CN114820679B

CN114820679B - Image labeling method and device electronic device and storage medium

Info

Publication number: CN114820679B
Application number: CN202210768430.2A
Authority: CN
Inventors: 柴亚捷; 张亚森; 时爱君
Original assignee: Beijing Xiaomi Pinecone Electronic Co Ltd; Xiaomi Automobile Technology Co Ltd
Current assignee: Beijing Xiaomi Pinecone Electronic Co Ltd; Xiaomi Automobile Technology Co Ltd
Priority date: 2022-07-01
Filing date: 2022-07-01
Publication date: 2023-04-18
Anticipated expiration: 2042-07-01
Also published as: CN114820679A

Abstract

The present disclosure provides an image labeling method, an image labeling apparatus, an electronic device, and a storage medium, where the method includes: the method comprises the steps of obtaining a slice image of an object according to an image to be annotated, obtaining a reference annotation region of the object in the image to be annotated according to the slice image, determining region adjustment information of the reference annotation region according to the slice image, and adjusting the reference annotation region according to the region adjustment information to obtain a target annotation region, wherein the target annotation region is used for annotating the image to be annotated, so that the automation degree of the image annotation process can be effectively improved, the time cost of image annotation is reduced, and the image annotation efficiency is effectively improved.

Description

Image annotation method and device, electronic equipment and storage medium

Technical Field

The disclosure relates to the technical field of automatic driving, in particular to an image annotation method, an image annotation device, electronic equipment and a storage medium.

Background

With the development of artificial intelligence technology and automobile manufacturing industry and the beauty of people for the future life intellectualization, the automatic driving method based on the artificial intelligence technology has become a research hotspot in academic and industrial fields. In the current big data era, the development of various intelligent technologies cannot leave the support of a large amount of data, and the automatic driving technology is in the initial development stage, so that the demand on the large amount of data is always greater.

In the related art, when the marking task of the image data in each road scene is performed, the degree of dependence on manual marking is high, so that the marking efficiency is low, the cost is high, and the method is not suitable for the image marking task with large data volume.

Disclosure of Invention

The present disclosure is directed to solving, at least in part, one of the technical problems in the related art.

Therefore, the present disclosure aims to provide an image annotation method, an image annotation device, an electronic device, and a storage medium, which can effectively improve the automation degree of an image annotation process, reduce the time cost of image annotation, and thereby effectively improve the image annotation efficiency.

An embodiment of the first aspect of the present disclosure provides an image annotation method, including: acquiring a slice image of an object according to an image to be marked; acquiring a reference marking area of an object in the image to be marked according to the slice image; determining the region adjustment information of the reference labeling region according to the slice image; and adjusting the reference labeling area according to the area adjustment information to obtain a target labeling area, wherein the target labeling area is used for labeling the image to be labeled.

According to the image annotation method provided by the embodiment of the first aspect of the disclosure, a slice image of an object is acquired according to an image to be annotated, a reference annotation region of the object in the image to be annotated is acquired according to the slice image, region adjustment information of the reference annotation region is determined according to the slice image, and the reference annotation region is adjusted according to the region adjustment information to obtain a target annotation region, wherein the target annotation region is used for annotating the image to be annotated, so that the automation degree of the image annotation process can be effectively improved, the time cost of image annotation is reduced, and the image annotation efficiency is effectively improved.

An image annotation device provided by an embodiment of a second aspect of the present disclosure includes: the first acquisition module is used for acquiring a slice image of an object according to an image to be marked; the second acquisition module is used for acquiring a reference marking area of an object in the image to be marked according to the slice image; the determining module is used for determining the area adjustment information of the reference labeling area according to the slice image; and the first processing module is used for adjusting the reference labeling area according to the area adjustment information to obtain a target labeling area, wherein the target labeling area is used for labeling the image to be labeled.

The image annotation device provided by the embodiment of the second aspect of the present disclosure obtains a slice image of an object according to an image to be annotated, obtains a reference annotation region of the object in the image to be annotated according to the slice image, determines region adjustment information of the reference annotation region according to the slice image, and adjusts the reference annotation region according to the region adjustment information to obtain a target annotation region, where the target annotation region is used for annotating the image to be annotated, so that the automation degree of the image annotation process can be effectively improved, the time cost of image annotation is reduced, and the image annotation efficiency is effectively improved.

An embodiment of a third aspect of the present disclosure provides an electronic device, including: the image annotation method comprises a memory, a processor and a computer program stored on the memory and capable of running on the processor, wherein the processor executes the program to realize the image annotation method as set forth in the embodiment of the first aspect of the disclosure.

A fourth aspect of the present disclosure provides a non-transitory computer-readable storage medium, on which a computer program is stored, where the computer program, when executed by a processor, implements the image annotation method as set forth in the first aspect of the present disclosure.

A fifth aspect of the present disclosure provides a computer program product, wherein when instructions in the computer program product are executed by a processor, the image annotation method as set forth in the first aspect of the present disclosure is performed.

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

Drawings

The foregoing and/or additional aspects and advantages of the present disclosure will become apparent and readily appreciated from the following description of the embodiments, taken in conjunction with the accompanying drawings of which:

fig. 1 is a schematic flowchart of an image annotation method according to an embodiment of the disclosure;

FIG. 2 is a schematic flowchart of an image annotation method according to another embodiment of the disclosure;

fig. 3 is a schematic view of an application scenario proposed in the embodiment of the present disclosure;

FIG. 4 is a flowchart illustrating an image annotation method according to another embodiment of the disclosure;

FIG. 5 is a flowchart illustrating an image annotation method according to another embodiment of the disclosure;

FIG. 6 is a schematic diagram of an image annotation process according to an embodiment of the disclosure;

fig. 7 is a schematic structural diagram of an image annotation device according to an embodiment of the disclosure;

FIG. 8 is a schematic structural diagram of an image annotation apparatus according to another embodiment of the disclosure;

FIG. 9 illustrates a block diagram of an exemplary electronic device suitable for use in implementing embodiments of the present disclosure.

Detailed Description

Reference will now be made in detail to the embodiments of the present disclosure, examples of which are illustrated in the accompanying drawings, wherein like or similar reference numerals refer to the same or similar elements or elements having the same or similar functions throughout. The embodiments described below with reference to the drawings are exemplary only for the purpose of illustrating the present disclosure and should not be construed as limiting the same. On the contrary, the embodiments of the disclosure include all changes, modifications and equivalents coming within the spirit and terms of the claims appended hereto.

Fig. 1 is a schematic flowchart of an image annotation method according to an embodiment of the disclosure.

It should be noted that an execution main body of the image annotation method of this embodiment is an image annotation device, the device may be implemented in a software and/or hardware manner, the device may be configured in an electronic device, the electronic device may include but is not limited to a terminal, a server, and the like, for example, the terminal may be a mobile phone, a palm computer, a vehicle-mounted controller, and the like.

As shown in fig. 1, the image annotation method includes:

s101: and acquiring a slice image of the object according to the image to be marked.

The image to be labeled is an image to be labeled, and the number of the image to be labeled may be one or more, which is not limited to this.

In the embodiment of the present disclosure, an image obtaining device may be configured in advance in the execution main body of the embodiment of the present disclosure to obtain an image to be labeled, for example, a camera device is configured in an intelligent vehicle to obtain a traffic image in a scene where the intelligent vehicle is located, and the traffic image is used as an image to be labeled, or a data interface may be configured in advance for the image labeling device, an image labeling request is received via the data interface, and then the image to be labeled is obtained by analyzing the image labeling request, which is not limited.

The object may refer to an object to be marked in the image to be marked, such as a person, an automobile, a bicycle, a traffic light, and the like, without limitation. The number of the objects may be one or more.

The slice image refers to an image obtained by slicing an image to be annotated, for example, a person image, a car image, a traffic light image, and the like in the image to be annotated. The number of slice images may correspond to the number of objects described above.

In some embodiments, when a slice image of an object is obtained according to an image to be labeled, the image to be labeled may be binarized to obtain a binarized image, a circular printing region is created, the binarized image is introduced into the circular printing region, a circle center of the circular printing region is used as a rotation center, a radius line of the circular printing region is used as a traverse line, the binarized image is traversed to obtain a traverse array, then data comparison is performed on data of each row of the traverse array line by line, a compared binarized value is output, and a converted slice image is obtained according to the binarized value.

In other embodiments, when the slice image of the object is obtained according to the image to be annotated, the texture feature enhancement processing may be performed on the image to be annotated, and then the object in the image to be annotated is sliced based on the texture feature after the enhancement processing, so as to obtain the slice image of the object.

Alternatively, in some embodiments, any other possible method may also be adopted to obtain a slice image of the object according to the image to be labeled, such as an engineering or machine learning model, and the like, which is not limited thereto.

It can be understood that the image to be labeled may include a plurality of objects and image data of other factors, and in the embodiment of the present disclosure, a slice image of an object is obtained according to the image to be labeled, so that independent segmentation for each object can be implemented, thereby providing a reliable analysis object for a subsequent image labeling process, and simultaneously effectively reducing interference of other factors on the image labeling process, thereby effectively improving an image labeling effect.

S102: and acquiring a reference marking area of an object in the image to be marked according to the slice image.

The labeling area refers to an area where each object is used for labeling and processing an image to be labeled. The reference labeling area refers to a reference labeling area of an object in an image to be labeled, which is acquired based on the slice image. The number of the reference labeling area may be one or more.

In some embodiments, when the reference labeling area of the object in the image to be labeled is obtained according to the slice image, the image in the slice image may be subjected to recognition processing, shape information of the object is determined according to a result of the recognition processing, and then the reference labeling area of the object in the image to be labeled is obtained based on the shape information.

In other embodiments, when the reference labeling area of the object in the image to be labeled is obtained according to the slice image, the slice image may also be input into a pre-trained machine learning model to obtain the reference labeling area of the object in the image to be labeled, and the reference labeling area is transmitted to the execution subject of the embodiment of the present disclosure.

Alternatively, any other possible method may also be adopted to obtain the reference labeling area of the object in the image to be labeled according to the slice image, which is not limited to this.

For example, in the embodiment of the present disclosure, a neural network model may be used to construct a full-automatic preliminary screening module for a potential object, so as to process an image to be labeled, and obtain a reference labeling area, such as a convolutional neural network. The neural network model can take a convolutional neural network as a backbone network and is connected with a multi-stage feature fusion output module, and each multi-stage feature fusion module can be connected with two full connection layers and a full convolutional network in parallel to respectively complete the functions of classification, positioning and identification; the neural network model can be subjected to self-supervision pre-training until convergence, and then the model parameters are obtained to be used as initial values of module training.

When the fully-automatic latent object preliminary screening module is used for pre-training, road scene data sets from public road scene data sets and/or internal labels (the covered label objects comprise pedestrians, various vehicles, lane lines, traffic lights, road signboards and the like) can be obtained as training data.

In the embodiment of the disclosure, by acquiring the reference labeling area of the object in the image to be labeled according to the slice image, the method can realize the preliminary determination of the labeling area in the image to be labeled, and the obtained reference labeling area can provide reliable reference basis for the subsequent determination of the target labeling area.

S103: and determining the area adjustment information of the reference labeling area according to the slice image.

The region adjustment information refers to information related to an adjustment process corresponding to the reference labeled region, and can be used to indicate an adjustment process of a subsequent reference labeled region.

In some disclosed embodiments, when determining the region adjustment information of the reference labeling region according to the slice image, a plurality of reference region adjustment information sets may be obtained in advance, and then the plurality of reference region adjustment information sets may be matched with the slice image, and the region adjustment information suitable for the reference labeling region may be determined from the plurality of reference region adjustment information sets according to a matching processing result.

In other embodiments, when determining the region adjustment information of the reference labeling region according to the slice image, a communication link between the execution main body and the big data server according to the embodiment of the disclosure may be established in advance, and then the region adjustment information of the reference labeling region may be acquired from the big data server according to the slice image.

Of course, any other possible method may be adopted to determine the region adjustment information of the reference labeling region according to the slice image, which is not limited to this.

It can be understood that there may be an error in the image annotation effect of the obtained reference annotation region, and when the region adjustment information of the reference annotation region is determined according to the slice image, a reliable reference basis can be provided for the subsequent adjustment process of the reference annotation region, so that the image annotation effect of the obtained target annotation region is effectively improved.

S104: and adjusting the reference labeling area according to the area adjustment information to obtain a target labeling area, wherein the target labeling area is used for labeling the image to be labeled.

The target labeling area refers to a labeling area used for labeling an image to be labeled.

In the embodiment of the present disclosure, when the reference annotation area is adjusted according to the area adjustment information to obtain the target annotation area, the adjustment object and the adjustment value corresponding to the adjustment object may be determined based on the area adjustment information, and then the corresponding adjustment object is adjusted based on the adjustment value to obtain the target annotation area, or the area adjustment information may be input into a pre-trained image adjustment model to obtain the target adjustment area and transmitted to the execution main body in the embodiment of the present disclosure, which is not limited thereto.

In the embodiment, a slice image of an object is acquired according to an image to be annotated, a reference annotation region of the object in the image to be annotated is acquired according to the slice image, region adjustment information of the reference annotation region is determined according to the slice image, and the reference annotation region is adjusted according to the region adjustment information to obtain a target annotation region, wherein the target annotation region is used for annotating the image to be annotated.

Fig. 2 is a schematic flowchart of an image annotation method according to another embodiment of the disclosure.

As shown in fig. 2, the image annotation method includes:

s201: and determining a local image area of the object in the image to be annotated.

The local image area may be an image area corresponding to an object in the annotation image.

The image labeling method provided by the embodiment of the disclosure can be applied to the technical field of automatic driving to realize the labeling process of various road scene data, and can be extended to other technical fields.

For example, as shown in fig. 3, fig. 3 is a schematic view of an application scene provided in the embodiment of the present disclosure, a user may sequentially click a person, an automobile, and a car in a left traffic scene to obtain a pixel-level label corresponding to a target object, and then the image labeling method in the embodiment of the present disclosure may assist in outputting a refined label of the target object to save manual labeling time and improve labeling efficiency.

S202: and performing at least one expansion processing on the regional boundary of the local image region in the image to be annotated to obtain a slice image region selected by the expanded regional boundary frame each time.

The slice image region is an image region obtained by performing expansion processing on a local image region.

For example, when the region boundary of the local image region in the image to be annotated is expanded, the center and the size information of the local image region may be determined, and then 100% of the obtained size information is expanded based on the center of the local image region to obtain the slice region image.

It can be understood that the characterization effect of the local image region on the object related information may have a defect, and when the region boundary of the local image region in the image to be annotated is subjected to at least one expansion process, the characterization integrity of the obtained slice image region on the corresponding object can be effectively improved, so that a reliable reference basis is provided for subsequently generated slice images.

S203: from the slice image region, a slice image is generated.

In the embodiment of the present disclosure, when generating a slice image according to a slice image region, a to-be-annotated image may be cut based on the slice image region to obtain a slice image corresponding to the slice image region.

That is to say, in the embodiment of the present disclosure, a local image region of an object in an image to be annotated may be determined, a region boundary of the local image region in the image to be annotated is subjected to at least one expansion process, a slice image region framed and selected by the expanded region boundary is obtained each time, and then a slice image is generated according to the slice image region, thereby, flexible expansion processing of the local image region may be implemented, corresponding representation integrity of the obtained slice region image is ensured, and a generation effect of the slice image is effectively improved.

S204: and acquiring an initial labeling area of an object in the image to be labeled.

In the embodiment of the present disclosure, a computer technology may be used to perform preliminary screening on an object that may become an annotation object in an image to be annotated to obtain a plurality of reference annotation objects, and an image area corresponding to the reference annotation image may be referred to as an initial annotation area.

Therefore, preliminary screening of the labeling area in the image to be labeled can be achieved, a reliable analysis object is provided for subsequently determining the reference labeling area, and the determination efficiency of the reference labeling area is effectively improved.

S205: and processing the initial labeling area according to the slice image to obtain a reference labeling area.

In some embodiments, when the initial labeling area is processed according to the slice image to obtain the reference labeling area, contour enhancement processing may be performed on an object in the initial labeling area, and then the initial labeling area after contour enhancement processing is processed based on the slice image to obtain the reference labeling area.

In other embodiments, when the initial annotation region is processed according to the slice image to obtain the reference annotation region, the attribute information of the object in the initial annotation image may be acquired, and then the initial annotation region is processed based on the attribute information of the object in the initial annotation image and the slice image to obtain the reference annotation region.

Of course, in some embodiments, any other possible method may be used to process the initial labeled region according to the slice image to obtain the reference labeled region, such as an engineering or mathematical method, which is not limited herein.

That is to say, according to the slice image region, after the slice image is generated, the initial annotation region of the object in the image to be annotated can be obtained, and then the initial annotation region is processed according to the slice image to obtain the reference annotation region, so that the annotation region in the image to be annotated can be screened to obtain the initial annotation image, and then the initial annotation region is processed by combining with the slice image, so that the reliability of the obtained reference annotation region can be ensured by combining with the slice image while the efficiency of determining the reference annotation region is effectively improved.

S206: slice image characteristics of the slice image are determined.

The slice image feature refers to a relevant feature in the slice image, and may be, for example, a position feature, a contour feature, a size feature, and the like of the object, which is not limited thereto.

In the embodiment of the disclosure, when the slice image features of the slice image are determined, the obtained slice image features can effectively represent the feature information corresponding to the slice image, so that a reliable reference basis is provided for subsequently determining the region adjustment information of the reference labeling region.

S207: and determining the area adjustment information of the reference labeling area according to the slice image and the characteristics of the slice image.

In some embodiments, when determining the region adjustment information of the reference labeling region according to the slice image and the characteristics of the slice image, size information of an object in the slice image may be acquired, and then the region adjustment information of the reference labeling region may be determined by combining the size information and the slice image.

In other embodiments, when determining the region adjustment information of the reference labeling region according to the slice image and the characteristics of the slice image, the chrominance characteristics of the object in the slice image may be acquired, and then the region adjustment information of the reference labeling region may be determined by combining the chrominance characteristics and the slice image.

Of course, in some embodiments, any other possible method may also be used to determine the region adjustment information of the reference labeling region according to the slice image and the characteristics of the slice image, such as an engineering or a combination of mathematical and numerical methods, which is not limited herein.

That is to say, according to the embodiment of the present disclosure, after the initial annotation region is processed according to the slice image to obtain the reference annotation region, the slice image feature of the slice image may be determined, and the region adjustment information of the reference annotation region may be determined according to the slice image and the slice image feature.

S208: and adjusting the reference labeling area according to the area adjustment information to obtain a target labeling area, wherein the target labeling area is used for labeling the image to be labeled.

For the description of S208, reference may be made to the foregoing embodiments, and details are not repeated herein.

S209: and acquiring the classification information of the object in the image to be labeled.

The classification information may be related information obtained by performing classification processing on a plurality of objects in the annotation image.

For example, in a traffic scene, the embodiments of the present disclosure may classify a plurality of objects in a corresponding image to be labeled according to the types of pedestrians, automobiles, traffic signs, other obstacles, and the like, so as to obtain corresponding classification information.

In the embodiment of the disclosure, by acquiring the classification information of the object in the image to be labeled, a reliable reference basis can be provided for the subsequent image labeling process.

S210: and labeling the image to be labeled according to the target labeling area and the classification information to obtain a target labeling image.

When the image to be labeled is labeled according to the target labeling area and the classification information, the labeling of each target labeling area can be performed based on the classification information so as to complete the labeling processing of the image to be labeled.

For example, when the full-automatic latent object prescreening module is used, an RGB three-channel image to be marked may be input, and then a latent object in a current image is output, where the output information includes: and (4) classifying information, position characteristics and an initial labeling area of the object, and outputting the information and the position characteristics as L1. Because the L1 outputs the classification information of the object, the correctness of the classification information only needs to be checked manually in the subsequent process; if the target cannot be effectively identified by the L1, the classification information of the object needs to be added later manually. Therefore, the classification information of the object is output in an auxiliary mode, and the manual classification cost is reduced.

That is to say, according to the embodiment of the present disclosure, after the reference labeling area is adjusted according to the area adjustment information to obtain the target labeling area, the classification information of the object in the image to be labeled can be obtained, and then the image to be labeled is labeled according to the target labeling area and the classification information to obtain the target labeling image.

In the embodiment, a local image area of an object in an image to be annotated is determined, area boundaries of the local image area in the image to be annotated are expanded at least once to obtain a slice image area framed by the boundary of the expanded area each time, then a slice image is generated according to the slice image area, so that flexible expansion processing of the local image area can be realized, corresponding characterization integrity of the image of the obtained slice area is ensured, and generation effect of the slice image is effectively improved Therefore, the reliability of the image annotation result is effectively improved while the accuracy of the image annotation range is ensured.

Fig. 4 is a flowchart illustrating an image annotation method according to another embodiment of the disclosure.

As shown in fig. 4, the image annotation method includes:

s401: and acquiring a slice image of the object according to the image to be marked.

S402: and acquiring a reference marking area of an object in the image to be marked according to the slice image.

S403: slice image features of the slice image are determined.

For the description of S401 to S403, reference may be made to the above embodiments, which are not described herein again.

S404: contour points of the object are acquired from the slice image.

It can be understood that the slice image may be composed of a plurality of pixel points, and the pixel points corresponding to the contour of the object in the slice image may be referred to as contour points of the object.

In the embodiment of the disclosure, by acquiring contour points of an object from a slice image, preliminary determination of boundary lines between the object and other image factors in the slice image can be realized.

S405: the position feature of the contour point in the slice image is determined.

The position feature may be related information describing a spatial position attribute of the contour point in the slice image, and may be, for example, a distance between the contour point and a peripheral boundary of the slice image.

When determining the position features of the contour points in the slice images, the disclosed embodiment may determine the minimum distance values between a plurality of contour points and the peripheral boundary of the slice images, and then determine the position features of the contour points in the slice images based on the obtained minimum distance values.

S406: and acquiring contour point characteristics corresponding to the contour points from the slice image characteristics.

The contour point feature may refer to feature information corresponding to the contour point, such as a number feature, a distribution feature, and the like of the contour point.

It can be understood that the contour point features have high relevance to the adjustment process of the reference labeling area, and when the contour point features corresponding to the contour points are acquired from the slice image features, reliable reference basis can be provided for subsequently determining the area adjustment information of the reference labeling area.

S407: and determining the area adjustment information of the reference labeling area according to the position characteristic, the contour point characteristic and the slice image.

In some embodiments, when the region adjustment information of the reference labeling region is determined according to the position feature, the contour point feature, and the slice image, the reference slice image may be determined in advance, the reference position feature and the reference contour point feature corresponding to the reference slice image are determined, then the position feature and the contour point feature are analyzed and compared with the reference position feature and the reference contour point feature, respectively, and then the region adjustment information of the reference labeling region is determined according to the analysis and comparison result.

In other embodiments, when the region adjustment information of the reference labeling region is determined according to the position feature, the contour point feature, and the slice image, the expansion processing or reduction processing information corresponding to the slice image may be determined according to the position feature and the contour point feature, and then the obtained expansion processing or reduction processing information may be used as the region adjustment information of the reference labeling region.

Of course, in some embodiments, any other possible method may also be adopted to determine the region adjustment information of the reference labeling region according to the position feature, the contour point feature, and the slice image, which is not limited to this.

Optionally, in some embodiments, when determining the region adjustment information of the reference labeling region according to the position feature, the contour point feature, and the slice image, a position adjustment direction and a position adjustment value corresponding to the position feature may be determined according to the contour point feature and the slice image, and the position adjustment direction and the position adjustment value are used as the region adjustment information, so that when the position adjustment direction and the position adjustment value are used as the region adjustment information, the adjustment direction and the adjustment value corresponding to the region adjustment information may be accurately quantized, thereby effectively improving the definition of the region adjustment information on the characterization of the adjustment content, and facilitating accurate adjustment of the reference labeling region.

The position adjustment direction may be an adjustment direction corresponding to the reference marked area when the position is adjusted.

The position adjustment value refers to an adjustment value corresponding to the reference marking area when the position adjustment is performed.

In the embodiment of the present disclosure, when determining the position adjustment direction and the position adjustment value corresponding to the position feature according to the contour point feature and the slice image, the center of gravity of the corresponding object in the slice image may be determined according to the contour point feature, and then the position adjustment direction and the position adjustment value corresponding to the position feature may be determined according to the position information of the center of gravity, or the minimum distance value between the contour point feature and the boundary of the slice image may be determined according to the contour point feature, and then the position adjustment direction and the position adjustment value corresponding to the position feature may be determined according to the minimum distance value, which is not limited herein.

Optionally, in some embodiments, when determining the position adjustment direction and the position adjustment value corresponding to the position feature according to the contour point feature and the slice image, the contour point feature and the slice image may be input into the first adjustment information determination model to obtain the position adjustment direction and the position adjustment value output by the first adjustment information determination model, so that the position adjustment direction and the position adjustment value may be determined quickly and accurately based on the first adjustment information determination model, and the degree of automation of the position adjustment direction and the position adjustment value determination process may be effectively improved.

The adjustment information determination model may be, for example, a neural network model, a machine learning model, or any other possible model capable of performing the adjustment information determination task, which is not limited herein. And the first adjustment information determination model refers to an adjustment information determination model used for processing the contour point feature and the slice image to determine a position adjustment direction and a position adjustment value.

For example, the embodiment of the present disclosure may construct a full-automatic fine tuning module for the selected object based on a neural network model, the structure of the module may be a convolutional neural network, and the input data may be a feature map corresponding to the slice image and a reference labeling area, so as to output a target labeling area.

The module can be composed of a multilayer neural network, and the using process can be as follows: extracting a plurality of contour points of an object in a slice image; acquiring contour point features and position features corresponding to the contour points at the same positions in the corresponding feature map, and marking the feature values of the contour points as f; traversing a plurality of contour points, inputting the characteristic value corresponding to each contour point into a multilayer neural network, judging the relative position of the contour point and an object by the multilayer neural network, and then outputting the corresponding position adjustment direction, namely, inwards contracting towards the inside of the object or outwards expanding towards the outside of the object; and repeatedly executing the steps until the characteristic points are input into the multilayer neural network, and then the multilayer neural network determines that the corresponding contour points are positioned at the edges of the objects so as to stop the fine tuning process, or when the directions of the contour point positions output by the multilayer neural network and the upper wheel output contour point positions are opposite, the fine tuning process can be stopped. After the fine adjustment is completed, the reference labeling area may be adjusted based on the fine-adjusted object contour to obtain a fine-adjusted pixel-level labeling area.

That is to say, after determining the slice image feature of the slice image, the embodiment of the present disclosure may acquire the contour point of the object from the slice image, determine the position feature of the contour point in the slice image, acquire the contour point feature corresponding to the contour point from the slice image feature, and determine the area adjustment information of the reference labeling area according to the position feature, the contour point feature, and the slice image, thereby, the position feature, the contour point feature, and the feature information of multiple dimensions of the slice consideration image may be effectively combined to implement comprehensive adjustment of the area adjustment information, thereby effectively improving the adaptability between the obtained area adjustment information and the reference labeling area, and improving the adjustment effect of the area adjustment information on the reference labeling area.

S408: and adjusting the reference labeling area according to the area adjustment information to obtain a target labeling area, wherein the target labeling area is used for labeling the image to be labeled.

For the description of S408, reference may be made to the above embodiments, which are not described herein again.

In the embodiment, the contour point of the object is acquired from the slice image, the position feature of the contour point in the slice image is determined, the contour point feature corresponding to the contour point is acquired from the slice image feature, and the region adjustment information of the reference labeling region is determined according to the position feature, the contour point feature and the slice image, so that the comprehensive consideration of the region adjustment information can be realized by effectively combining the position feature, the contour point feature and the feature information of multiple dimensions of the slice image, the adaptability between the obtained region adjustment information and the reference labeling region is effectively improved, the adjustment effect of the region adjustment information on the reference labeling region is improved, the position adjustment direction and the position adjustment value corresponding to the position feature are determined according to the contour point feature and the slice image, the position adjustment direction and the position adjustment value are used as area adjustment information, therefore, when the area adjustment information is used as the area adjustment information based on the position adjustment direction and the position adjustment value, the adjustment direction and the adjustment value corresponding to the area adjustment information can be accurately quantized, the representation definition of the area adjustment information on adjustment contents is effectively improved, so that accurate adjustment of a reference labeling area is conveniently realized, the contour point characteristics and the slice image are input into the first adjustment information determination model, so that the position adjustment direction and the position adjustment value output by the first adjustment information determination model are obtained, therefore, the position adjustment direction and the position adjustment value can be quickly and accurately determined based on the first adjustment information determination model, and the automation degree of the position adjustment direction and position adjustment value determination process can be effectively improved.

Fig. 5 is a flowchart illustrating an image annotation method according to another embodiment of the disclosure.

As shown in fig. 5, the image annotation method includes:

s501: and acquiring a slice image of the object according to the image to be marked.

S502: and acquiring an initial labeling area of an object in the image to be labeled.

For the description of S501 and S502, reference may be made to the above embodiments, and details are not repeated here.

S503: and acquiring positive click data and negative click data, wherein the positive click data is click data of which the click position is positioned in the object in the image to be marked, and the negative click data is click data of which the click position is positioned outside the object in the image to be marked.

The click data may be data generated by clicking an image to be annotated by a user.

The forward click data may be click data generated when the user clicks an area, which is not covered by the initial labeling area, in the image area where the object is located.

The negative click data may be click data generated by the user clicking an area outside the image area where the object is located and covered by the initial labeling area.

In the embodiment of the disclosure, by acquiring the positive click data and the negative click data, the rough adjustment of the initial labeling area can be completed by effectively combining with the manual labeling result in the image labeling process, so as to provide reliable reference data for subsequently determining the reference labeling area.

S504: and processing the initial labeling area according to the slice image, the positive click data and the negative click data to obtain a reference labeling area.

In some embodiments, when the initial annotation area is processed according to the slice image, the positive click data, and the negative click data to obtain the reference annotation area, feature enhancement processing may be performed on the initial annotation area based on the positive click data and the negative click data to obtain an outline boundary corresponding to an object in the initial annotation area, and then the outline boundary and the slice image are combined to determine the reference annotation area.

In other embodiments, when the initial annotation area is processed according to the slice image, the positive click data, and the negative click data to obtain the reference annotation area, distance information between the positive click data and the negative click data may be determined, and then the reference annotation area may be determined by combining the distance information and the slice image.

Alternatively, any other possible method may be adopted, and the initial labeling area is processed according to the slice image, the positive click data, and the negative click data, so as to obtain a reference labeling area, which is not limited to this.

Optionally, in some embodiments, when the initial annotation area is processed according to the slice image, the positive click data, and the negative click data to obtain the reference annotation area, reference adjustment information corresponding to the initial annotation area may be generated according to the slice image, the positive click data, and the negative click data, and then the initial annotation area is adjusted according to the reference adjustment information to obtain the reference annotation area.

The reference adjustment information may refer to adjustment information corresponding to the initial labeling area, and the reference adjustment information may be used to indicate an adjustment process corresponding to the initial labeling area to obtain the reference labeling area.

In the embodiment of the present disclosure, when the reference adjustment information corresponding to the initial labeling area is generated according to the slice image, the positive click data, and the negative click data, the position information of the positive click data and the negative click data in the slice image may be determined, and then the reference adjustment information corresponding to the initial labeling area may be determined according to the position information, or a third-party adjustment information generation device may be further used to process the slice image, the positive click data, and the negative click data to obtain the reference adjustment information corresponding to the initial labeling area, and transmit the reference adjustment information to the execution main body of the embodiment of the present disclosure, which is not limited thereto.

Optionally, in some embodiments, when reference adjustment information corresponding to the initial annotation area is generated according to the slice image, the positive click data, and the negative click data, the slice image, the positive click data, and the negative click data may be input into the second adjustment information determination model to obtain reference adjustment information output by the second adjustment information determination model, so that extraction and fusion processing of relevant feature information in the slice image, the positive click data, and the negative click data may be implemented based on the second adjustment information determination model, and an indication effect of the obtained reference adjustment information on an adjustment process of the initial annotation area may be effectively improved.

The second adjustment information determination model is an adjustment information determination model used for processing the slice image, the positive click data, and the negative click data to obtain the reference adjustment information.

For example, the embodiment of the present disclosure may use a neural network model to construct a semi-automatic coarse tuning module for the selected object, where the structure of the module may be a convolutional neural network, and the module is configured to process an RGB three-channel slice image, positive click data, negative click data, and an artificially labeled initial labeling area, so as to output an initial labeling area corresponding to the selected object.

When the module carries out model training, a public data set can be obtained as training data, positive click data and negative click data respectively represent that click positions are inside or outside an object, and an initial labeling area of manual labeling is empty; in the iterative training process of the model, the target marking area obtained in the previous iterative process can be used as the initial marking area of the current round, the iteration times can be 3 times, and 2 pieces of positive click data and 1 piece of negative click data can be input in each round of iterative process; the semi-automatic coarse adjustment module of the selected object can be formed by sequentially connecting a convolutional neural network, an attention fusion module and a full convolutional network so as to determine an initial labeling area of the selected object; the convolutional neural network can perform self-supervision pre-training in the public data set so as to train until the model converges, and then obtain the model parameters as initial values of module training.

During the use of the module, the module is, the slice image, the positive click data, the negative click data, and the initial annotation region can be processed to obtain a reference annotation region.

That is, after acquiring the initial annotation region of the object in the image to be annotated, the embodiment of the present disclosure may acquire positive click data and negative click data, where the positive click data is click data whose click position is located in the object in the image to be annotated, and the negative click data is click data whose click position is located outside the object in the image to be annotated, and process the initial annotation region according to the slice image, the positive click data, and the negative click data to obtain the reference annotation region.

S505: and determining the area adjustment information of the reference labeling area according to the slice image.

S506: and adjusting the reference labeling area according to the area adjustment information to obtain a target labeling area, wherein the target labeling area is used for labeling the image to be labeled.

For the description of S505 and S506, reference may be made to the above embodiments, which are not described herein again.

For example, as shown in fig. 6, fig. 6 is a schematic diagram of an image labeling process provided in the embodiment of the present disclosure, including:

(a) Acquiring an image to be marked;

(b) Primarily screening objects which can become marking objects in the whole image to be marked by using a computer, outputting a rough pixel level mark (namely an initial marking area) to potential objects appearing in all scenes, outputting the classification information of the potential objects, and recording the classification information as L1;

(c) Manually selecting an object to be labeled, if the object is repeated with the object obtained in the L1, taking the pixel-level label of the object in the L1 as an initial labeling area, expanding the local image area where the object in the L1 is located by 100% to obtain a slice image area, and cutting the image on the image to be labeled based on the obtained slice image area to be used as a slice image of the object. If the object is not repeated with the object obtained in the L1, the corresponding initial labeling area of the object is empty;

(d) Inputting the slice image and the manual click data into a computer system, and processing the slice image and the manual click data through a machine learning model to obtain a reference labeling area, for example, an initial labeling area of an object in L1 can be compared with an area where the object is actually located, and the reference labeling area is clicked once in an undetected area to serve as forward click data; clicking once in the area which is detected by mistake to serve as negative click data;

(e) Outputting a coarse adjustment result of the object, wherein the process can be iterated for 1 to 2 times, and the output result is recorded as L2;

(f) And inputting the L2 into a computer system to finely adjust the outline of the object, recording an output result as L3, finely adjusting the L3 again by manpower, and outputting a target labeling image L4.

In this embodiment, by obtaining the positive click data and the negative click data, wherein the positive click data is the click data whose click position is located inside the object in the image to be labeled, and the negative click data is the click data whose click position is located outside the object in the image to be labeled, and processing the initial labeling area according to the slice image, the positive click data and the negative click data to obtain a reference labeling area, since the acquisition process of the positive click data and the negative click data is simple and convenient, when the initial labeling area is processed based on the slice image, the positive click data, and the negative click data to obtain the reference labeling area, can effectively combine with the manual annotation result to realize the rapid determination of the reference annotation area, generates the reference adjustment information corresponding to the initial annotation area according to the slice image, the positive click data and the negative click data, then, the initial labeling area is adjusted according to the reference adjustment information to obtain a reference labeling area, and therefore, the reference adjustment information corresponding to the initial labeling area can be accurately determined by effectively combining the slice image, the positive click data and the negative click data, thereby providing reliable execution basis for the adjustment process corresponding to the initial labeling area, ensuring the adaptability between the obtained reference labeling area and the object, by inputting the slice image, the positive click data, and the negative click data into the second adjustment information determination model to obtain the reference adjustment information output by the second adjustment information determination model, therefore, extraction and fusion processing of relevant characteristic information in the slice image, the positive click data and the negative click data can be achieved based on the second adjustment information determination model, and the indication effect of the obtained reference adjustment information on the initial labeling area adjustment process can be effectively improved.

Fig. 7 is a schematic structural diagram of an image annotation device according to an embodiment of the disclosure.

As shown in fig. 7, the image labeling apparatus 70 includes:

a first obtaining module 701, configured to obtain a slice image of an object according to an image to be annotated;

a second obtaining module 702, configured to obtain, according to the slice image, a reference annotation region of an object in the image to be annotated;

a determining module 703, configured to determine, according to the slice image, area adjustment information of the reference labeling area;

the first processing module 704 is configured to adjust the reference labeling area according to the area adjustment information to obtain a target labeling area, where the target labeling area is used for labeling an image to be labeled.

In some embodiments of the present disclosure, as shown in fig. 8, fig. 8 is a schematic structural diagram of an image annotation apparatus according to another embodiment of the present disclosure, and the determining module 703 includes:

a first determining sub-module 7031 for determining a slice image feature of the slice image;

a second determining sub-module 7032 is configured to determine region adjustment information of the reference labeling region according to the slice image and the slice image feature.

In some embodiments of the present disclosure, the second determining submodule 7032 is specifically configured to:

acquiring contour points of an object from a slice image;

determining the position characteristics of the contour points in the slice images;

acquiring contour point characteristics corresponding to the contour points from the slice image characteristics;

and determining the area adjustment information of the reference labeling area according to the position characteristic, the contour point characteristic and the slice image.

In some embodiments of the present disclosure, second determining submodule 7032 is further configured to:

determining a position adjustment direction and a position adjustment value corresponding to the position characteristic according to the contour point characteristic and the slice image;

the position adjustment direction and the position adjustment value are used as area adjustment information.

and inputting the contour point features and the slice images into the first adjustment information determination model to obtain a position adjustment direction and a position adjustment value output by the first adjustment information determination model.

In some embodiments of the present disclosure, the second obtaining module 702 includes:

the obtaining sub-module 7021 is configured to obtain an initial labeling area of an object in the image to be labeled;

and the processing sub-module 7022 is configured to process the initial labeling area according to the slice image to obtain a reference labeling area.

In some embodiments of the present disclosure, processing sub-module 7022 is specifically configured to:

acquiring positive click data and negative click data, wherein the positive click data is click data of which the click position is positioned in an object in an image to be marked, and the negative click data is click data of which the click position is positioned outside the object in the image to be marked;

and processing the initial labeling area according to the slice image, the positive click data and the negative click data to obtain a reference labeling area.

In some embodiments of the disclosure, processing sub-module 7022 is further configured to:

generating reference adjustment information corresponding to the initial labeling area according to the slice image, the positive click data and the negative click data;

and adjusting the initial labeling area according to the reference adjustment information to obtain a reference labeling area.

In some embodiments of the present disclosure, processing submodule 7022 is further configured to:

and inputting the slice image, the positive click data and the negative click data into the second adjustment information determination model to obtain the reference adjustment information output by the second adjustment information determination model.

In some embodiments of the present disclosure, the first obtaining module 701 is specifically configured to:

determining a local image area of an object in an image to be marked;

performing at least one expansion processing on the regional boundary of a local image region in an image to be annotated to obtain a slice image region framed and selected by the expanded regional boundary each time;

from the slice image region, a slice image is generated.

In some embodiments of the disclosure, the apparatus further comprises:

a third obtaining module 705, configured to obtain classification information of an object in an image to be annotated;

and the second processing module 706 is configured to label the image to be labeled according to the target labeling area and the classification information, so as to obtain a target labeled image.

It should be noted that the above explanation of the image annotation method is also applicable to the image annotation apparatus of the present embodiment, and is not repeated here.

FIG. 9 illustrates a block diagram of an exemplary electronic device suitable for use in implementing embodiments of the present disclosure. The electronic device 12 shown in fig. 9 is only an example and should not bring any limitation to the functions and the scope of use of the embodiments of the present disclosure.

As shown in fig. 9, the electronic device 12 is in the form of a general purpose computing device. The components of electronic device 12 may include, but are not limited to: one or more processors or processing units 16, a system memory 28, and a bus 18 that couples various system components including the system memory 28 and the processing unit 16.

Bus 18 represents one or more of any of several types of bus structures, including a memory bus or memory controller, a peripheral bus, an accelerated graphics port, and a processor or local bus using any of a variety of bus architectures. These architectures include, but are not limited to, industry Standard Architecture (ISA) bus, micro Channel Architecture (MAC) bus, enhanced ISA bus, video Electronics Standards Association (VESA) local bus, and Peripheral Component Interconnect (PCI) bus, to name a few.

Electronic device 12 typically includes a variety of computer system readable media. Such media may be any available media that is accessible by electronic device 12 and includes both volatile and nonvolatile media, removable and non-removable media.

Memory 28 may include computer system readable media in the form of volatile Memory, such as Random Access Memory (RAM) 30 and/or cache Memory 32. Electronic device 12 may further include other removable/non-removable, volatile/nonvolatile computer system storage media. By way of example only, storage system 34 may be used to read from and write to non-removable, nonvolatile magnetic media (not shown in FIG. 9, and commonly referred to as a "hard drive").

Although not shown in FIG. 9, a disk drive for reading from and writing to a removable, nonvolatile magnetic disk (e.g., a "floppy disk") and an optical disk drive for reading from or writing to a removable, nonvolatile optical disk (e.g., a Compact disk Read Only Memory (CD-ROM), a Digital versatile disk Read Only Memory (DVD-ROM), or other optical media) may be provided. In these cases, each drive may be connected to bus 18 by one or more data media interfaces. Memory 28 may include at least one program product having a set (e.g., at least one) of program modules that are configured to carry out the functions of embodiments of the disclosure.

A program/utility 40 having a set (at least one) of program modules 42 may be stored, for example, in memory 28, such program modules 42 including, but not limited to, an operating system, one or more application programs, other program modules, and program data, each of which examples or some combination thereof may comprise an implementation of a network environment. Program modules 42 generally perform the functions and/or methodologies of the embodiments described in this disclosure.

Electronic device 12 may also communicate with one or more external devices 14 (e.g., keyboard, pointing device, display 24, etc.), with one or more devices that enable a person to interact with electronic device 12, and/or with any devices (e.g., network card, modem, etc.) that enable electronic device 12 to communicate with one or more other computing devices. Such communication may be through an input/output (I/O) interface 22. Also, the electronic device 12 may communicate with one or more networks (e.g., a Local Area Network (LAN), a Wide Area Network (WAN), and/or a public Network such as the Internet) via the Network adapter 20. As shown, the network adapter 20 communicates with other modules of the electronic device 12 via the bus 18. It should be understood that although not shown in the figures, other hardware and/or software modules may be used in conjunction with electronic device 12, including but not limited to: microcode, device drivers, redundant processing units, external disk drive arrays, RAID systems, tape drives, and data backup storage systems, to name a few.

The processing unit 16 executes various functional applications and image labeling, such as the image labeling method mentioned in the foregoing embodiments, by executing programs stored in the system memory 28.

In order to achieve the above embodiments, the present disclosure also proposes a non-transitory computer readable storage medium on which a computer program is stored, which when executed by a processor implements the image annotation method as proposed by the foregoing embodiments of the present disclosure.

In order to implement the foregoing embodiments, the present disclosure further provides a computer program product, which when executed by an instruction processor in the computer program product, executes the image annotation method as set forth in the foregoing embodiments of the present disclosure.

Other embodiments of the disclosure will be apparent to those skilled in the art from consideration of the specification and practice of the disclosure disclosed herein. This disclosure is intended to cover any variations, uses, or adaptations of the disclosure following, in general, the principles of the disclosure and including such departures from the present disclosure as come within known or customary practice within the art to which the disclosure pertains. It is intended that the specification and examples be considered as exemplary only, with a true scope and spirit of the disclosure being indicated by the following claims.

It will be understood that the present disclosure is not limited to the precise arrangements described above and shown in the drawings and that various modifications and changes may be made without departing from the scope thereof. The scope of the present disclosure is limited only by the appended claims.

It should be noted that, in the description of the present disclosure, the terms "first", "second", etc. are used for descriptive purposes only and are not to be construed as indicating or implying relative importance. Further, in the description of the present disclosure, "a plurality" means two or more unless otherwise specified.

Any process or method descriptions in flow charts or otherwise described herein may be understood as representing modules, segments, or portions of code which include one or more executable instructions for implementing specific logical functions or steps in the process, and alternate implementations are included within the scope of the preferred embodiment of the present disclosure in which functions may be executed out of order from that shown or discussed, including substantially concurrently or in reverse order, depending on the functionality involved, as would be understood by those reasonably skilled in the art of the embodiments of the present disclosure.

It should be understood that portions of the present disclosure may be implemented in hardware, software, firmware, or a combination thereof. In the above embodiments, various steps or methods may be implemented in software or firmware stored in a memory and executed by a suitable instruction execution system. For example, if implemented in hardware, as in another embodiment, any one or combination of the following techniques, which are known in the art, may be used: a discrete logic circuit having a logic gate circuit for implementing a logic function on a data signal, an application specific integrated circuit having an appropriate combinational logic gate circuit, a Programmable Gate Array (PGA), a Field Programmable Gate Array (FPGA), or the like.

It will be understood by those skilled in the art that all or part of the steps carried by the method for implementing the above embodiments may be implemented by hardware related to instructions of a program, which may be stored in a computer readable storage medium, and when the program is executed, the program includes one or a combination of the steps of the method embodiments.

In addition, functional units in the embodiments of the present disclosure may be integrated into one processing module, or each unit may exist alone physically, or two or more units are integrated into one module. The integrated module can be realized in a hardware mode, and can also be realized in a software functional module mode. The integrated module, if implemented in the form of a software functional module and sold or used as a stand-alone product, may also be stored in a computer readable storage medium.

The storage medium mentioned above may be a read-only memory, a magnetic or optical disk, etc.

In the description of the present specification, reference to the description of "one embodiment," "some embodiments," "an example," "a specific example," or "some examples" or the like means that a particular feature, structure, material, or characteristic described in connection with the embodiment or example is included in at least one embodiment or example of the present disclosure. In this specification, the schematic representations of the terms used above do not necessarily refer to the same embodiment or example. Furthermore, the particular features, structures, materials, or characteristics described may be combined in any suitable manner in any one or more embodiments or examples.

Although embodiments of the present disclosure have been shown and described above, it is understood that the above embodiments are exemplary and should not be construed as limiting the present disclosure, and that changes, modifications, substitutions and alterations may be made to the above embodiments by those of ordinary skill in the art within the scope of the present disclosure.

Claims

1. An image annotation method, comprising:

acquiring a slice image of an object according to an image to be marked;

acquiring a reference marking area of an object in the image to be marked according to the slice image;

determining region adjustment information of the reference labeling region according to the slice image, wherein slice image features of the slice image are determined, contour points of the object are acquired from the slice image, position features of the contour points in the slice image are determined, contour point features corresponding to the contour points are acquired from the slice image features, position adjustment directions and position adjustment values corresponding to the position features are determined according to the contour point features and the slice image, and the position adjustment directions and the position adjustment values are used as the region adjustment information;

adjusting the reference labeling area according to the area adjustment information to obtain a target labeling area, wherein the target labeling area is used for labeling the image to be labeled;

wherein the determining a position adjustment direction and a position adjustment value corresponding to the position feature according to the contour point feature and the slice image comprises: determining the gravity center of a corresponding object in the slice image according to the contour point characteristics, determining a position adjustment direction and a position adjustment value corresponding to the position characteristics according to the position information of the gravity center, or determining a minimum distance value between the contour point and the boundary of the slice image according to the contour point characteristics, and determining a position adjustment direction and a position adjustment value corresponding to the position characteristics according to the minimum distance value;

the acquiring a reference labeling area of an object in the image to be labeled according to the slice image includes:

acquiring an initial labeling area of an object in the image to be labeled;

acquiring positive click data and negative click data, wherein the positive click data is click data of which the click position is positioned in an object in the image to be marked, and the negative click data is click data of which the click position is positioned outside the object in the image to be marked;

and processing the initial labeling area according to the slice image, the positive click data and the negative click data to obtain the reference labeling area, wherein distance information between the positive click data and the negative click data is determined, and the reference labeling area is determined by combining the distance information and the slice image.

2. The method of claim 1, wherein determining a position adjustment direction and a position adjustment value corresponding to the position feature from the contour point feature and the slice image comprises:

and inputting the contour point features and the slice images into a first adjustment information determination model to obtain the position adjustment direction and the position adjustment value output by the first adjustment information determination model.

3. The method of claim 1, wherein said processing the initial labeled region based on the slice image, the positive click data, and the negative click data to obtain the reference labeled region comprises:

and adjusting the initial labeling area according to the reference adjustment information to obtain the reference labeling area.

4. The method of claim 3, wherein the generating reference adjustment information corresponding to the initial annotation region based on the slice image, the positive click data, and the negative click data comprises:

and inputting the slice image, the positive click data and the negative click data into a second adjustment information determination model to obtain the reference adjustment information output by the second adjustment information determination model.

5. The method according to any one of claims 1 to 4, wherein the obtaining of slice images of the object from the image to be annotated comprises:

determining a local image area of an object in the image to be annotated;

performing at least one expansion processing on the regional boundary of the local image region in the image to be annotated to obtain a slice image region framed and selected by the expanded regional boundary each time;

and generating the slice image according to the slice image area.

6. The method according to any one of claims 1 to 4, wherein after the adjusting the reference labeled region according to the region adjustment information to obtain the target labeled region, the method further comprises:

acquiring the classification information of the object in the image to be labeled;

and labeling the image to be labeled according to the target labeling area and the classification information to obtain a target labeling image.

7. An image annotation apparatus, comprising:

the first acquisition module is used for acquiring a slice image of an object according to an image to be marked;

the second acquisition module is used for acquiring a reference labeling area of an object in the image to be labeled according to the slice image;

the determining module is used for determining the area adjustment information of the reference labeling area according to the slice image;

the first processing module is used for adjusting the reference labeling area according to the area adjustment information to obtain a target labeling area, wherein the target labeling area is used for labeling the image to be labeled;

a first determination submodule for determining slice image characteristics of the slice image;

the second determining submodule is used for determining the area adjustment information of the reference labeling area according to the slice image and the characteristics of the slice image;

the second determining submodule is specifically configured to:

acquiring contour points of the object from the slice image;

determining a position feature of the contour point in the slice image;

acquiring contour point features corresponding to the contour points from the slice image features;

determining the region adjustment information of the reference labeling region according to the position feature, the contour point feature and the slice image;

the second determining submodule is further configured to:

determining a position adjustment direction and a position adjustment value corresponding to the position feature according to the contour point feature and the slice image;

taking the position adjustment direction and the position adjustment value as the area adjustment information;

the second obtaining module includes:

the obtaining sub-module is used for obtaining an initial labeling area of an object in the image to be labeled;

the processing submodule is used for acquiring positive click data and negative click data, wherein the positive click data is click data of which the click position is positioned in an object in the image to be marked, and the negative click data is click data of which the click position is positioned outside the object in the image to be marked;

8. The apparatus of claim 7, wherein the second determination submodule is further configured to:

9. The apparatus of claim 7, wherein the processing submodule is further operable to:

10. The apparatus of claim 9, wherein the processing submodule is further configured to:

11. The apparatus of any one of claims 7-10, wherein the first obtaining module is specifically configured to:

determining a local image area of an object in the image to be annotated;

and generating the slice image according to the slice image area.

12. The apparatus of any one of claims 7-10, further comprising:

the third acquisition module is used for acquiring the classification information of the object in the image to be labeled;

and the second processing module is used for labeling the image to be labeled according to the target labeling area and the classification information to obtain a target labeling image.

13. An electronic device, comprising:

at least one processor; and

a memory communicatively coupled to the at least one processor; wherein, the first and the second end of the pipe are connected with each other,

the memory stores instructions executable by the at least one processor to enable the at least one processor to perform the method of any one of claims 1-6.

14. A non-transitory computer readable storage medium having stored thereon computer instructions for causing the computer to perform the method of any one of claims 1-6.