CN116452957B - Quality detection method and device for image annotation data and electronic equipment - Google Patents
Quality detection method and device for image annotation data and electronic equipment Download PDFInfo
- Publication number
- CN116452957B CN116452957B CN202310733130.5A CN202310733130A CN116452957B CN 116452957 B CN116452957 B CN 116452957B CN 202310733130 A CN202310733130 A CN 202310733130A CN 116452957 B CN116452957 B CN 116452957B
- Authority
- CN
- China
- Prior art keywords
- image
- annotation data
- images
- processing model
- image processing
- Prior art date
- Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
- Active
Links
- 238000001514 detection method Methods 0.000 title claims abstract description 26
- 238000012545 processing Methods 0.000 claims abstract description 175
- 238000000034 method Methods 0.000 claims abstract description 45
- 238000012549 training Methods 0.000 claims abstract description 33
- 238000002372 labelling Methods 0.000 claims abstract description 28
- 230000008569 process Effects 0.000 claims abstract description 23
- 230000006870 function Effects 0.000 claims description 39
- 238000012795 verification Methods 0.000 claims description 9
- 238000004590 computer program Methods 0.000 description 6
- 238000004891 communication Methods 0.000 description 4
- 238000010586 diagram Methods 0.000 description 4
- 238000012986 modification Methods 0.000 description 3
- 230000004048 modification Effects 0.000 description 3
- 230000006978 adaptation Effects 0.000 description 2
- 230000004075 alteration Effects 0.000 description 2
- 230000001419 dependent effect Effects 0.000 description 2
- 238000013461 design Methods 0.000 description 2
- 230000003287 optical effect Effects 0.000 description 2
- 230000000644 propagated effect Effects 0.000 description 2
- 230000009286 beneficial effect Effects 0.000 description 1
- 238000013500 data storage Methods 0.000 description 1
- 238000005516 engineering process Methods 0.000 description 1
- 239000000835 fiber Substances 0.000 description 1
- 239000003550 marker Substances 0.000 description 1
- 230000008447 perception Effects 0.000 description 1
Classifications
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06V—IMAGE OR VIDEO RECOGNITION OR UNDERSTANDING
- G06V10/00—Arrangements for image or video recognition or understanding
- G06V10/98—Detection or correction of errors, e.g. by rescanning the pattern or by human intervention; Evaluation of the quality of the acquired patterns
- G06V10/993—Evaluation of the quality of the acquired pattern
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06V—IMAGE OR VIDEO RECOGNITION OR UNDERSTANDING
- G06V10/00—Arrangements for image or video recognition or understanding
- G06V10/70—Arrangements for image or video recognition or understanding using pattern recognition or machine learning
- G06V10/77—Processing image or video features in feature spaces; using data integration or data reduction, e.g. principal component analysis [PCA] or independent component analysis [ICA] or self-organising maps [SOM]; Blind source separation
- G06V10/771—Feature selection, e.g. selecting representative features from a multi-dimensional feature space
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06V—IMAGE OR VIDEO RECOGNITION OR UNDERSTANDING
- G06V10/00—Arrangements for image or video recognition or understanding
- G06V10/70—Arrangements for image or video recognition or understanding using pattern recognition or machine learning
- G06V10/77—Processing image or video features in feature spaces; using data integration or data reduction, e.g. principal component analysis [PCA] or independent component analysis [ICA] or self-organising maps [SOM]; Blind source separation
- G06V10/774—Generating sets of training patterns; Bootstrap methods, e.g. bagging or boosting
-
- Y—GENERAL TAGGING OF NEW TECHNOLOGICAL DEVELOPMENTS; GENERAL TAGGING OF CROSS-SECTIONAL TECHNOLOGIES SPANNING OVER SEVERAL SECTIONS OF THE IPC; TECHNICAL SUBJECTS COVERED BY FORMER USPC CROSS-REFERENCE ART COLLECTIONS [XRACs] AND DIGESTS
- Y02—TECHNOLOGIES OR APPLICATIONS FOR MITIGATION OR ADAPTATION AGAINST CLIMATE CHANGE
- Y02T—CLIMATE CHANGE MITIGATION TECHNOLOGIES RELATED TO TRANSPORTATION
- Y02T10/00—Road transport of goods or passengers
- Y02T10/10—Internal combustion engine [ICE] based vehicles
- Y02T10/40—Engine management systems
Abstract
The disclosure relates to a quality detection method and device for image annotation data and electronic equipment, wherein the method comprises the following steps: acquiring an image annotation data set and an initial image processing model; labeling data of a plurality of images in the image labeling data set are applied to the same image processing task; in the process of training an image processing model by adopting an image annotation data set, under the condition that the image processing model is in an under fitting state, respectively inputting a plurality of images into the image processing model to obtain prediction data of the plurality of images; then, selecting a problem image from the plurality of images by combining the labeling data of the plurality of images and the loss function; the problem images in the image annotation data set and the annotation data of the problem images are deleted, wherein the problem images are images with wrong annotation data or images with correct annotation data but difficult to learn, and the annotation data of the problem images are deleted, so that the quality of the annotation data in the image annotation data set can be improved.
Description
Technical Field
The disclosure relates to the technical field of automatic driving and intelligent perception, in particular to a quality detection method and device for image annotation data and electronic equipment.
Background
Currently, in an automatic driving system, the execution accuracy of each image processing task depends on the accuracy of an image processing model in the image processing task; the accuracy of the image processing model is greatly dependent on the quality of the labeling data in the image labeling data set during training.
However, the labeling data in the image labeling data set is mainly obtained by manual labeling, the quality of the labeling data is difficult to control, the accuracy of an image processing model is low, and the accuracy of an image processing task is low.
Disclosure of Invention
The disclosure provides a quality detection method and device for image annotation data and electronic equipment.
According to a first aspect of embodiments of the present disclosure, there is provided a quality detection method of image annotation data, the method including: acquiring an image annotation data set and an initial image processing model; the image annotation data set comprises annotation data of each image in the plurality of images; the plurality of annotation data are applied to the same image processing task; in the process of training the initial image processing model by adopting the image annotation data set, under the condition that the image processing model is in an under-fitting state, respectively inputting a plurality of images into the image processing model to acquire prediction data of each image in the plurality of images; selecting a problem image from the plurality of images based on the plurality of annotation data, the plurality of prediction data, and the loss function; and deleting the problem image in the image annotation data set and the annotation data of the problem image.
In one embodiment of the present disclosure, the selecting the problem image from the plurality of images according to the plurality of annotation data, the plurality of prediction data, and the loss function includes: determining a loss function value of each of a plurality of images according to the labeling data of the image, the prediction data of the image and the loss function; and selecting an image with the largest loss function value from the plurality of images as a problem image according to the plurality of loss function values.
In one embodiment of the present disclosure, the method further comprises: reporting the problem image to obtain a processing result of the problem image; performing a first marking process on the problem image in a case where the processing result indicates that the problem image is difficult to learn; the first mark is used for indicating the image processing model to perform key training processing on the problem image; and updating the problem image after the marking processing and the marking data of the problem image into the image marking data set.
In one embodiment of the present disclosure, before deleting the problem image in the image annotation data set and the annotation data of the problem image, the method further comprises: determining that the problem image is not provided with a first mark; the first mark is used for indicating the image processing model to perform key training processing on the problem image.
In one embodiment of the present disclosure, the method further comprises: and under the condition that the image processing model is not in an under-fitting state, adjusting the learning rate of the image processing model until the image processing model is in the under-fitting state.
In one embodiment of the present disclosure, the image processing task includes at least one of: a vehicle positioning task, an obstacle positioning task, a road detection task, and a traffic light identification task.
According to a second aspect of embodiments of the present disclosure, there is further provided a quality detection apparatus of image annotation data, the apparatus including: the first acquisition module is used for acquiring an image annotation data set and an initial image processing model; the image annotation data set comprises annotation data of each image in the plurality of images; the plurality of annotation data are applied to the same image processing task; the second acquisition module is used for respectively inputting a plurality of images into the image processing model under the condition that the image processing model is in an under-fitting state in the process of training the initial image processing model by adopting the image annotation data set, and acquiring prediction data of each image in the plurality of images; a selection module for selecting a problem image from the plurality of images based on the plurality of annotation data, the plurality of prediction data, and the loss function; and the processing module is used for deleting the problem image in the image annotation data set and the annotation data of the problem image.
According to a third aspect of embodiments of the present disclosure, there is also provided an electronic device, including: a processor; a memory for storing the processor-executable instructions; wherein the processor is configured to: the method for detecting the quality of the image annotation data is realized.
According to a fourth aspect of embodiments of the present disclosure, there is also provided a non-transitory computer-readable storage medium, which when executed by a processor, causes the processor to perform the quality detection method of image annotation data as described above.
The technical scheme provided by the embodiment of the disclosure at least brings the following beneficial effects:
acquiring an image annotation data set and an initial image processing model; the image annotation data set comprises annotation data of each of the plurality of images; the plurality of annotation data are applied to the same image processing task; in the process of training an initial image processing model by adopting an image annotation data set, under the condition that the image processing model is in an under fitting state, respectively inputting a plurality of images into the image processing model to obtain prediction data of each image in the plurality of images; selecting a problem image from the plurality of images based on the plurality of annotation data, the plurality of prediction data, and the loss function; deleting the problem images in the image annotation data set and the annotation data of the problem images, wherein the annotation data of the non-problem images can be quickly learned by the image processing model; the problem image which is difficult to quickly learn by the image processing model is an image with wrong annotation data or an image with correct annotation data but difficult to learn, and the annotation data of the problem image and the problem image are deleted, so that the quality of the annotation data in the image annotation data set can be improved, and the accuracy of the model obtained based on the training of the image annotation data set is further improved.
It is to be understood that both the foregoing general description and the following detailed description are exemplary and explanatory only and are not restrictive of the disclosure.
Drawings
The accompanying drawings, which are incorporated in and constitute a part of this specification, illustrate embodiments consistent with the disclosure and together with the description, serve to explain the principles of the disclosure and do not constitute an undue limitation on the disclosure.
FIG. 1 is a flow chart of a method for quality detection of image annotation data according to one embodiment of the disclosure;
FIG. 2 is a schematic structural diagram of a quality detection apparatus for image annotation data according to an embodiment of the present disclosure;
fig. 3 is a block diagram of an electronic device, according to an exemplary embodiment of the present disclosure.
Detailed Description
In order to enable those skilled in the art to better understand the technical solutions of the present disclosure, the technical solutions of the embodiments of the present disclosure will be clearly and completely described below with reference to the accompanying drawings.
It should be noted that the terms "first," "second," and the like in the description and claims of the present disclosure and in the foregoing figures are used for distinguishing between similar objects and not necessarily for describing a particular sequential or chronological order. It is to be understood that the data so used may be interchanged where appropriate such that the embodiments of the disclosure described herein may be capable of operation in sequences other than those illustrated or described herein. The implementations described in the following exemplary examples are not representative of all implementations consistent with the present disclosure. Rather, they are merely examples of apparatus and methods consistent with some aspects of the present disclosure as detailed in the accompanying claims.
Currently, in an automatic driving system, the execution accuracy of each image processing task depends on the accuracy of an image processing model in the image processing task; the accuracy of the image processing model is greatly dependent on the quality of the labeling data in the image labeling data set during training.
However, the labeling data in the image labeling data set is mainly obtained by manual labeling, the quality of the labeling data is difficult to control, the accuracy of an image processing model is low, and the accuracy of an image processing task is low.
Fig. 1 is a flowchart of a quality detection method of image annotation data according to an embodiment of the present disclosure. It should be noted that, the method for detecting the quality of the image annotation data according to the present embodiment may be applied to a device for detecting the quality of the image annotation data, where the device may be configured in an electronic device, so that the electronic device may perform a function of detecting the quality of the image annotation data.
The electronic device may be any device with computing capability, for example, may be a personal computer (Personal Computer, abbreviated as PC), a mobile terminal, a server, etc., and the mobile terminal may be, for example, a vehicle-mounted device, a mobile phone, a tablet computer, a personal digital assistant, a wearable device, etc., and may be a hardware device with various operating systems, a touch screen, and/or a display screen. In the following embodiments, an execution body is described as an example of an electronic device.
As shown in fig. 1, the method comprises the steps of:
step 101, acquiring an image annotation data set and an initial image processing model; the image annotation data set comprises annotation data of each of the plurality of images; multiple annotation data are applied to the same image processing task.
In the embodiment of the present disclosure, the image processing task may be a task related to image processing in various technical fields. Technical fields such as the automatic driving technical field, the intelligent sensing technical field, etc. In the field of autopilot technology, the image processing task may include at least one of: a vehicle positioning task, an obstacle positioning task, a road detection task, and a traffic light identification task. Here, the specific description of the image processing task is not limited, and may be extended according to actual needs.
In different image processing tasks, the images in the image annotation data set may be different, and the annotation data of the images may be different. For example, in a vehicle positioning task, the image in the image annotation dataset may be a vehicle surrounding image or a vehicle image; the labeling data of the image may be position data of the vehicle, position data of an object around the vehicle, or the like. For another example, in the obstacle locating task, the image in the image annotation dataset may be a vehicle front image or a vehicle surrounding image; the annotation data of the image may be position data of an obstacle in the image, etc.
The plurality of images in the image annotation data set and the annotation data of each image in the plurality of images can be used for training or checking an image processing model in an image processing task. The image annotation data may also be referred to as image annotation data.
Wherein the plurality of images may refer to all images in the image annotation dataset. For example, assume that there are 10 images in the image annotation dataset; correspondingly, a plurality of images may refer to all 10 images in the image annotation dataset.
The plurality of annotation data may refer to all the annotation data in the image annotation data set. For example, assuming that there are 10 images in the image annotation data set, one for each image, there are 10 annotation data in the image annotation data set; correspondingly, the plurality of annotation data may refer to all 10 annotation data in the image annotation data set.
Step 102, in the process of training an initial image processing model by adopting an image annotation data set, under the condition that the image processing model is in an under fitting state, respectively inputting a plurality of images into the image processing model to obtain the prediction data of each image in the plurality of images.
In the embodiment of the present disclosure, the training of the initial image processing model by using the image labeling data set may be, for example, inputting an image in the image labeling data set into the image processing model, and obtaining prediction data of an image output by the image processing model; combining the predicted data of the image, the labeling data of the image and the loss function to determine a loss function value; and carrying out parameter adjustment processing on the image processing model by combining the loss function value to realize training.
The electronic device may further obtain a verification data set corresponding to the image annotation data set. Correspondingly, in the training process of the image processing model, the electronic equipment can determine the processing accuracy of the image processing model on the image annotation data set and the verification accuracy of the image processing model on the verification data set; and determining whether the image processing model is in a lack-fitting state according to the processing accuracy, the verification accuracy and the accuracy threshold.
Wherein the accuracy threshold may be preset. And determining that the image processing model is in an under-fitting state under the condition that the processing accuracy is smaller than or equal to an accuracy threshold and the verification accuracy is smaller than or equal to the accuracy threshold.
The electronic device may also perform the following process: and under the condition that the image processing model is not in the under-fitting state, adjusting the learning rate of the image processing model until the image processing model is in the under-fitting state. After the learning rate of the image processing model is adjusted, training of the image processing model can be continued, and whether the image processing model is in an under-fitting state or not can be determined in real time.
Step 103, selecting a problem image from the plurality of images according to the plurality of labeling data, the plurality of prediction data and the loss function.
In the embodiment of the present disclosure, the electronic device performs the process of step 103 may, for example, determine, for each of the plurality of images, a loss function value of the image according to labeling data of the image, prediction data of the image, and a loss function; and selecting an image with the largest loss function value from the plurality of images as a problem image according to the plurality of loss function values.
The electronic equipment can sort the images in a descending order according to the loss function values to obtain a sorting result; and taking the image which is the forefront in the sorting result as a problem image. In addition, in order to further improve the quality of the labeling data in the image labeling data set, the electronic device may further use a plurality of images in front of the sorting result as problem images.
And step 104, deleting the problem image in the image annotation data set and the annotation data of the problem image.
In the embodiment of the disclosure, since there are two kinds of problem images, one is an image with wrong annotation data, and the other is an image which is difficult to learn. Thus, to ensure that the image processing model is able to learn the second problem image, the electronic device may further perform the following process after step 103: reporting the problem image to obtain a processing result of the problem image; performing first marking processing on the problem image in the case that the processing result indicates that the problem image is difficult to learn; the first mark is used for indicating the image processing model to carry out key training processing on the problem image; and updating the problem image after the marking processing and the marking data of the problem image into an image marking data set.
When the processing result indicates that the problem image annotation data is incorrect, the processing of the problem image is stopped. The processing result indicates that the problem image with the wrong label data of the problem image is a badcase; the processing result indicates a problem image that is difficult to learn, and may be hardcase, for example.
The image processing model may perform multiple training processes on an image in which a first marker is set in the image annotation data set, for example, to increase the probability that the image is sampled, so that the image may be sampled multiple times and used to train the image processing model.
In addition, in the case of performing the first marking process on the problem image, the corresponding electronic device may further perform the following steps before performing step 104: determining that the problem image is not provided with a first mark; the first mark is used for indicating the image processing model to carry out key training processing on the problem image. If the problem image is provided with the first mark, the electronic device may stop deleting the problem image and the mark data of the problem image.
In an embodiment of the present disclosure, the electronic device may repeat the above steps 101 to 104 for the image annotation dataset until at least one of the following conditions is met: the problem images which reach the preset execution times and are selected continuously for many times are difficult to learn, the number of the deleted problem images is larger than or equal to a preset number threshold, the duty ratio of the deleted problem images is larger than or equal to a preset duty ratio threshold, and the like. The conditions herein may be set according to actual needs, and are not particularly limited in this embodiment.
In the quality detection method of the image annotation data, an image annotation data set and an initial image processing model are obtained; the image annotation data set comprises annotation data of each of the plurality of images; the plurality of annotation data are applied to the same image processing task; in the process of training an initial image processing model by adopting an image annotation data set, under the condition that the image processing model is in an under fitting state, respectively inputting a plurality of images into the image processing model to obtain prediction data of each image in the plurality of images; selecting a problem image from the plurality of images based on the plurality of annotation data, the plurality of prediction data, and the loss function; deleting the problem images in the image annotation data set and the annotation data of the problem images, wherein the annotation data of the non-problem images can be quickly learned by the image processing model; the problem image which is difficult to quickly learn by the image processing model is an image with wrong annotation data or an image with correct annotation data but difficult to learn, and the annotation data of the problem image and the problem image are deleted, so that the quality of the annotation data in the image annotation data set can be improved, and the accuracy of the model obtained based on the training of the image annotation data set is further improved.
Fig. 2 is a schematic structural diagram of a quality detection device for image labeling data according to an embodiment of the disclosure.
As shown in fig. 2, the quality detection apparatus for image annotation data may include: a first acquisition module 201, a second acquisition module 202, a selection module 203 and a processing module 204;
the first acquiring module 201 is configured to acquire an image annotation dataset and an initial image processing model; the image annotation data set comprises annotation data of each image in the plurality of images; the plurality of annotation data are applied to the same image processing task;
a second obtaining module 202, configured to, in a process of training the initial image processing model by using the image annotation data set, respectively input a plurality of images into the image processing model when the image processing model is in a lack-fitting state, and obtain prediction data of each image in the plurality of images;
a selection module 203 for selecting a problem image from the plurality of images based on the plurality of annotation data, the plurality of prediction data, and the loss function;
the processing module 204 is configured to delete the problem image in the image annotation data set and annotation data of the problem image.
In one embodiment of the disclosure, the selecting module 203 is specifically configured to determine, for each image of a plurality of images, a loss function value of the image according to labeling data of the image, prediction data of the image, and the loss function; and selecting an image with the largest loss function value from the plurality of images as a problem image according to the plurality of loss function values.
In one embodiment of the present disclosure, the apparatus further comprises: a reporting module and an updating module; the reporting module is used for reporting the problem image to acquire a processing result of the problem image; the processing module 204 is further configured to perform a first marking process on the problem image if the processing result indicates that the problem image is difficult to learn; the first mark is used for indicating the image processing model to perform key training processing on the problem image; the updating module is used for updating the problem image after the marking processing and the marking data of the problem image into the image marking data set.
In one embodiment of the disclosure, the processing module 204 is further configured to determine that the problem image is not set with a first flag before deleting the problem image in the image annotation data set and the annotation data of the problem image; the first mark is used for indicating the image processing model to perform key training processing on the problem image.
In one embodiment of the present disclosure, the apparatus further comprises: and the adjusting module is used for adjusting the learning rate of the image processing model until the image processing model is in an under-fitting state under the condition that the image processing model is not in the under-fitting state.
In one embodiment of the present disclosure, the image processing task includes at least one of: a vehicle positioning task, an obstacle positioning task, a road detection task, and a traffic light identification task.
In the quality detection device of the image annotation data, an image annotation data set and an initial image processing model are obtained; the image annotation data set comprises annotation data of each of the plurality of images; the plurality of annotation data are applied to the same image processing task; in the process of training an initial image processing model by adopting an image annotation data set, under the condition that the image processing model is in an under fitting state, respectively inputting a plurality of images into the image processing model to obtain prediction data of each image in the plurality of images; selecting a problem image from the plurality of images based on the plurality of annotation data, the plurality of prediction data, and the loss function; deleting the problem images in the image annotation data set and the annotation data of the problem images, wherein the annotation data of the non-problem images can be quickly learned by the image processing model; the problem image which is difficult to quickly learn by the image processing model is an image with wrong annotation data or an image with correct annotation data but difficult to learn, and the annotation data of the problem image and the problem image are deleted, so that the quality of the annotation data in the image annotation data set can be improved, and the accuracy of the model obtained based on the training of the image annotation data set is further improved.
According to a third aspect of embodiments of the present disclosure, there is also provided an electronic device, including: a processor; a memory for storing processor-executable instructions, wherein the processor is configured to: the quality detection method of the image annotation data is realized.
In order to implement the above-described embodiments, the present disclosure also proposes a storage medium.
Wherein the instructions in the storage medium, when executed by the processor, enable the processor to perform the quality detection method of image annotation data as described above.
To achieve the above embodiments, the present disclosure also provides a computer program product.
Wherein the computer program product, when executed by a processor of an electronic device, enables the electronic device to perform the method as above.
Fig. 3 is a block diagram of an electronic device, according to an example embodiment. The electronic device shown in fig. 3 is merely an example and should not be construed to limit the functionality and scope of use of the disclosed embodiments.
As shown in fig. 3, the electronic device 1000 includes a processor 111 that can perform various appropriate actions and processes according to a program stored in a Read Only Memory (ROM) 112 or a program loaded from a Memory 116 into a random access Memory (RAM, random Access Memory) 113. In the RAM 113, various programs and data required for the operation of the electronic apparatus 1000 are also stored. The processor 111, the ROM 112, and the RAM 113 are connected to each other through a bus 114. An Input/Output (I/O) interface 115 is also connected to bus 114.
The following components are connected to the I/O interface 115: a memory 116 including a hard disk and the like; and a communication section 117 including a network interface card such as a local area network (Local Area Network, LAN) card, a modem, or the like, the communication section 117 performing communication processing via a network such as the internet; the drive 118 is also connected to the I/O interface 115 as needed.
In particular, according to embodiments of the present disclosure, the processes described above with reference to flowcharts may be implemented as computer software programs. For example, embodiments of the present disclosure include a computer program embodied on a computer readable medium, the computer program containing program code for performing the methods shown in the flowcharts. In such an embodiment, the computer program may be downloaded and installed from the network through the communication section 117. The above-described functions defined in the methods of the present disclosure are performed when the computer program is executed by the processor 111.
In an exemplary embodiment, a storage medium is also provided, such as a memory, comprising instructions executable by the processor 111 of the electronic device 1000 to perform the above-described method. Alternatively, the storage medium may be a non-transitory computer readable storage medium, for example, a ROM, random Access Memory (RAM), CD-ROM, magnetic tape, floppy disk, optical data storage device, and the like.
In the context of this disclosure, a computer-readable storage medium may be any tangible medium that can contain, or store a program for use by or in connection with an instruction execution system, apparatus, or device. In the present disclosure, however, the computer-readable signal medium may include a data signal propagated in baseband or as part of a carrier wave, with the computer-readable program code embodied therein. Such a propagated data signal may take any of a variety of forms, including, but not limited to, electro-magnetic, optical, or any suitable combination of the foregoing. A computer readable signal medium may also be any computer readable medium that is not a computer readable storage medium and that can communicate, propagate, or transport a program for use by or in connection with an instruction execution system, apparatus, or device. Program code embodied on a computer readable medium may be transmitted using any appropriate medium, including but not limited to: wireless, wire, fiber optic cable, RF, etc., or any suitable combination of the foregoing.
Furthermore, the word "exemplary" is used herein to mean serving as an example, instance, illustration. Any aspect or design described herein as "exemplary" is not necessarily to be construed as advantageous over other aspects or designs. Rather, the use of the word exemplary is intended to present concepts in a concrete fashion. As used herein, the term "or" is intended to mean an inclusive "or" rather than an exclusive "or". That is, unless specified otherwise, or clear from context, "X application a or B" is intended to mean any one of the natural inclusive permutations. I.e. if X applies a; x is applied with B; or both X applications a and B, "X application a or B" is satisfied under any of the foregoing examples. In addition, the articles "a" and "an" as used in this application and the appended claims are generally understood to mean "one or more" unless specified otherwise or clear from context to be directed to a singular form.
Also, although the disclosure has been shown and described with respect to one or more implementations, equivalent alterations and modifications will occur to others skilled in the art upon the reading and understanding of this specification and the annexed drawings. The present disclosure includes all such modifications and alterations and is limited only by the scope of the claims. In particular regard to the various functions performed by the above described components (e.g., elements, resources, etc.), the terms used to describe such components are intended to correspond, unless otherwise indicated, to any component which performs the specified function of the described component (which is functionally equivalent), even though not structurally equivalent to the disclosed structure. In addition, while a particular feature of the disclosure may have been disclosed with respect to only one of several implementations, such feature may be combined with one or more other features of the other implementations as may be desired and advantageous for any given or particular application. Furthermore, to the extent that the terms "includes," including, "" has, "" having, "or variants thereof are used in either the detailed description or the claims, such terms are intended to be inclusive in a manner similar to the term" comprising.
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 application is intended to cover any adaptations, 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 is to be understood that the present disclosure is not limited to the precise arrangements and instrumentalities shown in the drawings, and that various modifications and changes may be effected without departing from the scope thereof. The scope of the present disclosure is limited only by the appended claims.
Claims (12)
1. A method for quality detection of image annotation data, the method comprising:
acquiring an image annotation data set and an initial image processing model; the image annotation data set comprises annotation data of each image in the plurality of images; the plurality of annotation data are applied to the same image processing task;
in the process of training the initial image processing model by adopting the image annotation data set, under the condition that the image processing model is in an under-fitting state, respectively inputting a plurality of images into the image processing model to acquire prediction data of each image in the plurality of images;
selecting a problem image from a plurality of images based on a plurality of annotation data, a plurality of prediction data for the images, and a loss function;
deleting the problem image in the image annotation data set and the annotation data of the problem image;
under the condition that the image processing model is not in an under-fitting state, adjusting the learning rate of the image processing model until the image processing model is in the under-fitting state;
wherein determining that the image processing model is in an under-fit state comprises: acquiring a verification data set corresponding to the image annotation data set, determining the processing accuracy of the image processing model on the image annotation data set in the training process of the image processing model, and determining that the image processing model is in an under-fitting state under the condition that the processing accuracy is smaller than or equal to an accuracy threshold and the verification accuracy is smaller than or equal to the accuracy threshold.
2. The method of claim 1, wherein selecting the problem image from the plurality of images based on the plurality of annotation data, the plurality of prediction data, and the loss function comprises:
determining a loss function value of each of a plurality of images according to the labeling data of the image, the prediction data of the image and the loss function;
and selecting an image with the largest loss function value from the plurality of images as a problem image according to the plurality of loss function values.
3. The method according to claim 1, wherein the method further comprises:
reporting the problem image to obtain a processing result of the problem image;
performing a first marking process on the problem image in a case where the processing result indicates that the problem image is difficult to learn; the first mark is used for indicating the image processing model to perform key training processing on the problem image;
and updating the problem image after the marking processing and the marking data of the problem image into the image marking data set.
4. A method according to claim 1 or 3, wherein prior to deleting the problem image in the image annotation data set and the annotation data for the problem image, the method further comprises:
determining that the problem image is not provided with a first mark; the first mark is used for indicating the image processing model to perform key training processing on the problem image.
5. The method of claim 1, wherein the image processing task comprises at least one of: a vehicle positioning task, an obstacle positioning task, a road detection task, and a traffic light identification task.
6. A quality detection apparatus for image annotation data, the apparatus comprising:
the first acquisition module is used for acquiring an image annotation data set and an initial image processing model; the image annotation data set comprises annotation data of each image in the plurality of images; the plurality of annotation data are applied to the same image processing task;
the second acquisition module is used for respectively inputting a plurality of images into the image processing model under the condition that the image processing model is in an under-fitting state in the process of training the initial image processing model by adopting the image annotation data set, and acquiring prediction data of each image in the plurality of images;
a selection module for selecting a problem image from a plurality of images based on a plurality of annotation data, a plurality of prediction data for the images, and a loss function;
the processing module is used for deleting the problem image in the image annotation data set and the annotation data of the problem image;
the adjusting module is used for adjusting the learning rate of the image processing model until the image processing model is in an under-fitting state under the condition that the image processing model is not in the under-fitting state;
wherein determining that the image processing model is in an under-fit state comprises: acquiring a verification data set corresponding to the image annotation data set, determining the processing accuracy of the image processing model on the image annotation data set in the training process of the image processing model, and determining that the image processing model is in an under-fitting state under the condition that the processing accuracy is smaller than or equal to an accuracy threshold and the verification accuracy is smaller than or equal to the accuracy threshold.
7. The apparatus of claim 6, wherein the selection module is configured to,
determining a loss function value of each of a plurality of images according to the labeling data of the image, the prediction data of the image and the loss function;
and selecting an image with the largest loss function value from the plurality of images as a problem image according to the plurality of loss function values.
8. The apparatus of claim 6, wherein the apparatus further comprises: a reporting module and an updating module;
the reporting module is used for reporting the problem image to acquire a processing result of the problem image;
the processing module is further used for performing first marking processing on the problem image when the processing result indicates that the problem image is difficult to learn; the first mark is used for indicating the image processing model to perform key training processing on the problem image;
the updating module is used for updating the problem image after the marking processing and the marking data of the problem image into the image marking data set.
9. The apparatus of claim 6 or 8, wherein the processing module is further configured to determine that the problem image is not provided with a first flag prior to deleting the problem image in the image annotation data set and the annotation data for the problem image; the first mark is used for indicating the image processing model to perform key training processing on the problem image.
10. The apparatus of claim 6, wherein the image processing task comprises at least one of: a vehicle positioning task, an obstacle positioning task, a road detection task, and a traffic light identification task.
11. An electronic device, comprising:
a processor;
a memory for storing the processor-executable instructions;
wherein the processor is configured to:
a step of implementing the quality detection method of image annotation data according to any one of claims 1 to 5.
12. A non-transitory computer readable storage medium, which when executed by a processor, causes the processor to perform the quality detection method of image annotation data according to any of claims 1 to 5.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202310733130.5A CN116452957B (en) | 2023-06-20 | 2023-06-20 | Quality detection method and device for image annotation data and electronic equipment |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202310733130.5A CN116452957B (en) | 2023-06-20 | 2023-06-20 | Quality detection method and device for image annotation data and electronic equipment |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN116452957A CN116452957A (en) | 2023-07-18 |
| CN116452957B true CN116452957B (en) | 2023-09-26 |
Family
ID=87125949
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN202310733130.5A Active CN116452957B (en) | 2023-06-20 | 2023-06-20 | Quality detection method and device for image annotation data and electronic equipment |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN116452957B (en) |
Citations (4)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| WO2020098539A1 (en) * | 2018-11-13 | 2020-05-22 | 腾讯科技（深圳）有限公司 | Image processing method and apparatus, computer readable medium, and electronic device |
| CN111931863A (en) * | 2020-09-15 | 2020-11-13 | 深圳追一科技有限公司 | Method and device for screening mislabeling data and computer storage medium |
| CN114419397A (en) * | 2022-01-20 | 2022-04-29 | 中山大学·深圳 | Data set construction method and device based on data cleaning and data generation |
| WO2022095258A1 (en) * | 2020-11-03 | 2022-05-12 | 上海商汤智能科技有限公司 | Image object classification method and apparatus, device, storage medium and program |
Family Cites Families (1)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| US20220036126A1 (en) * | 2020-07-30 | 2022-02-03 | Toyota Research Institute, Inc. | System and method for training of a detector model to output an instance identifier indicating object consistency along the temporal axis |
-
2023
- 2023-06-20 CN CN202310733130.5A patent/CN116452957B/en active Active
Patent Citations (4)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| WO2020098539A1 (en) * | 2018-11-13 | 2020-05-22 | 腾讯科技（深圳）有限公司 | Image processing method and apparatus, computer readable medium, and electronic device |
| CN111931863A (en) * | 2020-09-15 | 2020-11-13 | 深圳追一科技有限公司 | Method and device for screening mislabeling data and computer storage medium |
| WO2022095258A1 (en) * | 2020-11-03 | 2022-05-12 | 上海商汤智能科技有限公司 | Image object classification method and apparatus, device, storage medium and program |
| CN114419397A (en) * | 2022-01-20 | 2022-04-29 | 中山大学·深圳 | Data set construction method and device based on data cleaning and data generation |
Also Published As
| Publication number | Publication date |
|---|---|
| CN116452957A (en) | 2023-07-18 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN109117831B (en) | Training method and device of object detection network | |
| CN111860300A (en) | Key point detection method and device, terminal equipment and storage medium | |
| CN112749081B (en) | User interface testing method and related device | |
| CN108427941B (en) | Method for generating face detection model, face detection method and device | |
| CN111784774A (en) | Target detection method and device, computer readable medium and electronic equipment | |
| CN110287817B (en) | Target recognition and target recognition model training method and device and electronic equipment | |
| CN113592033B (en) | Oil tank image recognition model training method, oil tank image recognition method and device | |
| CN112381092A (en) | Tracking method, device and computer readable storage medium | |
| CN111340015A (en) | Positioning method and device | |
| CN109064464B (en) | Method and device for detecting burrs of battery pole piece | |
| CN114359681A (en) | Image recognition model training method, image recognition device, image recognition equipment and image recognition medium | |
| CN112712795B (en) | Labeling data determining method, labeling data determining device, labeling data determining medium and electronic equipment | |
| CN112330602B (en) | Intelligent trapping direction judgment method and device | |
| CN113409393B (en) | Method and device for identifying traffic sign | |
| CN116452957B (en) | Quality detection method and device for image annotation data and electronic equipment | |
| CN112633341A (en) | Interface testing method and device, computer equipment and storage medium | |
| CN114998570B (en) | Method and device for determining object detection frame, storage medium and electronic device | |
| CN113255819B (en) | Method and device for identifying information | |
| CN115861809A (en) | Rod detection and training method and device for model thereof, electronic equipment and medium | |
| CN113033539A (en) | Calligraphy practicing grid detection method and device, readable medium and electronic equipment | |
| CN112016513A (en) | Video semantic segmentation method, model training method, related device and electronic equipment | |
| CN112749293A (en) | Image classification method and device and storage medium | |
| CN112434591B (en) | Lane line determination method and device | |
| CN116563840B (en) | Scene text detection and recognition method based on weak supervision cross-mode contrast learning | |
| CN109492695B (en) | Sample processing method and device for data modeling, electronic equipment and readable medium |
Legal Events
| Date | Code | Title | Description |
|---|---|---|---|
| PB01 | Publication | ||
| PB01 | Publication | ||
| SE01 | Entry into force of request for substantive examination | ||
| SE01 | Entry into force of request for substantive examination | ||
| GR01 | Patent grant | ||
| GR01 | Patent grant |