Machine learning model training method, machine learning model data processing method, equipment and medium

Abstract

The embodiment of the invention discloses a machine learning model training method and device, a data processing method and device and a medium. The method comprises the steps of obtaining rail transit scene data; training a machine learning model to be trained according to first rail transit scene data and second rail transit scene data in the rail transit scene data; wherein the first rail transit scene data has a data label, and the second rail transit scene data does not have a data label.

Description

Technical Field

[0001] The present disclosure relates to the field of rail transit technology, and more specifically, to a machine learning model training, data processing method, device and medium. Background Art

[0002] In the rail transit scene perception system, many types of sensors are required, and these devices need the support of artificial intelligence algorithms to perform well.

[0003] In related technologies, machine learning algorithms based on artificial intelligence are mostly modeled by learning from large amounts of data. During the modeling process, a large amount of collected data needs to be labeled. However, this method of training the machine learning model only after all the collected data has been labeled not only consumes a lot of manpower and material resources, but also takes a long time. Summary of the invention

[0004] The disclosed embodiments provide a new technical solution for training a machine learning model.

[0005] According to a first aspect of the present disclosure, a method for training a machine learning model is provided, the method comprising:

[0006] Obtain rail transit scene data;

[0007] Training the machine learning model to be trained according to the first rail transit scene data and the second rail transit scene data in the rail transit scene data;

[0008] Among them, the first rail transit scene data has a data label, and the second rail transit scene data does not have a data label.

[0009] Optionally, obtain rail transit scene data, including:

[0010] Acquire original rail transit scene data; wherein the original rail transit scene data involves at least one of the following: text data in the rail transit scene, video data in the rail transit scene, image data in the rail transit scene, and audio data in the rail transit scene;

[0011] The original rail transit scene data is cleaned to obtain the rail transit scene data; wherein the data cleaning involves at least one of the following: missing value processing, noise data processing, and duplicate data processing.

[0012] Optionally, before training the machine learning model to be trained according to the first rail transit scene data and the second rail transit scene data in the rail transit scene data, the method further includes:

[0013] Data labeling is performed on the first rail transit scene data in the rail transit scene data so that the first rail transit scene data has the data label.

[0014] Optionally, the performing data labeling on the first rail transit scene data in the rail transit scene data so that the first rail transit scene data has the data label, includes:

[0015] According to the type of machine learning task, the rail transit scene data is divided into multiple types of rail transit scene data; wherein different types of machine learning tasks correspond to different types of machine learning models;

[0016] Data labeling is performed on the first rail transit scene data in each type of the rail transit scene data so that the first rail transit scene data in each type of the rail transit scene data has the data label.

[0017] Optionally, the training of the machine learning model to be trained according to the first rail transit scene data and the second rail transit scene data in the rail transit scene data includes:

[0018] According to the first rail transit scene data in each category of the rail transit scene data and the second rail transit scene data in each category of the rail transit scene data, the corresponding machine learning model to be trained is trained.

[0019] Optionally, the training of the machine learning model to be trained according to the first rail transit scene data and the second rail transit scene data in the rail transit scene data includes:

[0020] According to the first rail transit scene data in each category of the rail transit scene data and the second rail transit scene data in each category of the rail transit scene data, the corresponding machine learning model to be trained is trained.

[0021] Optionally, the data annotation involves at least one of the following: box annotation, semantic segmentation, and point cloud annotation.

[0022] According to a second aspect of the present disclosure, a data processing method is provided, the method comprising:

[0023] Obtain rail transit scene data to be processed;

[0024] Processing the to-be-processed rail transit scene data according to the trained machine learning model;

[0025] Among them, the machine learning model is trained according to the training method of the machine learning model described in the first aspect above.

[0026] Optionally, the processing of the to-be-processed rail transit scene data according to the trained machine learning model includes:

[0027] According to the trained machine learning model, the rail transit scene data to be processed is marked with a frame to obtain the posture information of pedestrians and / or vehicles in the rail transit scene data to be processed; or,

[0028] According to the trained machine learning model, semantic segmentation is performed on the rail transit scene data to be processed to obtain semantic information of the drivable area in the rail transit scene data to be processed; or

[0029] According to the trained machine learning model, point cloud annotation is performed on the rail transit scene data to be processed to obtain the track area and / or pedestrian information on the track area in the rail transit scene data to be processed.

[0030] According to a third aspect of the present disclosure, an electronic device is provided, comprising a memory and a processor, wherein the memory is used to store an executable computer program; the computer program is used to control the processor to execute the method according to the first aspect or the second aspect above.

[0031] According to a fourth aspect of the present disclosure, a computer-readable storage medium is provided, on which a computer program is stored. When the computer program is executed by a processor, the method described in the first aspect or the second aspect is implemented.

[0032] One beneficial effect of the disclosed embodiment is that after acquiring the rail transit scene data, the machine learning model is trained based on the first rail transit scene data with data labels in the rail transit scene data and the second rail transit scene data without data labels in the rail transit scene data. That is, in the rail transit scene, the present embodiment can use part of the rail transit scene data with data labels and part of the rail transit scene data without labels to train the machine learning model, so that the effect of model training with part of the labeled data in the rail transit scene data is similar to that of all the labeled data of the rail transit scene data, which reduces manpower, material resources and time.

[0033] Other features and advantages of the present disclosure will become apparent from the following detailed description of exemplary embodiments of the present disclosure with reference to the accompanying drawings. BRIEF DESCRIPTION OF THE DRAWINGS

[0034] The accompanying drawings, which are incorporated in and constitute a part of the specification, illustrate embodiments of the present disclosure and, together with the description, serve to explain the principles of the present disclosure.

[0035] Figure 1 is a schematic diagram of the hardware configuration of an electronic device according to an embodiment of the present disclosure;

[0036] Figure 2 is a flowchart of a method for training a machine learning model according to an embodiment of the present disclosure;

[0037] Figure 3 is a flowchart of a method for training a machine learning model according to an example of the present disclosure;

[0038] Figure 4 is a flowchart of a data processing method according to an embodiment of the present disclosure;

[0039] Figure 5 is a principle block diagram of a training device for a machine learning model according to an embodiment of the present disclosure;

[0040] Figure 6 is a principle block diagram of a data processing device according to an embodiment of the present disclosure;

[0041] Figure 7 is a schematic diagram of the hardware configuration of an electronic device according to another embodiment of the present disclosure. DETAILED DESCRIPTION

[0042] Various exemplary embodiments of the present disclosure will now be described in detail with reference to the accompanying drawings. It should be noted that the relative arrangement of components and steps, numerical expressions and numerical values ​​set forth in these embodiments do not limit the scope of the present disclosure unless otherwise specifically stated.

[0043] The following description of at least one exemplary embodiment is merely illustrative in nature and is in no way intended to limit the present disclosure, its application, or uses.

[0044] Technologies, methods, and equipment known to ordinary technicians in the relevant art may not be discussed in detail, but where appropriate, the technologies, methods, and equipment should be considered as part of the specification.

[0045] In all examples shown and discussed herein, any specific values ​​should be interpreted as merely exemplary and not as limiting. Therefore, other examples of the exemplary embodiments may have different values.

[0046] It should be noted that like reference numerals and letters refer to similar items in the following figures, and therefore, once an item is defined in one figure, it need not be further discussed in subsequent figures.

[0047] <Hardware Configuration>

[0048] Figure 1 A block diagram showing a hardware configuration of an electronic device 1000 that can implement an embodiment of the present disclosure is shown.

[0049] The electronic device 1000 may be a portable computer, a desktop computer, a mobile phone, a tablet computer, etc. Figure 1 As shown, the electronic device 1000 may include a processor 1100, a memory 1200, an interface device 1300, a communication device 1400, a display device 1500, an input device 1600, a speaker 1700, a microphone 1800, and the like. Among them, the processor 1100 may be a central processing unit CPU, a microprocessor MCU, and the like. The memory 1200 includes, for example, a ROM (read-only memory), a RAM (random access memory), a non-volatile memory such as a hard disk, and the like. The interface device 1300 includes, for example, a USB interface, a headphone interface, and the like. The communication device 1400 is, for example, capable of wired or wireless communication, and may specifically include Wifi communication, Bluetooth communication, 2G / 3G / 4G / 5G communication, and the like. The display device 1500 is, for example, a liquid crystal display screen, a touch display screen, and the like. The input device 1600 may include, for example, a touch screen, a keyboard, a somatosensory input, and the like. The user may input / output voice information through the speaker 1700 and the microphone 1800.

[0050] Figure 1 The electronic device shown is merely illustrative and does not in any way imply any limitation on the present disclosure, its application or use. In the embodiments of the present disclosure, the memory 1200 of the electronic device 1000 is used to store instructions, and the instructions are used to control the processor 1100 to operate to perform any one of the machine learning model training methods or data processing methods provided in the embodiments of the present disclosure. It should be understood by those skilled in the art that although Figure 1 In the electronic device 1000, multiple devices are shown, but the present disclosure may only involve some of the devices, for example, the electronic device 1000 only involves the processor 1100 and the storage device 1200. A technician can design instructions according to the scheme disclosed in the present disclosure. How instructions control the processor to operate is well known in the art, so it will not be described in detail here.

[0051] <Method Example 1>

[0052] The present disclosure discloses a method for training a machine learning model. Figure 2 As shown, the training method of the machine learning model includes the following steps S2100 to S2200.

[0053] Step S2100, obtaining rail transit scene data.

[0054] Among them, the rail transit scene data can be obtained through the original rail transit scene data collected by sensors in the rail transit scene perception system, and the sensors include, for example but not limited to, cameras, point clouds and millimeter wave radars.

[0055] In an optional embodiment, the step S2100 of acquiring rail transit scene data may further include the following steps S2110 to S2120:

[0056] Step S2110, obtaining original rail transit scene data.

[0057] Among them, the original rail transit scene data involves at least one of the following: text data in the rail transit scene, video data in the rail transit scene, image data in the rail transit scene, and audio data in the rail transit scene.

[0058] Optionally, the step S2110 of acquiring the original rail transit scene data may further include: providing a data input interface in response to a request for acquiring the rail transit scene data; acquiring the rail transit scene data input through the data input interface as the original rail transit scene data.

[0059] The data input interface may be any entry for a user to input information, such as an input box, a drop-down list, a voice input entry, etc., and is not limited here.

[0060] It should be noted that after the sensors in the rail transit scene perception system collect the original rail transit scene data, the collected original rail transit scene data will be stored in the electronic device. When it is necessary to train the machine learning model in the rail transit scene, the user can initiate a request to obtain the rail transit scene data. In response to the acquisition request, the electronic device provides a data input interface and obtains the rail transit scene data input through the data input interface as the original rail transit scene data.

[0061] Step S2120, performing data cleaning on the original rail transit scene data to obtain the rail transit scene data.

[0062] The data cleaning involves at least one of the following: missing value processing, noise data processing, and duplicate data processing.

[0063] In this step S2120, since the collected original rail transit scene data is usually unstructured, it is usually necessary to perform data cleaning on the collected original rail transit scene data, such as performing missing value processing, noise data processing, and duplicate data processing on the collected original rail transit scene data, so that the original rail transit scene data can be cleaned into data suitable for labeling. In order to distinguish the original rail transit scene data from the cleaned data, the cleaned data is called rail transit scene data.

[0064] According to the above steps S2110 to S2120, the original rail transit scene data will be cleaned to obtain rail transit scene data for training the machine learning model. Such processing can improve the effect of the machine learning model trained thereby to a certain extent.

[0065] Step S2200: training the machine learning model to be trained according to the first rail transit scene data and the second rail transit scene data in the rail transit scene data.

[0066] Among them, the first rail transit scene data has a data label, and the second rail transit scene data does not have a data label. This can also be understood as that the first rail transit scene data in the rail transit scene data is part of the rail transit scene information after manual data annotation, and the second rail transit scene data in the rail transit scene data is another part of the rail transit scene information that has not been manually labeled. In this way, it is possible to use part of the rail transit scene data with data annotation and part of the rail transit scene data without annotation to train the machine learning model, so that the effect of model training with part of the annotated data in the rail transit scene data is similar to that of all the annotated data of the rail transit scene data, reducing manpower, material resources and time. The training method of the machine learning model of this embodiment can be understood as semi-supervised learning.

[0067] Furthermore, the first rail transit scene data is less than the second rail transit scene data. In this way, a small amount of manually annotated data and a large amount of unlabeled data can be used to train the machine learning model, and the effect of using less labeled data can be achieved similar to that of using a large amount of labeled data for model training.

[0068] In this embodiment, after obtaining the first rail transit scene data with data labels and the second rail transit scene data without data labels, a semi-supervised training method can be used to train the machine learning model to be trained using the first rail transit scene data and the second rail transit scene data.

[0069] It should be noted that the semi-supervised training method mentioned in this embodiment is self-training in the semi-supervised training method, that is, pseudo-labeling technology. The basic idea is to first train on the labeled data, then predict the unlabeled data, take the one with the highest prediction confidence and directly define the label for it, and then include such samples in the current training samples to continue training until the prediction results of the model no longer change.

[0070] In an optional embodiment, before executing step 2200 to train the machine learning model to be trained according to the first rail transit scene data and the second rail transit scene data in the rail transit scene data, the training method of the machine learning model of the embodiment of the present disclosure further includes step S3100: data labeling the first rail transit scene data in the rail transit scene data so that the first rail transit scene data has the data label.

[0071] Optionally, this step S3100 performs data labeling on the first rail transit scene data in the rail transit scene data so that the first rail transit scene data has the data label, which may further include the following steps S3110 to S3120:

[0072] Step S3110: divide the rail transit scene data into multiple categories of rail transit scene data according to the type of machine learning task.

[0073] Among them, different types of machine learning tasks correspond to different types of machine learning models.

[0074] For example, if the machine learning task is an obstacle detection task, then the machine learning model is an obstacle detection model.

[0075] For another example, if the machine learning task is a semantic segmentation task, then the machine learning model is a semantic segmentation model.

[0076] For another example, if the machine learning task is a track area recognition task, then the machine learning model is a track area recognition model.

[0077] For example, if there are three machine learning tasks, such as obstacle detection task, semantic segmentation task and track area identification task, it can be determined that the rail transit scene data needs to be divided into a first type of rail transit scene data for obstacle detection task, a second type of rail transit scene data for semantic segmentation task, and a third type of rail transit scene data for track area identification task.

[0078] Step S3120: perform data labeling on the first rail transit scene data in each type of the rail transit scene data, so that the first rail transit scene data in each type of the rail transit scene data has the data label.

[0079] The data annotation involves at least one of the following: box annotation, semantic segmentation, and point cloud annotation.

[0080] Among them, the box annotation can mark the designated target object in the image or video in a simple and quick way, such as with a 2D, 3D, or polygonal box. In this embodiment, by performing box annotation on the rail transit scene data, the position information of the vehicle and pedestrians can be output in the drivable scene picture.

[0081] Among them, semantic segmentation can divide complex and irregular images into regions according to the attributes of objects, mark the content and location in the image, perform pixel-level segmentation according to the attributes, and colorize. In this embodiment, by performing semantic segmentation on rail transit scene data, the semantic information of the drivable area can be output in the drivable scene image.

[0082] Point cloud annotation is one of the ways to express three-dimensional data. It can collect various obstacles and their position coordinates through sensors such as laser radar, and the annotator needs to classify these dense point clouds and annotate them with different attributes. In this embodiment, by annotating the rail transit scene data with point clouds, the information of pedestrians on the track area and on the track area can be obtained.

[0083] Continuing with the above example, for the first category of rail transit scene data, the first rail transit scene data in this category of rail transit scene data may be marked with a box, and the second rail transit scene data in this category of rail transit scene data may not be marked with a box, so that the first rail transit scene in this category of rail transit scene data has a data label. For another example, for the second category of rail transit scene data, semantic segmentation may be performed on the first rail transit scene data in this category of rail transit scene data, and semantic segmentation may not be performed on the second rail transit scene data in this category of rail transit scene data, so that the first rail transit scene in this category of rail transit scene data has a data label. For another example, for the third category of rail transit scene data, point cloud annotation may be performed on the first rail transit scene data in this category of rail transit scene data, and point cloud annotation may not be performed on the second rail transit scene data in this category of rail transit scene data, so that the first rail transit scene in this category of rail transit scene data has a data label.

[0084] According to the above steps S3110 to S3120, the rail transit scene data to be labeled will be divided into different types of rail transit scene data according to different machine learning tasks, and the different types of rail transit scene data will be labeled separately to train the machine learning model corresponding to the corresponding machine learning task.

[0085] In this embodiment, this step S2200 trains the machine learning model to be trained according to the first rail transit scene data and the second rail transit scene data in the rail transit scene data, specifically including: training the corresponding machine learning model to be trained according to the first rail transit scene data in each category of the rail transit scene data and the second rail transit scene data in each category of the rail transit scene data.

[0086] Specifically, after the first rail transit scene data in different types of rail transit scene data are respectively labeled according to the above step S3100, the corresponding type of machine learning model to be trained can be trained based on the second rail transit scene data without data labeling in each type of rail transit scene data and the first rail transit scene data with data labeling according to step S2200. This can also be understood as the training of the corresponding type of machine learning model can be carried out in a semi-supervised training manner.

[0087] For example, the obstacle detection model may be trained based on the first rail transit scene data and the second rail transit scene data in the first category of rail transit scene data.

[0088] For another example, a semantic segmentation model may be trained based on the first rail transit scene data and the second rail transit scene data in the second category of rail transit scene data.

[0089] For another example, the track area recognition model may be trained based on the first rail transit scene data and the second rail transit scene data in the third category of rail transit scene data.

[0090] According to the method of the embodiment of the present disclosure, after acquiring the rail transit scene data, the machine learning model will be trained based on the first rail transit scene data with data labels in the rail transit scene data and the second rail transit scene data without data labels in the rail transit scene data. That is, in the rail transit scene, this embodiment can use part of the rail transit scene data with data labels and part of the rail transit scene data without labels to train the machine learning model, so that the effect of model training with part of the labeled data in the rail transit scene data is close to that of all the labeled data of the rail transit scene data, which reduces manpower, material resources and time.

[0091] In one embodiment, after executing the above step S3100 to perform data labeling on the first rail transit scene data in the rail transit scene data so that the first rail transit scene data has the data label, before executing the above step S2200 to train the machine learning model to be trained according to the first rail transit scene data and the second rail transit scene data in the rail transit scene data, the training method of the machine learning model in the embodiment of the present disclosure further includes the following steps S4100 to S4200:

[0092] Step S4100: According to the set audit items, the first rail transit scene data is audited to obtain the audit results.

[0093] Among them, setting the review items may include whether the data labels of the first rail transit scene data are standardized.

[0094] In this embodiment, when the data label of the first rail transit scene data is standardized, an audit result of passing the audit is obtained, and when the data label of the first rail transit scene data is not standardized, an audit result of failing the audit is obtained.

[0095] For example, for the first rail transit scene data in the first category of rail transit scene data, it can be reviewed whether the pull-box marking in the first rail transit scene data is standardized.

[0096] For another example, for the first rail transit scene data in the second category of rail transit scene data, it can be examined whether the semantic segmentation in the first rail transit scene data is standardized.

[0097] For another example, for the first rail transit scene data in the third category of rail transit scene data, it can be reviewed whether the point cloud annotation in the first rail transit scene data is standardized.

[0098] Step S4200, when the audit result indicates that the first rail transit scene data is approved, the machine learning model to be trained is trained based on the first rail transit scene data and the second rail transit scene data in the rail transit scene data.

[0099] In this embodiment, training of the machine learning model to be trained will be performed based on the first rail transit scene data and the second rail transit scene data in the rail transit scene data only when the audit result indicates that the first rail transit scene data has passed the audit.

[0100] For example, only when the first rail transit scene data in the first category of rail transit scene data is standardized, the obstacle detection model will be trained according to the first rail transit scene data and the second rail transit scene data in the first category of rail transit scene data.

[0101] For another example, only when the first rail transit scene data in the second category of rail transit scene data is standardized, the semantic segmentation model will be trained based on the first rail transit scene data and the second rail transit scene data in the first category of rail transit scene data.

[0102] For another example, only when the first rail transit scene data in the third category of rail transit scene data is standardized, the track area recognition model will be trained according to the first rail transit scene data and the second rail transit scene data in the first category of rail transit scene data.

[0103] According to this embodiment, the first rail transit scene data in the rail transit scene data will be reviewed, and the machine learning model will be trained only after the review is passed to ensure the labeling quality of the first rail transit scene data.

[0104] <Example>

[0105] Next, we will show an example of a machine learning model training method. In this example, refer to Figure 3 , the training method of the machine learning model includes:

[0106] Step S301, obtaining text data in a rail transit scene, video data in a rail transit scene, image data in a rail transit scene, and audio data in a rail transit scene.

[0107] Step S302, performing data cleaning on the acquired rail transit scene data to obtain the cleaned rail transit scene data as training data, for example, performing missing value processing, noise data processing, and duplicate data processing.

[0108] Step S303, according to different requirements, the training data is divided into different training data packets, and for different training data packets, only a small amount of training data in the training data packets is labeled to obtain training samples corresponding to different training data packets.

[0109] Step S304: training the corresponding machine learning model for the corresponding training samples.

[0110] <Method Example 2>

[0111] The present disclosure embodiment discloses a data processing method, such as Figure 4As shown, the data processing method includes the following steps S5100 to S5200.

[0112] Step S5100, obtaining rail transit scene data to be processed.

[0113] Among them, the rail transit scene data to be processed includes, for example but not limited to, text data in the rail transit scene, video data in the rail transit scene, image data in the rail transit scene, and audio data in the rail transit scene.

[0114] Step S5200: Process the rail transit scene data to be processed according to the trained machine learning model.

[0115] Among them, the machine learning model is trained according to the training method of the machine learning model of any one of the above method embodiments.

[0116] In an optional embodiment, the step S5200 of processing the to-be-processed rail transit scene data according to the trained machine learning model may further include one of the following:

[0117] Item 1: Based on the trained machine learning model, the rail transit scene data to be processed is marked with a frame to obtain the posture information of pedestrians and / or vehicles in the rail transit scene data to be processed.

[0118] The second item: According to the trained machine learning model, semantic segmentation is performed on the rail transit scene data to be processed to obtain semantic information of the drivable area in the rail transit scene data to be processed.

[0119] The third item: Based on the trained machine learning model, point cloud annotation is performed on the rail transit scene data to be processed to obtain the track area and / or pedestrian information on the track area in the rail transit scene data to be processed.

[0120] For example, if the machine learning model is an obstacle detection model, the rail transit scene data to be processed can be marked with boxes according to the trained obstacle detection model to obtain the posture information of pedestrians and / or vehicles in the rail transit scene data to be processed.

[0121] For another example, if the machine learning model is a semantic segmentation model, the rail transit scene data to be processed can be semantically segmented according to the trained semantic segmentation model to obtain the semantic information of the drivable area in the image data of the rail transit scene.

[0122] For another example, if the machine learning model is a track area recognition model, point cloud annotation can be performed on the rail transit scene data to be processed according to the trained track area recognition model to obtain the track area and / or pedestrian information on the track area in the rail transit scene data to be processed.

[0123] <Device Example 1>

[0124] In this embodiment, a training device 500 for a machine learning model is provided, such as Figure 5 As shown, the training device 500 of the machine learning model may include a first acquisition module 510 and a training module 520.

[0125] A first acquisition module 510 is used to acquire rail transit scene data;

[0126] A training module 520, configured to train a machine learning model to be trained according to the first rail transit scene data and the second rail transit scene data in the rail transit scene data;

[0127] Among them, the first rail transit scene data has a data label, and the second rail transit scene data does not have a data label.

[0128] In one embodiment, the first acquisition module 510 is specifically used to: acquire original rail transit scene data; wherein the original rail transit scene data involves at least one of the following: text data in the rail transit scene, video data in the rail transit scene, image data in the rail transit scene, and audio data in the rail transit scene; perform data cleaning on the original rail transit scene data to obtain the rail transit scene data; wherein the data cleaning involves at least one of the following: missing value processing, noise data processing, and duplicate data processing.

[0129] In one embodiment, the first acquisition module 510 is specifically used to: provide a data input interface in response to an acquisition request for obtaining rail transit scene data; and acquire the rail transit scene data input through the data input interface as the original rail transit scene data.

[0130] In one embodiment, the training module 520 is further used to: perform data labeling on the first rail transit scene data in the rail transit scene data so that the first rail transit scene data has the data label.

[0131] In one embodiment, the training module 520 is also used to: divide the rail transit scene data into multiple categories of rail transit scene data according to the type of machine learning task; wherein different types of machine learning tasks correspond to different types of machine learning models; and perform data labeling on the first rail transit scene data in each category of the rail transit scene data so that the first rail transit scene data in each category of the rail transit scene data has the data label.

[0132] In one embodiment, the training module 520 is specifically used to: train the corresponding machine learning model to be trained according to the first rail transit scene data in each category of the rail transit scene data and the second rail transit scene data in each category of the rail transit scene data.

[0133] In one embodiment, the training module 520 is also used to: audit the first rail transit scene data according to the set audit items to obtain an audit result; when the audit result indicates that the first rail transit scene data has passed the audit, train the machine learning model to be trained based on the first rail transit scene data and the second rail transit scene data in the rail transit scene data.

[0134] In one embodiment, the data annotation involves at least one of the following: box annotation, semantic segmentation, and point cloud annotation.

[0135] According to this embodiment, after acquiring the rail transit scene data, it will train the machine learning model based on the first rail transit scene data with data labels in the rail transit scene data and the second rail transit scene data without data labels in the rail transit scene data, and the first rail transit scene data is less than the second rail transit scene data. That is, in the rail transit scene, it can use part of the rail transit scene data with data labels and part of the rail transit scene data without labels to train the machine learning model, so that the part of the labeled data in the rail transit scene data can achieve the effect of model training similar to all the labeled data of the rail transit scene data, reducing manpower, material resources and time.

[0136] <Device Example 2>

[0137] In this embodiment, a data processing device 600 is provided. Figure 6 As shown, the data processing device 600 may include a second acquisition module 610 and a processing module 620 .

[0138] The second acquisition module 610 is used to acquire the video to be classified and the text description information corresponding to the video to be classified in different categories;

[0139] The processing module 620 is used to obtain the rail transit scene data to be processed;

[0140] A processing module, used for processing the rail transit scene data to be processed according to the trained machine learning model;

[0141] Among them, the machine learning model is trained according to the training method of the machine learning model of any one of the above method embodiments.

[0142] In one embodiment, the processing module 620 is specifically used to: perform frame marking on the rail transit scene data to be processed according to the trained machine learning model to obtain the posture information of pedestrians and / or vehicles in the rail transit scene data to be processed; or, perform semantic segmentation on the rail transit scene data to be processed according to the trained machine learning model to obtain the semantic information of the drivable area in the rail transit scene data to be processed; or, perform point cloud marking on the rail transit scene data to be processed according to the trained machine learning model to obtain the track area and / or pedestrian information on the track area in the rail transit scene data to be processed.

[0143] <Equipment Embodiment>

[0144] In this embodiment, an electronic device 700 is also provided. Figure 7 As shown, it includes a memory 720 and a processor 710 .

[0145] The memory 720 is used to store an executable computer program; the computer program is used to control the processor 710 to execute any one of the machine learning model training methods provided in the above method embodiment 1, or any one of the data processing methods provided in the above method embodiment 2.

[0146] <Computer Readable Storage Medium>

[0147] In this embodiment, a computer-readable storage medium is further provided, on which a computer program is stored. When the computer program is executed by a processor, the method of any embodiment of the present disclosure is implemented.

[0148] The present disclosure may be a system, a method and / or a computer program product. The computer program product may include a computer-readable storage medium carrying computer-readable program instructions for causing a processor to implement various aspects of the present disclosure.

[0149] A computer-readable storage medium may be a tangible device that can hold and store instructions used by an instruction execution device. A computer-readable storage medium may be, for example, but not limited to, an electrical storage device, a magnetic storage device, an optical storage device, an electromagnetic storage device, a semiconductor storage device, or any suitable combination of the foregoing. More specific examples of computer-readable storage media (a non-exhaustive list) include: a portable computer disk, a hard disk, a random access memory (RAM), a read-only memory (ROM), an erasable programmable read-only memory (EPROM or flash memory), a static random access memory (SRAM), a portable compact disk read-only memory (CD-ROM), a digital versatile disk (DVD), a memory stick, a floppy disk, a mechanical encoding device, such as a punch card or a raised structure in a groove on which instructions are stored, and any suitable combination of the foregoing. As used herein, a computer-readable storage medium is not to be interpreted as a transient signal per se, such as a radio wave or other freely propagating electromagnetic wave, an electromagnetic wave propagating through a waveguide or other transmission medium (e.g., a light pulse through a fiber optic cable), or an electrical signal transmitted through a wire.

[0150] The computer-readable program instructions described herein can be downloaded from a computer-readable storage medium to each computing / processing device, or downloaded to an external computer or external storage device via a network, such as the Internet, a local area network, a wide area network, and / or a wireless network. The network can include copper transmission cables, optical fiber transmissions, wireless transmissions, routers, firewalls, switches, gateway computers, and / or edge servers. The network adapter card or network interface in each computing / processing device receives the computer-readable program instructions from the network and forwards the computer-readable program instructions for storage in the computer-readable storage medium in each computing / processing device.

[0151] The computer program instructions for performing the operation of the present disclosure may be assembly instructions, instruction set architecture (ISA) instructions, machine instructions, machine-related instructions, microcode, firmware instructions, state setting data, or source code or object code written in any combination of one or more programming languages, including object-oriented programming languages, such as Smalltalk, C++, etc., and conventional procedural programming languages, such as "C" language or similar programming languages. Computer-readable program instructions may be executed completely on a user's computer, partially on a user's computer, as an independent software package, partially on a user's computer, partially on a remote computer, or completely on a remote computer or server. In the case of a remote computer, the remote computer may be connected to the user's computer via any type of network, including a local area network (LAN) or a wide area network (WAN), or may be connected to an external computer (e.g., using an Internet service provider to connect via the Internet). In some embodiments, an electronic circuit, such as a programmable logic circuit, a field programmable gate array (FPGA), or a programmable logic array (PLA), may be customized by utilizing the state information of the computer-readable program instructions, and the electronic circuit may execute the computer-readable program instructions, thereby realizing various aspects of the present disclosure.

[0152] Various aspects of the present disclosure are described herein with reference to the flowcharts and / or block diagrams of the methods, devices (systems) and computer program products according to the embodiments of the present disclosure. It should be understood that each box in the flowchart and / or block diagram and the combination of each box in the flowchart and / or block diagram can be implemented by computer-readable program instructions.

[0153] These computer-readable program instructions can be provided to a processor of a general-purpose computer, a special-purpose computer, or other programmable data processing device, thereby producing a machine, so that when these instructions are executed by the processor of the computer or other programmable data processing device, a device that implements the functions / actions specified in one or more boxes in the flowchart and / or block diagram is generated. These computer-readable program instructions can also be stored in a computer-readable storage medium, and these instructions cause the computer, programmable data processing device, and / or other equipment to work in a specific manner, so that the computer-readable medium storing the instructions includes a manufactured product, which includes instructions for implementing various aspects of the functions / actions specified in one or more boxes in the flowchart and / or block diagram.

[0154] Computer-readable program instructions may also be loaded onto a computer, other programmable data processing apparatus, or other device so that a series of operating steps are performed on the computer, other programmable data processing apparatus, or other device to produce a computer-implemented process, thereby causing the instructions executed on the computer, other programmable data processing apparatus, or other device to implement the functions / actions specified in one or more boxes in the flowchart and / or block diagram.

[0155] The flowcharts and block diagrams in the accompanying drawings show the possible architecture, functions and operations of the systems, methods and computer program products according to multiple embodiments of the present disclosure. In this regard, each box in the flowchart or block diagram can represent a module, a program segment or a part of an instruction, and the module, a program segment or a part of an instruction contains one or more executable instructions for realizing the specified logical function. In some alternative implementations, the functions marked in the box can also occur in a different order from the order marked in the accompanying drawings. For example, two consecutive boxes can actually be executed substantially in parallel, and they can sometimes be executed in the opposite order, depending on the functions involved. It should also be noted that each box in the block diagram and / or the flowchart, and the combination of the boxes in the block diagram and / or the flowchart can be implemented by a dedicated hardware-based system that performs the specified function or action, or can be implemented by a combination of dedicated hardware and computer instructions. It is well known to those skilled in the art that it is equivalent to implement it by hardware, implement it by software, and implement it by combining software and hardware.

[0156] The embodiments of the present disclosure have been described above, and the above description is exemplary, not exhaustive, and is not limited to the disclosed embodiments. Many modifications and changes will be apparent to those of ordinary skill in the art without departing from the scope and spirit of the described embodiments. The choice of terms used herein is intended to best explain the principles of the embodiments, practical applications, or technical improvements in the marketplace, or to enable other persons of ordinary skill in the art to understand the embodiments disclosed herein. The scope of the present disclosure is defined by the appended claims.

Claims

1. A training method for a machine learning model, characterized in that: The method comprises: Obtain rail transit scene data; Training the machine learning model to be trained according to the first rail transit scene data and the second rail transit scene data in the rail transit scene data; Among them, the first rail transit scene data has a data label, and the second rail transit scene data does not have a data label.

2. The method according to claim 1, characterized in that Obtain rail transit scene data, including: Acquire original rail transit scene data; wherein the original rail transit scene data involves at least one of the following: text data in the rail transit scene, video data in the rail transit scene, image data in the rail transit scene, and audio data in the rail transit scene; The original rail transit scene data is cleaned to obtain the rail transit scene data; wherein the data cleaning involves at least one of the following: missing value processing, noise data processing, and duplicate data processing.

3. The method according to claim 2, characterized in that The obtaining of original rail transit scene data includes: In response to a request for obtaining rail transit scene data, a data input interface is provided; The rail transit scene data input through the data input interface is acquired as the original rail transit scene data.

4. The method according to claim 1, characterized in that: Before training the machine learning model to be trained according to the first rail transit scene data and the second rail transit scene data in the rail transit scene data, the method further includes: Data labeling is performed on the first rail transit scene data in the rail transit scene data so that the first rail transit scene data has the data label.

5. The method according to claim 4, characterized in that The step of labeling the first rail transit scene data in the rail transit scene data so that the first rail transit scene data has the data label includes: According to the type of machine learning task, the rail transit scene data is divided into multiple types of rail transit scene data; wherein different types of machine learning tasks correspond to different types of machine learning models; Data labeling is performed on the first rail transit scene data in each type of the rail transit scene data so that the first rail transit scene data in each type of the rail transit scene data has the data label.

6. The method according to claim 5, characterized in that The training of the machine learning model to be trained according to the first rail transit scene data and the second rail transit scene data in the rail transit scene data includes: According to the first rail transit scene data in each category of the rail transit scene data and the second rail transit scene data in each category of the rail transit scene data, the corresponding machine learning model to be trained is trained.

7. The method according to claim 4, characterized in that After the data is labeled for the first rail transit scene data in the rail transit scene data, the method further includes: According to the set audit items, the first rail transit scenario data is audited to obtain the audit result; When the audit result indicates that the first rail transit scenario data has passed the audit, the machine learning model to be trained is trained based on the first rail transit scenario data and the second rail transit scenario data in the rail transit scenario data.

8. The method according to claim 4, characterized in that The data annotation involves at least one of the following: box annotation, semantic segmentation, and point cloud annotation.

9. A data processing method, characterized in that: The method comprises: Obtain rail transit scene data to be processed; Processing the to-be-processed rail transit scene data according to the trained machine learning model; Wherein, the machine learning model is trained according to the training method of the machine learning model according to any one of claims 1-8.

10. The method according to claim 9, characterized in that The processing of the to-be-processed rail transit scene data according to the trained machine learning model includes: According to the trained machine learning model, the rail transit scene data to be processed is marked with a frame to obtain the posture information of pedestrians and / or vehicles in the rail transit scene data to be processed; or, According to the trained machine learning model, semantic segmentation is performed on the rail transit scene data to be processed to obtain semantic information of the drivable area in the rail transit scene data to be processed; or According to the trained machine learning model, point cloud annotation is performed on the rail transit scene data to be processed to obtain the track area and / or pedestrian information on the track area in the rail transit scene data to be processed.

11. An electronic device, characterized in that: The method comprises a memory and a processor, wherein the memory is used to store an executable computer program; and the computer program is used to control the processor to execute the method according to any one of claims 1 to 10.

12. A computer-readable storage medium, characterized in that: A computer program is stored thereon, and when the computer program is executed by a processor, the method according to any one of claims 1 to 10 is implemented.

Images