CN211827319U

CN211827319U - Sample data generation system for machine learning

Info

Publication number: CN211827319U
Application number: CN202020398364.0U
Authority: CN
Inventors: 张和辉; 吴臻志; 王红伟
Original assignee: Beijing Lynxi Technology Co Ltd
Current assignee: Beijing Lynxi Technology Co Ltd
Priority date: 2020-03-25
Filing date: 2020-03-25
Publication date: 2020-10-30
Anticipated expiration: 2030-03-25

Abstract

The present disclosure relates to a sample data generation system for machine learning. The method comprises the following steps: the display device is used for displaying the target object; the movement device is used for bearing the display device to move; an image pickup device for acquiring a target image of the target object in motion; the speed measuring device is used for acquiring the movement speed of the movement device; and the processing device is used for acquiring the category and the current time of the target object, determining the current position of the target object according to the motion speed and the current time, and generating sample data for machine learning based on the category, the current position and the target image. The sample data generation system for machine learning can acquire image data moving at a high speed in a simple and rapid mode under the condition of low space requirement, can label the image data in an automatic and efficient mode to generate sample data for machine learning, and is high in labeling efficiency.

Description

Sample data generation system for machine learning

Technical Field

The disclosure relates to the field of computer information processing, in particular to a sample data generation system for machine learning.

Background

To construct a visual system based on a deep learning technology, a series of processes such as data acquisition, data cleaning, deep network training, data post-processing and the like are inevitably required, and the data acquisition is the first step in the whole process and is also a very critical step. In recent years, computer vision-related systems have become relatively sophisticated, and in particular, in terms of target detection and identification, detection and identification of static objects or low-speed moving objects have reached a high level. However, for some detection and identification tasks aiming at a high-speed moving target, such as an unmanned vehicle avoiding pedestrians suddenly rushing into the field of view, an unmanned vehicle quickly avoiding obstacles, factory high-speed gear fault detection and the like, a lot of difficulties still exist, and a large part of reasons are that image data acquisition of high-speed moving is difficult, high-quality data cannot be acquired or is difficult to acquire, and even if data acquisition, data acquisition and labeling work can be carried out, the time and the labor are usually wasted.

Therefore, a new sample data generation system for machine learning is required.

The above information disclosed in this background section is only for enhancement of understanding of the background of the disclosure and therefore it may contain information that does not constitute prior art that is already known to a person of ordinary skill in the art.

SUMMERY OF THE UTILITY MODEL

In view of this, the present disclosure provides a sample data generation system for machine learning, which can acquire high-speed moving image data in a simple and fast manner under a condition of low space requirement, and can label the image data in an automatic and efficient manner to generate sample data for machine learning, and the labeling efficiency is high.

Additional features and advantages of the disclosure will be set forth in the detailed description which follows, or in part will be obvious from the description, or may be learned by practice of the disclosure.

According to an aspect of the present disclosure, a sample data generation system for machine learning is provided, the system including: the display device is used for displaying the target object; the movement device is used for bearing the display device to move; an image pickup device for generating a target image of the target object in motion; the speed measuring device is used for acquiring the movement speed of the movement device; and the processing device is used for acquiring the category and the current time of the target object, determining the current position of the target object according to the motion speed and the current time, and generating sample data for machine learning based on the category, the current position and the target image.

In an exemplary embodiment of the present disclosure, the display device includes: the display screen is used for displaying the target object; and the display screen controller is used for determining the category of the target object and controlling the display screen to display the target object.

In an exemplary embodiment of the present disclosure, the display screen controller is configured to send a category of the target object to the processing apparatus when the display screen is controlled to display the target object;

the camera device is used for sending the acquired target image of the target object to the processing device;

the processing device is used for receiving the category of the target object sent by the display screen controller and the target image of the target object sent by the camera device, determining the current position of the target object according to the motion speed and the current time, and generating sample data for machine learning based on the category, the current position and the target image.

In an exemplary embodiment of the present disclosure, the processing device is configured to determine an initial time and an initial position of the target object, determine a target angle according to the movement speed, the current time and the initial time, and determine a current position of the target object according to the target angle and the initial position.

In an exemplary embodiment of the present disclosure, the exercise device includes: the driving motor is used for driving the bearing unit to rotate; the speed regulating device is used for controlling the rotating speed of the driving motor; and the bearing unit is used for bearing the display device to move.

In an exemplary embodiment of the present disclosure, the exercise device further includes: a fixing device for fixing the driving motor; and a damping device for reducing rotation of the driving motor.

In an exemplary embodiment of the present disclosure, the bearing unit includes: a rotating disk or drum.

In an exemplary embodiment of the disclosure, the processing device is configured to take the target image as raw data in machine learning, take the category and the current position as tags of the raw data, and generate the sample data from the tagged raw data.

In an exemplary embodiment of the present disclosure, the image pickup device includes one of a dynamic vision sensor camera, an event camera, and a retina imitation camera.

In an exemplary embodiment of the present disclosure, the storage module is configured to store the sample data.

According to an aspect of the present disclosure, an electronic device is provided, the electronic device including: one or more processors; storage means for storing one or more programs; when executed by one or more processors, cause the one or more processors to implement a method as above.

According to an aspect of the disclosure, a computer-readable medium is proposed, on which a computer program is stored, which program, when being executed by a processor, carries out the method as above.

According to the sample data generation system for machine learning disclosed by the invention, a target image of a moving target object is acquired; acquiring the category and the movement speed of the target object; determining the current position of the target object according to the motion speed and the current time; the mode of generating the sample data for machine learning based on the target image, the category and the current position can acquire the high-speed moving target image in a simple and rapid mode under the condition of low space requirement, and can label the target image in an automatic and efficient mode to generate the sample data for machine learning, thereby greatly reducing the manual workload.

It is to be understood that both the foregoing general description and the following detailed description are exemplary and explanatory only and are not restrictive of the disclosure.

Drawings

The above and other objects, features and advantages of the present disclosure will become more apparent by describing in detail exemplary embodiments thereof with reference to the attached drawings. The drawings described below are merely some embodiments of the present disclosure, and other drawings may be derived from those drawings by those of ordinary skill in the art without inventive effort.

Fig. 1 is a block diagram illustrating a sample data generation system for machine learning, according to an example embodiment.

Fig. 2 is a schematic diagram illustrating a sample data generation system for machine learning, according to an example embodiment.

Fig. 3 is a schematic diagram illustrating a sample data generation system for machine learning, according to another exemplary embodiment.

Fig. 4 is a flowchart illustrating a sample data generation method for machine learning according to an exemplary embodiment.

FIG. 5 is a block diagram illustrating an electronic device in accordance with an example embodiment.

FIG. 6 is a block diagram illustrating a computer-readable medium in accordance with an example embodiment.

Detailed Description

Example embodiments will now be described more fully with reference to the accompanying drawings. Example embodiments may, however, be embodied in many different forms and should not be construed as limited to the embodiments set forth herein; rather, these embodiments are provided so that this disclosure will be thorough and complete, and will fully convey the concept of example embodiments to those skilled in the art. The same reference numerals denote the same or similar parts in the drawings, and thus, a repetitive description thereof will be omitted.

Furthermore, the described features, structures, or characteristics may be combined in any suitable manner in one or more embodiments. In the following description, numerous specific details are provided to give a thorough understanding of embodiments of the disclosure. One skilled in the relevant art will recognize, however, that the subject matter of the present disclosure can be practiced without one or more of the specific details, or with other methods, components, devices, steps, and so forth. In other instances, well-known methods, system implementations, or operations are not shown or described in detail to avoid obscuring aspects of the disclosure.

The block diagrams shown in the figures are functional entities only and do not necessarily correspond to physically separate entities. I.e. these functional entities may be implemented in the form of software, or in one or more hardware modules or integrated circuits, or in different networks and/or processor means and/or microcontroller means.

The flow charts shown in the drawings are merely illustrative and do not necessarily include all of the contents and operations/steps, nor do they necessarily have to be performed in the order described. For example, some operations/steps may be decomposed, and some operations/steps may be combined or partially combined, so that the actual execution sequence may be changed according to the actual situation.

It will be understood that, although the terms first, second, third, etc. may be used herein to describe various components, these components should not be limited by these terms. These terms are used to distinguish one element from another. Thus, a first component discussed below may be termed a second component without departing from the teachings of the disclosed concept. As used herein, the term "and/or" includes any and all combinations of one or more of the associated listed items.

It is to be understood by those skilled in the art that the drawings are merely schematic representations of exemplary embodiments, and that the blocks or processes shown in the drawings are not necessarily required to practice the present disclosure and are, therefore, not intended to limit the scope of the present disclosure.

The inventor of the present disclosure finds that, for the verification data acquisition of the high-speed target detection algorithm, the existing data acquisition method is time-consuming and labor-consuming, and cannot meet the requirement of fast iteration of the existing high-speed target detection algorithm, and the specific disadvantages are as follows:

1) the existing method has extremely high space requirement for data acquisition, and enough space must be provided to support the camera/target to move.

2) The existing method is difficult to control the moving speed of the target, and particularly under the condition of extremely high speed, the speed is difficult to control.

3) The existing method has high data acquisition cost, and a camera stabilizer, a guide rail capable of supporting a body to move at a high speed, a prop for simulating an environment and the like need to be purchased.

4) The time cost and the labor cost for manually labeling the data after the data are collected and collecting and labeling the data are extremely high.

5) The data acquisition scene is difficult to reproduce, and the performance of the algorithm is difficult to test in real time or the demonstration of the algorithm is difficult.

In order to solve the problems in the prior art, the present disclosure provides a sample data generation system for machine learning, which can acquire high-speed moving image data at a low cost under the condition of low space requirement, and can flexibly control the moving speed of a target in cooperation with a speed regulator. The method can also easily reproduce scenes during data acquisition, and can easily perform algorithm performance test or algorithm effect demonstration on the target detection and identification algorithm.

The following are embodiments of the disclosed apparatus that may be used to perform embodiments of the disclosed methods. For details not disclosed in the embodiments of the apparatus of the present disclosure, refer to the embodiments of the method of the present disclosure.

Fig. 1 is a block diagram illustrating a sample data generation system for machine learning, according to an example embodiment. As shown in fig. 1, the sample data generation system 10 for machine learning includes: the device comprises a display device 102, a motion device 104, a camera device 106, a speed measuring device 108 and a processing device 110.

The display device 102 is used for displaying the target object; wherein the display device 102 may comprise: the display screen is used for displaying the target object; and the display screen controller is used for determining the category of the target object and controlling the display image of the display screen.

The movement device 104 is used for carrying the display device to move; the motion device 14 may include: the driving motor is used for driving the bearing unit to rotate; the speed regulating device is used for controlling the rotating speed of the driving motor; and the bearing unit is used for bearing the display device to move. The motion device 104 may also include: a fixing device for fixing the driving motor; and a damping device for reducing rotation of the driving motor. Wherein, the bearing unit includes: a rotating disk or drum.

The camera 106 is used for generating a target image of the target object in motion;

the speed measuring device 108 is used for acquiring the movement speed of the movement device; and

the processing device 110 is configured to obtain a category and a current time of the target object, determine a current position of the target object according to the motion speed and the current time, and generate sample data for machine learning based on the category, the current position, and the target image.

The display controller in the display device 102 may send the category of the target object to the processing device 110, the processing device 110 receives the category of the target object sent by the display controller in the display device 102, and the processing device 110 may also read the category of the target object from the memory of the display controller based on a predetermined rule, which is not limited in this disclosure.

According to the sample data generation method and device for machine learning, a target image of a moving target object is acquired; acquiring the category and the movement speed of the target object; determining the current position of the target object according to the motion speed and the current time; the mode of generating the sample data for machine learning based on the target image, the category and the current position can acquire the target image moving at high speed in a simple and rapid mode under the condition of low space requirement, and can label the target image in an automatic and efficient mode to generate the sample data for machine learning, so that the generation efficiency of the sample is improved, and the generation cost of the sample is reduced.

In a possible implementation manner, the display screen controller is configured to send the category of the target object to the processing apparatus when controlling the display screen to display the target object;

In a possible implementation manner, the processing device is configured to determine an initial time and an initial position of the target object, determine a target angle according to the movement speed, the current time, and the initial time, and determine a current position of the target object according to the target angle and the initial position.

In a possible implementation manner, the processing device is configured to take the target image as raw data in machine learning, take the category and the current position as tags of the raw data, and generate the sample data from the tagged raw data.

In one possible implementation, the camera device includes one of a dynamic vision sensor camera, an event camera, a retinal imitation camera, and other high-speed cameras.

For example, the time when the processing device receives the real-time rotation speed sent by the tachometer is recorded as an initial time, which may be t 0; at the same time, recording a certain preset fixed position coordinate (usually the position coordinate of the display device) of the detection turntable (or the rotary drum) of the velocimeter at the time t0, and recording the position coordinate as (x0, y 0); the time when the processing device acquires the target image of the target object is recorded as the current time, the current time can be recorded as T1, and the time interval between the current time T1 and the initial time T0 is T1-T0 (unit: seconds).

The preset fixed position can be any position on the rotating disc, and the position can be indicated through obvious marks. The two-dimensional plane perpendicular to the tachometer detection turntable (or rotary drum) can be determined as a coordinate system in the present disclosure, and the origin of the coordinate system can be the circle center position of the tachometer detection turntable (or rotary drum).

When the current rotational speed of the dial is denoted by v0 (revolutions per second), the angle by which the display device displaying the target object rotates within the time T is 360 × v0 × T. Based on this angle θ, the position (x1, y1) of the target at time t1 can be calculated according to the following formula:

the high-speed target image frame of the current time sent by the high-speed camera can be recorded as Data, the target category/Label transferred by the controller can be recorded as Label, and the sample Data of the user performing the robot output at the moment is [ Data, (Label, x1, y1) ].

In one possible implementation, the method includes:

and the storage module is used for storing the sample data.

Therefore, the target image can be labeled in an automatic and efficient mode to generate the sample data for machine learning, the generation efficiency of the sample is improved, and the generation cost of the sample is reduced.

FIG. 2 is a schematic diagram illustrating sample data generation and system for machine learning, according to an example embodiment. As shown in fig. 2, the sample data generation and system 20 for machine learning may include:

1. a drive motor may be located in the motion device 104 for rotating the turntable.

2. A speed regulating device may be located in the movement device 104 for controlling the rotational speed of the rotor of the drive motor, which may be a continuously variable motor as shown.

3. A fixing device may be located in the moving device 104 for fixing the position of the driving motor so that it is not shaken sharply while rotating.

4. A damping device may be provided in the moving device 104 for reducing vibration generated when the driving motor rotates at a high speed;

5. a high speed camera may be located in the camera 106 for capturing images of a target rotating at high speed.

6. The turntable may be located in the motion device 104, rotated by a drive motor, and carry a display screen.

7. The display screen may be located in the display device 102, carried by the carousel, for dynamically displaying the target.

8. A tachometer may be located in the tachometer means 108 for measuring the rotational speed of the turntable in real time.

As shown in fig. 2, the sample data generation and system 20 for machine learning may further include:

a display screen controller (abbreviated as controller in the figure) may be located in the display device 102, and is configured to control the objects displayed on the display screen, and meanwhile, transmit the categories of the objects to the integration module;

the integration module may be located in the processing device 110, and is configured to receive real-time data acquired by the high-speed camera, a target type sent by the display screen controller, a real-time rotation speed of the turntable sent by the velocimeter, a current time, and a current position of the target object, integrate the information through a fixed algorithm process, and transmit the integrated data and a corresponding tag to the storage module.

And the storage module is used for storing the marked data.

As shown in fig. 2, the driving motor is located on the fixing device, the rotating head thereof drives the rotating disc to rotate at a high speed, and the rotating speed of the rotating head is adjusted by the speed adjusting device. The display screen can be pasted in the blank position of the front surface of the turntable, and the content displayed by the display screen is controlled by the display screen controller. And after the display screen controller controls the display screen to display the target image, the target type/label of the target is transmitted to the integration module in a wireless mode. Meanwhile, the high-speed camera transmits the real-time acquired target image of the high-speed moving target object to the integration module.

The tachometer detects the fixed position of carousel, the display screen is located directly behind, and the tachometer calculates the real-time rotational speed of carousel according to the time interval that detects the display screen twice to send for the integration module.

The integration module may have five inputs, respectively:

1) high-speed target image transmitted by high-speed camera

2) Object categories/labels delivered by display screen controller

3) Real-time rotating speed of rotary disc detected by velocimeter

4) Current time

5) The coordinates of the fixed position detected by the velocimeter are the positions detected by the velocimeter, and the coordinates of the fixed position can be used for calculating the current position, and the specific calculation process can refer to the embodiment corresponding to fig. 4.

Through the above inputs, the integration module may generate sample data for machine learning.

Fig. 3 is a schematic diagram illustrating a sample data generation system for machine learning, according to another exemplary embodiment. As shown in fig. 3, the sample data generation and system 30 for machine learning may include:

5. a high speed camera may be located in the camera 106 for capturing a target image of a target object rotating at a high speed.

6. The rotating drum may be located in the motion device 104, rotated by a drive motor, and carry a display screen.

8. The tachometer may be located in the tachometer means 108 to measure the rotational speed of the turntable in real time.

As shown in fig. 3, the sample data generation and system 30 for machine learning may further include:

the display screen controller (simply referred to as the controller in the figure) is used for controlling the targets displayed by the display screen and transmitting the types of the targets to the integration module;

the integration module may be located in the processing device 110, and is configured to receive real-time data acquired by the high-speed camera, a target type sent by the controller, a real-time rotation speed of the drum sent by the tachometer, determine a current time and a current position of the target object, integrate the information through a fixed algorithm process, and transmit the integrated data and the corresponding tag to the storage module.

And the storage module is used for storing the marked data.

In the sample data generating system for machine learning of fig. 3, not the turntable but the rotating drum is mounted with the display screen, and the display screen may be located on the side of the rotating drum, and the rotating direction of the driving motor is perpendicular to the plane. Otherwise the same as in the embodiment of fig. 2.

The embodiment in fig. 3 can achieve similar effects as the embodiment in fig. 2, but slightly different, the moving direction of the display screen of the embodiment in fig. 3 can be regarded as almost horizontal movement in the field of view of the high-speed camera, and the moving track of the display screen is obviously circular/arc-shaped for the high-speed camera in the embodiment in fig. 2.

According to the sample data generation system for machine learning, the characteristics of the stepless speed change rotating machine can be fully utilized to be matched with the rotary table or the rotary drum, the target image of the target object moving at a high speed is acquired at a lower cost under the condition of lower space requirement, and the moving speed of the target can be flexibly controlled by matching with the speed regulator; make full use of electronic display screen cooperation control module, infrared tachymeter, integration module to automatic, efficient mode is marked data, the reduction of very big degree manual work volume.

Based on the characteristics, the sample data generation system for machine learning can reproduce scenes in data acquisition easily, and can easily perform algorithm performance test or algorithm effect demonstration on the target detection and identification algorithm.

Fig. 4 is a flowchart illustrating a sample data generation method for machine learning according to an exemplary embodiment. The sample data generation method 40 for machine learning includes at least steps S402 to S408.

As shown in fig. 4, in S402, a target image of a moving target object is acquired. Can include the following steps: displaying the target object on a display device; and acquiring a target image of the target object while the display device is moving.

In one embodiment, the display device is fixed on a moving device, and the moving device drives the display device to move, and the moving device comprises: a rotating disk or drum.

In S404, the category and the movement speed of the target object are acquired. The method comprises the following steps: acquiring the category of the target object through a display device; and acquiring the movement speed of the target object through a velocimeter.

In one embodiment, obtaining the moving speed of the target object by a velocimeter includes: and the velocimeter acquires the rotating speed of the movement device as the movement speed of the target object.

In S406, the current position of the target object is determined from the movement speed and the current time. The method comprises the following steps: acquiring an initial time and an initial position of the movement of a target object; determining a target angle according to the movement speed, the current time and the initial moment; and determining the current position of the target object through the target angle and the initial position.

For example, the time when the processing device receives the real-time rotation speed sent by the tachometer may be recorded as an initial time, which may be t 0; at the same time, recording a certain preset fixed position coordinate (usually the position coordinate of the display device) of the detection turntable (or the rotary drum) of the velocimeter at the time t0, and recording the position coordinate as (x0, y 0); the time when the processing apparatus acquires the target image of the target object is recorded as the current time, the current time may be recorded as T1, and the time interval between the current time T1 and the initial time T0 is T1-T0 (unit: seconds).

The method may further include outputting to the storage module the data and tag pairs at all times during the recording period.

In S408, sample data for machine learning is generated based on the target image, the category, and the current position. The method comprises the following steps: taking the target image as original data in machine learning; taking the category and the current location as labels of the original data; and generating the sample data through the original data with the label.

According to the sample data generation method for machine learning, a target image of a moving target object is acquired; acquiring the category and the movement speed of the target object; determining the current position of the target object according to the motion speed and the current time; the mode of generating the sample data for machine learning based on the target image, the category and the current position can acquire the target image of the target object moving at a high speed in a simple and rapid mode under the condition of occupying less system space, and can label the image in an automatic and efficient mode to generate the sample data for machine learning, thereby greatly reducing the manual workload.

According to the sample data generation method for machine learning disclosed by the invention, the image acquisition device is high in acquisition efficiency, low in construction cost, flexible in speed adjustment, low in space requirement and easy in target replacement, the efficiency of high-speed moving image acquisition can be greatly improved, and the time cost and the labor cost of data acquisition are reduced.

It should be clearly understood that this disclosure describes how to make and use particular examples, but the principles of this disclosure are not limited to any details of these examples. Rather, these principles can be applied to many other embodiments based on the teachings of the present disclosure.

Those skilled in the art will appreciate that all or part of the steps implementing the above embodiments are implemented as computer programs executed by a CPU. When executed by the CPU, performs the functions defined by the above-described methods provided by the present disclosure. The program may be stored in a computer readable storage medium, which may be a read-only memory, a magnetic or optical disk, or the like.

Furthermore, it should be noted that the above-mentioned figures are only schematic illustrations of the processes involved in the methods according to exemplary embodiments of the present disclosure, and are not intended to be limiting. It will be readily understood that the processes shown in the above figures are not intended to indicate or limit the chronological order of the processes. In addition, it is also readily understood that these processes may be performed synchronously or asynchronously, e.g., in multiple modules.

Claims

1. A sample data generation system for machine learning, comprising:

the display device is used for displaying the target object;

the movement device is used for bearing the display device to move;

an image pickup device for acquiring a target image of the target object in motion;

the speed measuring device is used for acquiring the movement speed of the movement device; and

and the processing device is used for acquiring the category and the current time of the target object, determining the current position of the target object according to the motion speed and the current time, and generating sample data for machine learning based on the category, the current position and the target image.

2. The system of claim 1, wherein the display device comprises:

the display screen is used for displaying the target object; and

and the display screen controller is used for determining the category of the target object and controlling the display screen to display the target object.

3. The system of claim 2,

the display screen controller is used for sending the category of the target object to the processing device when the display screen is controlled to display the target object;

4. The system according to any one of claims 1-3,

the processing device is used for determining the initial time and the initial position of the target object, determining a target angle according to the movement speed, the current time and the initial time, and determining the current position of the target object according to the target angle and the initial position.

5. The system of claim 1, wherein the motion device comprises:

the driving motor is used for driving the bearing unit to rotate;

the speed regulating device is used for controlling the rotating speed of the driving motor; and

and the bearing unit is used for bearing the display device to move.

6. The system of claim 5, wherein the motion device further comprises:

a fixing device for fixing the driving motor; and

and the damping device is used for reducing the rotation of the driving motor.

7. The system of claim 5 or 6, wherein the carrying unit comprises: a rotating disk or drum.

8. The system of claim 1, wherein the processing means is configured to take the target image as raw data in machine learning, take the category and the current position as tags of the raw data, and generate the sample data from the tagged raw data.

9. The system of claim 1, wherein the camera device comprises one of a dynamic vision sensor camera, an event camera, and a retinal imitation camera.

10. The system of claim 1, comprising:

and the storage module is used for storing the sample data.