGG 16.
In a possible implementation manner, the apparatus further includes:
the third acquisition unit is used for acquiring a verification running image of the hardware;
the input unit is used for inputting the verification running image of the hardware into the enabling model and obtaining a running state prediction result of the hardware at the next moment;
and the updating unit is used for taking the verification running image of the hardware as the training running image of the hardware again to update the enabling model when the running state prediction result of the hardware at the next moment is inconsistent with the running state marking result of the hardware at the next moment corresponding to the verification running image of the hardware.
An embodiment of the present application further provides an AI enabling apparatus for hardware, including: a processor, a memory, a system bus;
the processor and the memory are connected through the system bus;
the memory is for storing one or more programs, the one or more programs including instructions, which when executed by the processor, cause the processor to perform any one of the implementations of the hardware AI-enabled method described above.
An embodiment of the present application further provides a computer-readable storage medium, where instructions are stored in the computer-readable storage medium, and when the instructions are executed on a terminal device, the terminal device is caused to execute any implementation manner of the hardware AI enabling method.
An embodiment of the present application further provides a computer program product, which, when running on a terminal device, enables the terminal device to execute any implementation manner of the above-mentioned AI enabling method for hardware.
According to the AI enabling method, the device, the storage medium and the equipment of the hardware, firstly, the running image of the hardware to be enabled is obtained, and then the running image is sent to a pre-constructed enabling model so as to predict the running state of the hardware to be enabled at the next moment; the hardware to be energized and the equipment where the energized model is located are located in the same local area network, and then the hardware to be energized can be controlled to automatically run according to the running state of the next moment. Therefore, in the embodiment of the application, the pre-trained enabling model is used for enabling the AI to be customized for the hardware to be enabled so as to control the hardware to be enabled to automatically operate at the next moment according to the operation state predicted by the model, and as the hardware to be enabled and the equipment where the enabling model is located are located in the same local area network, data transmission between the hardware to be enabled and the equipment is only carried out in the local area network, so that the safety of data transmission can be ensured.
Therefore, in the application scene of artificial intelligence innovation education, teachers and students can utilize a pre-trained self-defined enabling model to carry out self-defined AI enabling on open source hardware (such as an automatic patrol trolley) in the scene, data transmission safety can be guaranteed, the classroom learning requirements of students are met, and the learning experience and the learning interest of the students in AI courses are improved.
Drawings
In order to more clearly illustrate the embodiments of the present application or the technical solutions in the prior art, the drawings needed to be used in the description of the embodiments or the prior art will be briefly introduced below, and it is obvious that the drawings in the following description are some embodiments of the present application, and it is obvious for those skilled in the art to obtain other drawings based on these drawings without creative efforts.
FIG. 1 is a schematic flow chart of an AI enabling method for hardware according to an embodiment of the present disclosure;
FIG. 2 is a schematic structural diagram of an energized model provided in an embodiment of the present application;
FIG. 3 is a schematic flow chart for constructing an enabling model according to an embodiment of the present application;
FIG. 4 is a schematic flow chart of a verification enabling model provided by an embodiment of the present application;
fig. 5 is a schematic composition diagram of a hardware AI energization apparatus according to an embodiment of the present application.
Detailed Description
When the AI capability is provided for some open source hardware devices with low computation capability, the AI function of the open source hardware can be realized only depending on the AI capability of the cloud due to the limitation of the computation capability of the processor of the open source hardware, that is, the open source hardware needs to transmit the acquired data to the cloud AI engine through the wide area network, and then returns an AI processing result to the open source hardware after the computation of the cloud AI engine is completed. However, in the AI enabling mode for hardware, for an artificial intelligence innovative education application scenario with a slow network transmission speed, not only is the data transmission rate low, but also the learning interest of students on AI courses cannot be well stimulated only by simple AI capability calling, and students can only see the processing result of the AI capability called from the cloud, cannot well participate in and experience the specific implementation process of the AI capability, cannot fully understand AI knowledge points, and are poor in experience.
In order to solve the above defects, an embodiment of the present application provides an AI enabling method for hardware, which includes acquiring a running image of hardware to be enabled, and then sending the running image to a pre-constructed enabling model to predict a running state of the hardware to be enabled at a next moment; the hardware to be energized and the equipment where the energized model is located are located in the same local area network, and then the hardware to be energized can be controlled to automatically run according to the running state of the next moment. Therefore, the embodiment of the application utilizes the pre-trained enabling model to perform self-defined AI enabling on the hardware to be enabled so as to control the hardware to be enabled to automatically operate at the next moment according to the operation state predicted by the model.
Therefore, in the application scene of artificial intelligence innovation education, teachers and students can participate in the construction process of energized models, and the pre-trained self-defined energized models can be used for carrying out self-defined AI energization on open source hardware (such as an automatic patrol trolley) in the scene, so that the data transmission safety can be guaranteed, the classroom learning requirements of students are met, and the learning experience of the students in AI courses is improved. The enabling model is mainly used for predicting the operating state of hardware to be enabled at the next moment, and any model with the function of the operating state can be used as the enabling model of the application. Such as neural network models, gaussian models, and the like.
In order to make the objects, technical solutions and advantages of the embodiments of the present application clearer, the technical solutions in the embodiments of the present application will be clearly and completely described below with reference to the drawings in the embodiments of the present application, and it is obvious that the described embodiments are some embodiments of the present application, but not all embodiments. All other embodiments, which can be derived by a person skilled in the art from the embodiments given herein without making any creative effort, shall fall within the protection scope of the present application.
First embodiment
Referring to fig. 1, a flow chart of an AI energizing method for hardware provided in this embodiment is schematically illustrated, where the method includes the following steps:
s101: a running image of the hardware to be energized is acquired.
In this embodiment, any open source hardware that implements AI enabling by using this embodiment is defined as hardware to be enabled, and in order to enable flexible custom AI capability to the hardware to be enabled, this application first needs to acquire a running image of the hardware to be enabled. For performing the following step S102. It should be noted that the present embodiment does not limit the type of the running image, for example, the running image may be a color image composed of three primary colors of red (G), green (G), and blue (B), or may be a grayscale image.
Moreover, it should be noted that the running image is acquired by the open source hardware, but not acquired cloud fixed data, for example, an image of a scene around the running hardware to be energized may be acquired by a camera carried by the hardware to be energized, and the image is used as the running image of the hardware to be energized.
For example, the following steps are carried out: assuming that in an application scene of an artificial intelligence innovation education classroom, the hardware to be energized is an automatic patrol trolley, in order to energize an AI for the trolley, the automatic patrol trolley is controlled to automatically run according to the prediction state of a model so as to improve the learning experience of students in an AI course, 20 runway scene pictures at two sides of the trolley during running can be shot and stored by using a vehicle-mounted camera of the trolley to serve as a running image of the trolley, and the subsequent step S102 is executed.
Or in another implementation manner, a camera around the hardware to be energized can acquire images of a scene around the hardware to be energized in real time during operation and transmit the images to the hardware to be energized, and the images received by the hardware to be energized in real time are used as operation images of the hardware to be energized.
S102: sending the running image to a pre-constructed enabling model to predict the running state of the hardware to be enabled at the next moment; and the hardware to be energized and the equipment where the energized model is located are positioned in the same local area network.
In this embodiment, after the operation image of the hardware to be energized is acquired in step S101, the operation image may be further sent to an energizing model that is constructed in advance, so that after the model performs feature extraction on the operation image, according to the extracted image features, the operation state of the hardware to be energized at the next time is predicted, so as to execute subsequent step S103. And the hardware to be energized and the equipment where the energized model is located are positioned in the same local area network.
Specifically, in an alternative implementation, the enabled model structure is as shown in fig. 2, and the model may include a feature extraction network composed of Resnet-18 and VGG16, a fully connected layer, and an output layer, after the running image is sent to the enabled model, image features (including shallow edge texture features and deep semantic features of the image) of the running image may be first extracted using the feature extraction network composed of Resnet-18 and VGG16, and output features of the last convolutional layer of Resnet-18 and VGG16 may be connected, so that 512-layer features of a single network may be doubled to 1024-layer features to obtain more various information of the running image. And then, inputting the spliced image features into a subsequent full-connection layer and an output layer so as to output the running state of the hardware to be energized at the next moment through the output layer, thereby reducing the risk of model overfitting and improving the prediction accuracy and the effectiveness of the model.
For example, the following steps are carried out: based on the above example, in the application scenario of the artificial intelligence innovation education classroom, after 20 runway scene pictures at two sides of the vehicle during driving are captured and stored by using the vehicle-mounted camera of the automatic patrol car and are used as the running image of the car, the 20 runway scene pictures can be further input into the feature extraction network composed of Resnet-18 and VGG16 in the enabled model shown in fig. 2 to extract the picture features (including the shallow edge texture feature and the deep semantic feature of the image) of the 20 runway scene pictures, and the image features after the connection of the output features of the last convolutional layer of the Resnet-18 and VGG16 are input into the subsequent full connection layer and output layer, so as to output the target position of the automatic patrol car to advance at the next moment through the output layer.
For example, the following steps are carried out: based on the above example, in the application scenario of the artificial intelligence innovation education classroom, after the on-board camera of the automatic patrol car is used to shoot the runway scene pictures at two sides of the car during driving in real time to serve as the running image of the car, the runway scene pictures can be further input into the feature extraction network composed of Resnet-18 and VGG16 in the enabling model shown in fig. 2 to extract the picture features (including the shallow edge texture feature and the deep semantic feature of the image) of the runway scene pictures, and the image features after the output features of the last convolutional layer of the Resnet-18 and VGG16 are connected are input into the subsequent full connection layer and output layer, so as to output the target position of the automatic patrol car to advance at the next moment through the output layer.
It should be noted that, in a possible implementation manner of the embodiment of the present application, in order to reduce network delay and improve real-time performance and safety of data transmission, before sending the running image to the pre-constructed enabling model, a communication connection needs to be established in the same local area network with the device where the enabling model is located in advance. And then the running image can be sent to a pre-constructed enabling model through the communication connection so as to predict the running state of the hardware to be enabled at the next moment through the model.
In this implementation, the open source hardware may determine whether an enabling model exists in the same lan by sending a broadcast. Specifically, the open source hardware may create a socket of a SOCK _ DGRAM type based on a User Datagram Protocol (UDP), so as to calculate a local area network broadcast address (referred to as an address for sending and receiving broadcast information) "192.168.40.255" after acquiring a network segment address of a to-be-local area network (e.g., "192.168.40"), and send a command to a preset custom port of the broadcast address (e.g., 8001 port, 8002 port, etc.), where the command includes an IP of the open source hardware, so as to call a recvfrom method to implement a function of waiting for a device IP and port information returned by a device where an enabling model is located, so as to perform a communication connection with the device where the enabling model is located.
Accordingly, for the device where the enabling model is located, it can be determined whether there is open source hardware that needs to perform AI enabling in the same lan by receiving a BROADCAST, and specifically, the device can also create a socket of a socket _ DGRAM type based on a UDP protocol and set SO _ BROADCAST supported BROADCAST, monitor a preset custom port (e.g., 8001 port, 8002 port, 12346 port, etc.) and receive a BROADCAST message of the open source hardware that needs to perform AI enabling in the lan, and feed back information of the local network IP and the port of the device itself as response information to the open source hardware.
On the basis, the open source hardware can transmit data to the equipment where the enabling model is located based on a Hypertext Transfer Protocol (http) request and a wireless network WIFI in a local area network according to the IP and the port of the equipment where the enabling model is located, and the equipment is returned by broadcasting, so that the data is only transmitted and used inside the local area network, and the safety of data transmission is guaranteed.
It should be noted that the device where the enabling model is located may be an edge computing device deployed in a computing device such as a server, and the enabling model may be a user-defined edge computing engine. Therefore, the most near-end AI enabling service can be provided for open source hardware nearby based on the characteristics of edge computing, specifically, an AI enabling application program is initiated at the edge side, so that a faster network service response can be generated, and the basic requirements of real-time business, application intelligence, safety, privacy protection and the like can be met.
It should be noted that, in order to implement this step S102, an enabling model needs to be constructed in advance, and a specific construction process can be referred to in the following description of the second embodiment.
S103: and controlling the hardware to be energized to automatically run according to the running state of the next moment.
In this embodiment, after the operation state of the to-be-enabled hardware at the next time is predicted in step S102, the to-be-enabled hardware may be further controlled to automatically operate according to the predicted operation state at the next time, so as to enable the AI of the to-be-enabled hardware.
For example, the following steps are carried out: based on the above example, in the application scenario of the artificial intelligence innovation education classroom, after the energized model is used for outputting the target position of the automatic patrol trolley to advance at the next moment, the trolley can automatically make corresponding motion behaviors such as advancing, left-turning, right-turning and the like at the next moment according to the prediction result until the trolley advances to the predicted target position. Therefore, teachers and students can well participate in and experience the process of realizing the AI capacity, so that the knowledge points of the AI are fully understood, and the learning experience of the students in the artificial intelligence course is improved.
In summary, in the AI enabling method for hardware provided in this embodiment, first, a running image of the hardware to be enabled is obtained, and then, the running image is sent to a pre-constructed enabling model to predict a running state of the hardware to be enabled at the next moment; the hardware to be energized and the equipment where the energized model is located are located in the same local area network, and then the hardware to be energized can be controlled to automatically run according to the running state of the next moment. Therefore, in the embodiment of the application, the pre-trained enabling model is used for enabling the AI to be customized for the hardware to be enabled so as to control the hardware to be enabled to automatically operate at the next moment according to the operation state predicted by the model, and as the hardware to be enabled and the equipment where the enabling model is located are located in the same local area network, data transmission between the hardware to be enabled and the equipment is only carried out in the local area network, so that the safety of data transmission can be ensured.
Therefore, in the application scene of artificial intelligence innovation education, teachers and students can utilize a pre-trained self-defined enabling model to carry out self-defined AI enabling on open source hardware (such as an automatic patrol trolley) in the scene, data transmission safety can be guaranteed, the classroom learning requirements of students are met, and the learning experience and the learning interest of the students in AI courses are improved.
Second embodiment
This embodiment will describe a process of constructing the energized model mentioned in the above embodiment.
Referring to fig. 3, there is shown a schematic flow chart of constructing an enabling model provided by the embodiment, where the flow chart includes the following steps:
s301: and acquiring a training running image of the hardware.
In the present embodiment, in order to construct an enabling model, a large amount of preparation work needs to be performed in advance, and first, a large amount of hardware running images need to be collected as training images. For example, taking a trolley in an application scene of an artificial intelligent innovative education classroom as hardware as an example, in order to endow the trolley with automatic patrol AI capability, when an enabling model is constructed, teachers and students can shoot and store 100 runway scene pictures of two sides of the trolley during driving as sample image data, and meanwhile, in order to enrich the number of samples and improve the model training effect, a method of transfer learning (namely, using a feature extraction model which is pre-trained on other data sets) can be adopted to obtain more running images of the trolley as the sample image data, so that the learning time cost of the teachers and students can be reduced, and the number of the sample images can be greatly improved. Moreover, it is necessary to manually mark the running state of the hardware corresponding to the sample images at the next time in the current scene, for example, the position (as can be represented by coordinate points in the scene image) of the trolley to be advanced at the next time in the sample images, so as to train the energized model.
S302: training the initial enabling model according to the training running image of the hardware and the running state label of the hardware at the next moment corresponding to the training running image of the hardware to generate an enabling model; wherein the hardware and the device where the initial enabling model is located are located in the same local area network.
In this embodiment, after the training run image of the hardware is acquired in step S301, the training run image can be input into an initial enabling model in the same local area network for training, so as to generate an enabling model. Wherein the initial enabling model comprises a feature extraction network, a fully connected layer and an output layer. And the feature extraction network may in turn comprise a convolutional neural network Resnet-18 and/or a convolutional neural network VGG 16.
Specifically, during the current round of training, the running image of the hardware to be energized in the first embodiment may be replaced by the sample image obtained in the current round, and the running state of the hardware at the next time corresponding to the sample image may be predicted according to the execution process in the first embodiment through the current initial energizing model. Then, the predicted running state of the hardware at the next moment corresponding to the sample image can be compared with the running state of the manually marked hardware at the next moment corresponding to the sample image, the model parameters are updated according to the difference between the predicted running state of the hardware at the next moment and the predicted running state of the manually marked hardware at the next moment, and the updating of the model parameters is stopped until a preset condition is met, for example, the difference has a small change amplitude, the training of the enabling model is completed, and the trained enabling model is generated.
For example, the following steps are carried out: based on the above example, in the current round of training, the sample image is the running image of the cart in the application scenario of the artificial intelligent innovation education classroom, the target position (which can be represented by the coordinate result) where the cart corresponding to the running image will advance at the next time can be predicted by the above method, then the mean-square error (MSE for short) between the coordinate result of the target position where the cart corresponding to the running image will advance at the next time and the coordinate of the position where the cart to advance at the next time is manually labeled can be calculated, and the network parameter is updated by using the back propagation algorithm, for example, a random gradient descent method (mini-batch sgd) can be used as an optimization strategy, a small batch of multiple sample images are used for training each time to adjust the model parameter, so that all training data can be iterated by fewer times, the convergence of the model is accelerated, the generation time of the enabling model is further reduced, and the trial and error cost of teachers and students is also reduced.
By the embodiment, the energized model can be generated according to the sample image training, and further, the energized model can be verified by using the verification image. The specific verification process may include the following steps S401 to S403:
step S401: and acquiring a verification running image of the hardware.
In this embodiment, in order to implement verification of the enabling model, a verification run image of the hardware is first acquired, where the verification run image of the hardware refers to data information that can be used for verification of the enabling model, and after the verification data is acquired, the subsequent step S402 can be continued.
Step S402: and inputting the verification running image of the hardware into an enabling model to obtain a running state prediction result of the hardware at the next moment.
After the verification operation image of the hardware is acquired in step S401, the verification operation image can be further input to the enabling model, so that the operation state prediction result of the hardware at the next time is obtained through the execution process in step S102 in the first embodiment, and the obtained enabling model is verified.
Step S403: and when the running state prediction result of the hardware at the next moment is inconsistent with the running state marking result of the hardware at the next moment corresponding to the verification running image of the hardware, taking the verification running image of the hardware as the training running image of the hardware again, and updating the enabling model.
After the operation state prediction result of the hardware at the next time is obtained in step S402, if the operation state prediction result of the hardware at the next time is inconsistent with the operation state result of the hardware manually labeled at the next time corresponding to the verification operation image of the hardware, the verification operation image of the hardware may be used again as the sample image to perform parameter update on the enabling model.
Through the embodiment, the energized model can be effectively verified by using the verification running image of the hardware, and when the running state prediction result of the hardware at the next moment corresponding to the verification running image is inconsistent with the running state result of the manually marked hardware at the next moment corresponding to the verification running image, the energized model can be timely adjusted and updated, so that the prediction precision and accuracy of the energized model can be improved.
In summary, the enabling model trained by the embodiment can be used for quickly and accurately predicting the operating state of the hardware to be enabled at the next moment by using the operating image of the hardware to be enabled, so that effective AI enabling can be performed on the hardware to be enabled subsequently. And owing to can support the user to construct self-defined enabling model to can improve the user and carry out the participation degree when AI enables to hardware, and then can improve user experience, for example, in the application scene in artificial intelligence innovation education classroom, teachers and students give the AI ability of automatic tour for the dolly through constructing self-defined enabling model, thereby can improve student's experience of learning in the AI course and interest in learning.
Third embodiment
In this embodiment, a hardware AI enabling apparatus will be described, and for related contents, refer to the above method embodiment.
Referring to fig. 5, a schematic composition diagram of a hardware AI energizing apparatus provided in this embodiment is shown, where the apparatus 500 includes:
a first obtaining unit 501, configured to obtain an operation image of hardware to be energized;
a sending unit 502, configured to send the running image to a pre-constructed enabling model, so as to predict a running state of the hardware to be enabled at a next moment; the hardware to be energized and the equipment where the edge calculation AI model is located are located in the same local area network;
and a control unit 503, configured to control the hardware to be enabled to automatically operate according to the operation state at the next time.
In an implementation manner of this embodiment, the apparatus further includes:
the establishing unit is used for establishing communication connection with the equipment where the enabling model is located in the same local area network in advance;
the sending unit 502 is specifically configured to:
and sending the running image to a pre-constructed enabling model through the communication connection so as to predict the running state of the hardware to be enabled at the next moment.
In an implementation manner of this embodiment, the apparatus further includes:
the second acquisition unit is used for acquiring a training running image of the hardware;
the training unit is used for training an initial energized model according to the training running image of the hardware and the running state label of the hardware at the next moment corresponding to the training running image of the hardware, and generating the energized model; the hardware and the device where the initial enabling model is located are located in the same local area network.
In one implementation of this embodiment, the initial enabling model includes a feature extraction network, a fully connected layer, and an output layer.
In one implementation of this embodiment, the feature extraction network includes a convolutional neural network Resnet-18 and/or a convolutional neural network VGG 16.
In an implementation manner of this embodiment, the apparatus further includes:
the third acquisition unit is used for acquiring a verification running image of the hardware;
the input unit is used for inputting the verification running image of the hardware into the enabling model and obtaining a running state prediction result of the hardware at the next moment;
and the updating unit is used for taking the verification running image of the hardware as the training running image of the hardware again to update the enabling model when the running state prediction result of the hardware at the next moment is inconsistent with the running state marking result of the hardware at the next moment corresponding to the verification running image of the hardware.
Further, an embodiment of the present application also provides an AI enabling apparatus for hardware, including: a processor, a memory, a system bus;
the processor and the memory are connected through the system bus;
the memory is for storing one or more programs, the one or more programs including instructions, which when executed by the processor, cause the processor to perform any of the above-described hardware AI enabling methods.
Further, an embodiment of the present application also provides a computer-readable storage medium, in which instructions are stored, and when the instructions are executed on a terminal device, the instructions cause the terminal device to execute any implementation method of the AI enabling method of the hardware.
Further, an embodiment of the present application also provides a computer program product, which, when running on a terminal device, causes the terminal device to execute any implementation method of the above-mentioned AI enabling method for hardware.
As can be seen from the above description of the embodiments, those skilled in the art can clearly understand that all or part of the steps in the above embodiment methods can be implemented by software plus a necessary general hardware platform. Based on such understanding, the technical solution of the present application may be essentially or partially implemented in the form of a software product, which may be stored in a storage medium, such as a ROM/RAM, a magnetic disk, an optical disk, etc., and includes several instructions for enabling a computer device (which may be a personal computer, a server, or a network communication device such as a media gateway, etc.) to execute the method according to the embodiments or some parts of the embodiments of the present application.
It should be noted that, in the present specification, the embodiments are described in a progressive manner, each embodiment focuses on differences from other embodiments, and the same and similar parts among the embodiments may be referred to each other. The device disclosed by the embodiment corresponds to the method disclosed by the embodiment, so that the description is simple, and the relevant points can be referred to the method part for description.
It is further noted that, herein, relational terms such as first and second, and the like may be used solely to distinguish one entity or action from another entity or action without necessarily requiring or implying any actual such relationship or order between such entities or actions. Also, the terms "comprises," "comprising," or any other variation thereof, are intended to cover a non-exclusive inclusion, such that a process, method, article, or apparatus that comprises a list of elements does not include only those elements but may include other elements not expressly listed or inherent to such process, method, article, or apparatus. Without further limitation, an element defined by the phrase "comprising an … …" does not exclude the presence of other identical elements in a process, method, article, or apparatus that comprises the element.
The previous description of the disclosed embodiments is provided to enable any person skilled in the art to make or use the present application. Various modifications to these embodiments will be readily apparent to those skilled in the art, and the generic principles defined herein may be applied to other embodiments without departing from the spirit or scope of the application. Thus, the present application is not intended to be limited to the embodiments shown herein but is to be accorded the widest scope consistent with the principles and novel features disclosed herein.
Claims (10)
1. A method for AI enablement of hardware, comprising:
acquiring a running image of hardware to be energized;
sending the running image to a pre-constructed enabling model to predict the running state of the hardware to be enabled at the next moment; the hardware to be energized and the equipment where the energized model is located are located in the same local area network;
and controlling the hardware to be energized to automatically operate according to the operating state of the next moment.
2. The method of claim 1, wherein prior to sending the running image to a pre-constructed energized model, further comprising:
establishing communication connection with the equipment where the enabling model is located in the same local area network in advance;
then, the sending the running image to a pre-constructed enabling model to predict the running state of the hardware to be enabled at the next moment comprises:
and sending the running image to a pre-constructed enabling model through the communication connection so as to predict the running state of the hardware to be enabled at the next moment.
3. The method of claim 1, wherein constructing the energized model comprises:
acquiring a training running image of hardware;
training an initial enabling model according to the training running image of the hardware and the running state label of the hardware at the next moment corresponding to the training running image of the hardware to generate the enabling model; the hardware and the device where the initial enabling model is located are located in the same local area network.
4. The method of claim 3, wherein the initial energization model comprises a feature extraction network, a fully connected layer, and an output layer.
5. The method of claim 4, wherein the feature extraction network comprises a convolutional neural network Resnet-18 and/or a convolutional neural network VGG 16.
6. The method according to any one of claims 3 to 5, further comprising:
acquiring a verification running image of the hardware;
inputting the verification running image of the hardware into the enabling model to obtain a running state prediction result of the hardware at the next moment;
and when the running state prediction result of the hardware at the next moment is inconsistent with the running state marking result of the hardware at the next moment corresponding to the verification running image of the hardware, taking the verification running image of the hardware as the training running image of the hardware again, and updating the enabling model.
7. An AI enabling apparatus for hardware, comprising:
the first acquisition unit is used for acquiring an operation image of hardware to be energized;
the sending unit is used for sending the running image to a pre-constructed enabling model so as to predict the running state of the hardware to be enabled at the next moment; the hardware to be energized and the equipment where the edge calculation AI model is located are located in the same local area network;
and the control unit is used for controlling the hardware to be energized to automatically run according to the running state of the next moment.
8. An AI-enabling apparatus for hardware, comprising: a processor, a memory, a system bus;
the processor and the memory are connected through the system bus;
the memory is to store one or more programs, the one or more programs comprising instructions, which when executed by the processor, cause the processor to perform the method of any of claims 1-6.
9. A computer-readable storage medium having stored therein instructions that, when executed on a terminal device, cause the terminal device to perform the method of any one of claims 1-6.
10. A computer program product, characterized in that the computer program product, when run on a terminal device, causes the terminal device to perform the method of any of claims 1-6.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202011641320.7A CN112785466A (en) | 2020-12-31 | 2020-12-31 | AI enabling method and device of hardware, storage medium and equipment |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202011641320.7A CN112785466A (en) | 2020-12-31 | 2020-12-31 | AI enabling method and device of hardware, storage medium and equipment |
Publications (1)
| Publication Number | Publication Date |
|---|---|
| CN112785466A true CN112785466A (en) | 2021-05-11 |
Family
ID=75753491
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN202011641320.7A Pending CN112785466A (en) | 2020-12-31 | 2020-12-31 | AI enabling method and device of hardware, storage medium and equipment |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN112785466A (en) |
Citations (5)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN110534009A (en) * | 2019-09-05 | 2019-12-03 | 北京青橙创客教育科技有限公司 | A kind of unmanned course teaching aid of artificial intelligence |
| CN111047048A (en) * | 2019-11-22 | 2020-04-21 | 支付宝(杭州)信息技术有限公司 | Energized model training and merchant energizing method and device, and electronic equipment |
| CN111316286A (en) * | 2019-03-27 | 2020-06-19 | 深圳市大疆创新科技有限公司 | Trajectory prediction method and device, storage medium, driving system and vehicle |
| WO2020168660A1 (en) * | 2019-02-19 | 2020-08-27 | 平安科技（深圳）有限公司 | Method and apparatus for adjusting traveling direction of vehicle, computer device and storage medium |
| CN111890365A (en) * | 2020-07-31 | 2020-11-06 | 平安科技（深圳）有限公司 | Target tracking method and device, computer equipment and storage medium |
-
2020
- 2020-12-31 CN CN202011641320.7A patent/CN112785466A/en active Pending
Patent Citations (5)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| WO2020168660A1 (en) * | 2019-02-19 | 2020-08-27 | 平安科技（深圳）有限公司 | Method and apparatus for adjusting traveling direction of vehicle, computer device and storage medium |
| CN111316286A (en) * | 2019-03-27 | 2020-06-19 | 深圳市大疆创新科技有限公司 | Trajectory prediction method and device, storage medium, driving system and vehicle |
| CN110534009A (en) * | 2019-09-05 | 2019-12-03 | 北京青橙创客教育科技有限公司 | A kind of unmanned course teaching aid of artificial intelligence |
| CN111047048A (en) * | 2019-11-22 | 2020-04-21 | 支付宝(杭州)信息技术有限公司 | Energized model training and merchant energizing method and device, and electronic equipment |
| CN111890365A (en) * | 2020-07-31 | 2020-11-06 | 平安科技（深圳）有限公司 | Target tracking method and device, computer equipment and storage medium |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| US11587344B2 (en) | Scene understanding and generation using neural networks | |
| CN109990790B (en) | Unmanned aerial vehicle path planning method and device | |
| CN109768940B (en) | Flow distribution method and device for multi-service SDN | |
| EP3998550A1 (en) | Scene interaction method and apparatus, electronic device, and computer storage medium | |
| US20190126472A1 (en) | Reinforcement and imitation learning for a task | |
| DE112011105104B4 (en) | Collaborative image control | |
| JP2021503134A (en) | Unsupervised learning of image depth and egomotion prediction neural networks | |
| Khetarpal et al. | Options of interest: Temporal abstraction with interest functions | |
| WO2019104713A1 (en) | Machine learning method, master node, work node, and system | |
| CN109313484A (en) | Virtual reality interactive system, method and computer storage medium | |
| Conway-Jones et al. | Demonstration of federated learning in a resource-constrained networked environment | |
| CN109782600A (en) | A method of autonomous mobile robot navigation system is established by virtual environment | |
| CN114584581A (en) | Federal learning system and federal learning training method for smart city Internet of things and letter fusion | |
| Li et al. | Adaptive traffic signal control model on intersections based on deep reinforcement learning | |
| CN102413378A (en) | Adaptive neural network-based lost packet recovery method in video transmission | |
| CN115249073A (en) | Method and device for federated learning | |
| CN111860777B (en) | Distributed reinforcement learning training method and device for super real-time simulation environment | |
| CN117529728A (en) | Privacy-aware pruning in machine learning | |
| Liu et al. | Task offloading optimization of cruising UAV with fixed trajectory | |
| WO2016114261A4 (en) | Autonomous learning system using video/audio clips | |
| CN112785466A (en) | AI enabling method and device of hardware, storage medium and equipment | |
| JP4857675B2 (en) | Application integrated management system, apparatus and program, integrated session management server, system, program and server housing, communication system, session control server, and integrated application server | |
| CN115713128A (en) | Federal learning method based on equipment training time fairness | |
| Debner et al. | Scalability of a machine learning environment for autonomous driving research | |
| CN110149390B (en) | Network simulation method and system applied to online officer examination training platform |
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 |