Target detection model training method and device, readable storage medium and electronic equipment
By screening and clustering the reflow dataset to remove the labeled noise data, the problem of low accuracy of existing target detection models in complex scenarios is solved, and the efficiency and accuracy of model training are achieved.
Patent Information
- Application Number
- CN202510637984.2
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- Filing Date
- 2025-05-16
- Publication Date
- 2025-09-23
AI Technical Summary
Existing object detection models have low accuracy in practical applications due to the influence of annotation noise in reflow datasets in complex scenarios.
By filtering the noise data of the reflow dataset, removing the labeled noise data, using the labeled information and model prediction information for clustering comparison, filtering out the noise dataset, and training the model based on the filtered dataset, a higher quality target detection model is obtained.
The accuracy of the target detection model in practical applications is improved, ensuring that the model learns higher quality data features.
Smart Images
Figure CN120689875A_ABST
Abstract
Description
Technical Field
[0001] The present application belongs to the field of target detection technology, and in particular relates to a target detection model training method, device, computer-readable storage medium and electronic device. Background Art
[0002] With the continuous development of artificial intelligence (AI) vision technology, sweepers have become increasingly capable of perceiving the environment and autonomous navigation. In particular, in obstacle detection tasks, AI vision models are widely used to identify and avoid obstacles. Existing sweeper systems can collect visual data of the surrounding environment through image sensors and use target detection algorithms to detect and classify obstacles in real time. However, sweepers often face the emergence of new data in complex actual scenarios. These new data may contain previously unseen obstacle types or environmental changes, resulting in a decrease in the detection capabilities of existing models. In order to maintain the efficiency and accuracy of the model, the data reflow process of the sweeper is very critical, that is, the model is optimized and trained using the newly collected data set (i.e., the reflow data set).
[0003] During the annotation process of reflow datasets, due to manual intervention or the limitations of automatic annotation systems, a certain degree of annotation noise often occurs in the dataset, which can negatively impact the training process. The model may learn incorrect features from incorrect annotations, resulting in low accuracy in practical applications. Summary of the Invention
[0004] In view of this, the embodiments of the present application provide a target detection model training method, device, computer-readable storage medium and electronic device to solve the problem that the trained model of the existing target detection model training method has low accuracy in practical applications.
[0005] A first aspect of an embodiment of the present application provides a method for training a target detection model, which may include:
[0006] In a specific implementation of the first aspect, the training the first object detection model based on the joint training dataset to obtain the trained second object detection model may include:
[0007] Obtain a first target detection model; wherein the first target detection model is a target detection model trained based on an existing labeled dataset;
[0008] Obtaining a reflow dataset and determining labeling information for each sample in the reflow dataset;
[0009] Performing target detection on each sample in the reflow dataset using the first target detection model to obtain model prediction information for each sample in the reflow dataset;
[0010] Screening noise data in the reflux dataset according to the annotation information and the model prediction information to obtain a noise dataset;
[0011] The first target detection model is trained based on the filtered reflow dataset to obtain a second target detection model; wherein the filtered reflow dataset is a dataset obtained after removing the noise dataset from the reflow dataset.
[0012] In a specific implementation of the first aspect, filtering noise data in the reflow dataset according to the annotation information and the model prediction information to obtain the noise dataset may include:
[0013] Clustering the annotation boxes in the annotation information and the prediction boxes in the model prediction information to obtain clusters;
[0014] Comparing the annotation information and the model prediction information in each cluster to obtain a cluster comparison result;
[0015] Determining each noise data in the reflow dataset according to the cluster comparison result;
[0016] The noise data set is determined according to each noise data in the reflow data set.
[0017] In a specific implementation of the first aspect, clustering the annotation boxes in the annotation information and the prediction boxes in the model prediction information to obtain clusters may include:
[0018] Determining an intersection-over-union ratio between the labeled box and the predicted box;
[0019] The labeled box and the predicted box are clustered according to an intersection-over-union ratio between the labeled box and the predicted box to obtain clusters.
[0020] In a specific implementation of the first aspect, clustering the labeled box and the predicted box according to the intersection-over-union ratio between the labeled box and the predicted box to obtain clusters may include:
[0021] When an intersection-over-union (IoU) ratio between the marked frame and the predicted frame is greater than a preset IoU threshold, determining the marked frame and the predicted frame as belonging to the same cluster;
[0022] When an intersection-over-union (IoU) ratio between the labeled box and the predicted box is less than or equal to the IoU threshold, the labeled box and the predicted box are determined to be in different clusters.
[0023] In a specific implementation of the first aspect, determining the noise dataset according to each noise data in the reflow dataset may include:
[0024] Determine the noise score of each noise data separately;
[0025] The noise data set is composed of several noise data with the highest noise scores.
[0026] In a specific implementation of the first aspect, determining the noise score of each noise data may include:
[0027] A noise score of each noise data is determined according to the confidence of the prediction box; wherein the noise score is positively correlated with the confidence.
[0028] In a specific implementation of the first aspect, after training the first object detection model based on the filtered reflow dataset to obtain the second object detection model, the method may further include:
[0029] Obtain the image to be detected;
[0030] Use the second target detection model to perform target detection on the image to be detected to obtain a target detection result of the image to be detected.
[0031] A second aspect of the embodiments of the present application provides a target detection model training device, which may include:
[0032] A model acquisition module is used to acquire a first target detection model; wherein the first target detection model is a target detection model trained based on an existing labeled data set;
[0033] a labeling information determination module, configured to obtain a reflow dataset and determine labeling information for each sample in the reflow dataset;
[0034] a model prediction information determination module, configured to perform target detection on each sample in the reflow dataset using the first target detection model to obtain model prediction information for each sample in the reflow dataset;
[0035] A noise data screening module, configured to screen noise data in the return data set according to the annotation information and the model prediction information to obtain a noise data set;
[0036] A model training module is used to perform model training on the first target detection model based on the filtered reflow data set to obtain a second target detection model; wherein the filtered reflow data set is a data set obtained after removing the noise data set from the reflow data set.
[0037] In a specific implementation of the second aspect, the noise data screening module may include:
[0038] A clustering submodule, configured to cluster the annotation boxes in the annotation information and the prediction boxes in the model prediction information to obtain clusters;
[0039] An information comparison submodule, used to compare the annotation information and the model prediction information in each cluster to obtain a cluster comparison result;
[0040] A noise data determination submodule, configured to determine each noise data in the reflow dataset according to the cluster comparison result;
[0041] The noise data set determining submodule is configured to determine the noise data set according to each noise data in the reflow data set.
[0042] In a specific implementation of the second aspect, the clustering submodule may include:
[0043] an IoU determining unit, configured to determine an IoU between the labeled frame and the predicted frame;
[0044] A clustering unit is configured to cluster the labeled box and the predicted box according to an intersection-over-union ratio between the labeled box and the predicted box to obtain clusters.
[0045] In a specific implementation of the second aspect, the clustering unit can be specifically used to: when the intersection-and-union (IoU) between the labeled box and the predicted box is greater than a preset IoU threshold, determine the labeled box and the predicted box as the same cluster; when the IoU between the labeled box and the predicted box is less than or equal to the IoU threshold, determine the labeled box and the predicted box as different clusters.
[0046] In a specific implementation of the second aspect, the noise data set determination submodule may include:
[0047] a noise score determination unit, configured to determine the noise score of each noise data;
[0048] The noise data set determining unit is configured to form the noise data set from the plurality of noise data with the highest noise scores.
[0049] In a specific implementation of the second aspect, the noise score determination unit can be specifically used to: determine the noise score of each noise data according to the confidence of the prediction box; wherein the noise score is positively correlated with the confidence.
[0050] In a specific implementation of the second aspect, the target detection model training device may further include:
[0051] The target detection module is used to obtain an image to be detected; use the second target detection model to perform target detection on the image to be detected, and obtain a target detection result of the image to be detected.
[0052] A third aspect of an embodiment of the present application provides a computer-readable storage medium, which stores a computer program. When the computer program is executed by a processor, it implements the steps of any of the above-mentioned target detection model training methods.
[0053] The fourth aspect of an embodiment of the present application provides an electronic device, comprising a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein the processor implements the steps of any one of the above-mentioned target detection model training methods when executing the computer program.
[0054] A fifth aspect of an embodiment of the present application provides a computer program product, which, when executed on an electronic device, enables the electronic device to execute the steps of any one of the above-mentioned target detection model training methods.
[0055] Compared with the prior art, the beneficial effects of the embodiments of the present application are as follows: the embodiments of the present application obtain a first target detection model; wherein, the first target detection model is a target detection model obtained by training based on an existing labeled data set; obtain a reflow data set, and determine the labeling information of each sample in the reflow data set; use the first target detection model to perform target detection on each sample in the reflow data set, and obtain model prediction information for each sample in the reflow data set; perform noise data screening in the reflow data set according to the labeling information and the model prediction information to obtain a noise data set; perform model training on the first target detection model based on the screened reflow data set to obtain a second target detection model; wherein, the screened reflow data set is a data set obtained after removing the noise data set from the reflow data set. Through the embodiments of the present application, noise data can be screened in the reflow data set according to the labeling information and the model prediction information, and the noise data in the reflow data set can be removed, so that higher quality data features can be learned during model training, thereby improving the accuracy of the trained model in practical applications. BRIEF DESCRIPTION OF THE DRAWINGS
[0056] In order to more clearly illustrate the technical solutions in the embodiments of the present application, the following briefly introduces the drawings required for use in the embodiments or descriptions of the prior art. Obviously, the drawings described below are only some embodiments of the present application. For ordinary technicians in this field, other drawings can be obtained based on these drawings without any creative work.
[0057] Figure 1 This is a flow chart of an embodiment of a target detection model training method in an embodiment of the present application;
[0058] Figure 2 Output diagram of target detection model for input image;
[0059] Figure 3 A schematic flowchart for filtering noisy data in a reflow dataset based on annotation information and model prediction information;
[0060] Figure 4 This is a structural diagram of an embodiment of a target detection model training device in an embodiment of the present application;
[0061] Figure 5 This is a schematic block diagram of an electronic device in an embodiment of the present application. DETAILED DESCRIPTION
[0062] In order to make the purpose, features, and advantages of the invention of this application more obvious and easy to understand, the technical solutions in the embodiments of this application will be clearly and completely described below in conjunction with the drawings in the embodiments of this application. Obviously, the embodiments described below are only part of the embodiments of this application, not all of the embodiments. Based on the embodiments in this application, all other embodiments obtained by ordinary technicians in this field without making creative work are within the scope of protection of this application.
[0063] It will be understood that when used in this specification and the appended claims, the term "comprising" indicates the presence of described features, integers, steps, operations, elements and / or components, but does not preclude the presence or addition of one or more other features, integers, steps, operations, elements, components and / or groups thereof.
[0064] It should also be understood that the terms used in this specification are for the purpose of describing specific embodiments only and are not intended to limit the present application. As used in this specification and the appended claims, the singular forms "a," "an," and "the" are intended to include the plural forms unless the context clearly indicates otherwise.
[0065] It should be further understood that the term "and / or" used in this specification and the appended claims refers to and includes any and all possible combinations of one or more of the associated listed items.
[0066] As used in this specification and the appended claims, the term "if" can be interpreted as "when" or "upon" or "in response to determining" or "in response to detecting," depending on the context. Similarly, the phrase "if it is determined" or "if [described condition or event] is detected" can be interpreted as meaning "upon determination" or "in response to determining" or "upon detection of [described condition or event]" or "in response to detecting [described condition or event]," depending on the context.
[0067] In addition, in the description of the present application, the terms "first", "second", "third", etc. are only used to distinguish the description and cannot be understood as indicating or implying relative importance.
[0068] With the continuous development of artificial intelligence (AI) vision technology, sweepers have become increasingly capable of perceiving the environment and autonomous navigation. In particular, in obstacle detection tasks, AI vision models are widely used to identify and avoid obstacles. Existing sweeper systems can collect visual data of the surrounding environment through image sensors and use target detection algorithms to detect and classify obstacles in real time. However, sweepers often face the emergence of new data in complex actual scenarios. These new data may contain previously unseen obstacle types or environmental changes, resulting in a decrease in the detection capabilities of existing models. In order to maintain the efficiency and accuracy of the model, the data reflow process of the sweeper is very critical, that is, the model is optimized and trained using the newly collected data set (i.e., the reflow data set).
[0069] During the annotation process of reflow datasets, due to manual intervention or the limitations of automatic annotation systems, a certain degree of annotation noise often occurs in the dataset, which can negatively impact the training process. The model may learn incorrect features from incorrect annotations, resulting in low accuracy in practical applications.
[0070] In view of this, the embodiments of the present application provide a target detection model training method, device, computer-readable storage medium and electronic device to solve the problem that the trained model of the existing target detection model training method has low accuracy in practical applications.
[0071] Through the embodiments of the present application, noise data can be screened in the reflow dataset based on the annotation information and model prediction information, and the noise data in the reflow dataset can be eliminated, so that higher quality data features can be learned during model training, thereby improving the accuracy of the trained model in practical applications.
[0072] The execution subject of the embodiment of the present application can be an electronic device, which may include but is not limited to mobile phones, tablet computers, desktop computers, notebooks, PDAs, sweepers, robots, and servers. It should be noted that the electronic device for model training and the electronic device for target detection can be the same electronic device or different electronic devices. That is, the model training can be performed in one electronic device, and then the trained model can be deployed to another electronic device for target detection.
[0073] See also Figure 1 In one embodiment of the present application, a method for training a target detection model may include:
[0074] Step S101: Acquire a first target detection model.
[0075] The first target detection model is a target detection model trained based on an existing labeled dataset. The existing labeled dataset is a dataset collected and labeled in a known stable scene, where the sample labels are of high quality.
[0076] In the embodiment of the present application, a preset artificial intelligence model can be trained based on an existing labeled dataset to obtain a trained first object detection model. The model structure of the artificial intelligence model can be flexibly set according to actual conditions and may include but is not limited to YOLO, EfficientDet, RetinaNet, or other model structures, which are not specifically limited in the embodiment of the present application.
[0077] like Figure 2 As shown, the first target detection model can perform target detection on the input image, and the model prediction information output by it can include the prediction box and the corresponding category and confidence information.
[0078] Step S102: Acquire a reflow dataset and determine the labeling information of each sample in the reflow dataset.
[0079] The robot vacuum collects visual data in a new environment, annotates it, and uses it as a reflow dataset for subsequent screening and training. The samples in the reflow dataset include various complex situations encountered by the robot vacuum in real-world scenarios.
[0080] During the annotation process of reflow datasets, due to manual intervention or the limitations of automatic annotation systems, a certain degree of annotation noise often occurs in the dataset, which can negatively impact the training process. The model may learn incorrect features from incorrect annotations, resulting in low accuracy in practical applications.
[0081] Step S103: Use the first target detection model to perform target detection on each sample in the reflow dataset to obtain model prediction information of each sample in the reflow dataset.
[0082] Step S104: filtering noise data in the return data set according to the annotation information and the model prediction information to obtain a noise data set.
[0083] like Figure 3 As shown, step S104 may specifically include the following process:
[0084] Step S1041: cluster the annotation boxes in the annotation information and the prediction boxes in the model prediction information to obtain clusters.
[0085] Taking any sample in the reflow dataset as an example, we can first determine the intersection over union (IoU) between all the annotations and predictions in the sample. The IoU is a key indicator for measuring the similarity between the annotations and the predictions, as shown in the following formula:
[0086] IoU = (A∩B) / (A∪B)
[0087] Among them, A and B are the areas where the annotation box and the prediction box are located respectively, A∩B is the intersection area of the annotation box and the prediction box, and A∪B is the union area of the annotation box and the prediction box.
[0088] Then, the annotation box and the prediction box can be clustered according to the intersection-and-union ratio between the annotation box and the prediction box to obtain various clusters. Specifically, when the intersection-and-union ratio between the annotation box and the prediction box is greater than the preset intersection-and-union ratio threshold, the annotation box and the prediction box can be determined as the same cluster; when the intersection-and-union ratio between the annotation box and the prediction box is less than or equal to the intersection-and-union ratio threshold, the annotation box and the prediction box can be determined as different clusters. In this way, all annotation boxes and prediction boxes will be clustered into various categories according to their similarities to form several clusters. Among them, the specific value of the intersection-and-union ratio threshold can be flexibly set according to actual conditions, and the embodiments of the present application do not specifically limit this.
[0089] Step S1042: Compare the annotation information and model prediction information in each cluster to obtain a cluster comparison result.
[0090] Step S1043: Determine each noise data in the reflow dataset according to the cluster comparison result.
[0091] Specifically, if a cluster contains a predicted box but no corresponding labeled box, it can be determined to be missing noise data. If the labeled box and the predicted box in a cluster have different categories, it can be determined to be mislabeled noise data. If the labeled box and the predicted box in a cluster have the same category but different positions (box coordinates), it can be determined to be mispositioned noise data.
[0092] Step S1044: Determine a noise data set according to each noise data in the reflow data set.
[0093] In an embodiment of the present application, the noise score of each noise data can be determined separately. For noise data, the higher the confidence (Confidence) of its prediction box, the greater the possibility of labeling error. Therefore, the noise score (Noise_Score) of each noise data can be determined separately according to the confidence of the prediction box. Among them, the noise score is positively correlated with the confidence, that is, the higher the confidence of the prediction box, the higher the noise score, and conversely, the lower the confidence of the prediction box, the lower the noise score. In a specific implementation method of an embodiment of the present application, the confidence of the prediction box can be directly used as the noise score, that is: Noise_Score = Confidence.
[0094] After determining the noise scores of the noise data, the N noise data with the highest noise scores can be combined into a noise data set, where N is a positive integer, and its specific value can be flexibly set according to actual conditions, and is not specifically limited in the present embodiment.
[0095] Step S105: Perform model training on the first target detection model based on the filtered reflow data set to obtain a second target detection model.
[0096] The filtered reflow dataset is a dataset obtained by removing the noise dataset from the reflow dataset.
[0097] In each batch of model training, the training loss value of the batch can be calculated according to the preset loss function, and the model parameters can be adjusted according to the training loss value. Among them, the specific loss function used can be flexibly set according to the actual situation, and the embodiments of this application do not specifically limit this.
[0098] In the embodiment of the present application, it is assumed that in the initial state, the model parameter is W1, and the training loss value is back-propagated to modify the model parameter W1 to obtain the modified model parameter W2. After modifying the parameters, the next training process is continued. In this training process, the training loss value is recalculated and the training loss value is back-propagated to modify the model parameter W2 to obtain the modified model parameter W3, ..., and so on. The above process is repeated continuously. The model parameters can be modified in each training process until the preset training conditions are met. The training condition can be that the number of training times reaches a preset number threshold. The number threshold can be set according to the actual situation. For example, it can be set to thousands, tens of thousands, hundreds of thousands or even larger values. The training condition can also be model convergence. Since the number of training times may not reach the number threshold, but the model has converged, unnecessary work may be repeated. Or the model may not converge, which may lead to an infinite loop and the training process cannot be terminated. Based on the above two situations, the training condition can also be that the number of training times reaches the number threshold or the model converges. When the training conditions are met, the trained second target detection model can be obtained.
[0099] After the second target detection model is trained, the second target detection model can be used to perform actual target detection. Specifically, an image to be detected can be obtained, and the second target detection model can be used to perform target detection on the image to be detected, thereby obtaining a target detection result for the image to be detected.
[0100] In summary, the embodiment of the present application obtains a first target detection model; wherein, the first target detection model is a target detection model obtained by training based on an existing labeled data set; obtains a reflow data set, and determines the labeling information of each sample in the reflow data set; uses the first target detection model to perform target detection on each sample in the reflow data set, and obtains model prediction information for each sample in the reflow data set; performs noise data screening in the reflow data set according to the labeling information and the model prediction information to obtain a noise data set; performs model training on the first target detection model based on the screened reflow data set to obtain a second target detection model; wherein, the screened reflow data set is a data set obtained after removing the noise data set from the reflow data set. Through the embodiment of the present application, noise data can be screened in the reflow data set according to the labeling information and the model prediction information, and the noise data in the reflow data set can be removed, so that higher quality data features can be learned during model training, thereby improving the accuracy of the trained model in practical applications.
[0101] It should be understood that the size of the serial numbers of the steps in the above embodiments does not mean the order of execution. The execution order of each process should be determined by its function and internal logic, and should not constitute any limitation on the implementation process of the embodiments of this application.
[0102] Corresponding to the target detection model training method described in the above embodiment, Figure 4 A structural diagram of an embodiment of a target detection model training device provided in an embodiment of the present application is shown.
[0103] In this embodiment, a target detection model training device may include:
[0104] The model acquisition module 401 is used to acquire a first target detection model; wherein the first target detection model is a target detection model trained based on an existing labeled dataset;
[0105] The labeling information determination module 402 is used to obtain a reflow dataset and determine labeling information of each sample in the reflow dataset;
[0106] a model prediction information determination module 403, configured to perform target detection on each sample in the reflow dataset using the first target detection model to obtain model prediction information for each sample in the reflow dataset;
[0107] A noise data screening module 404 is configured to screen noise data in the return flow dataset according to the annotation information and the model prediction information to obtain a noise dataset;
[0108] The model training module 405 is used to perform model training on the first target detection model based on the filtered reflow dataset to obtain a second target detection model; wherein the filtered reflow dataset is a dataset obtained after removing the noise dataset from the reflow dataset.
[0109] In a specific implementation of the embodiment of the present application, the noise data screening module may include:
[0110] A clustering submodule, configured to cluster the annotation boxes in the annotation information and the prediction boxes in the model prediction information to obtain clusters;
[0111] An information comparison submodule, used to compare the annotation information and the model prediction information in each cluster to obtain a cluster comparison result;
[0112] A noise data determination submodule, configured to determine each noise data in the reflow dataset according to the cluster comparison result;
[0113] The noise data set determining submodule is configured to determine the noise data set according to each noise data in the reflow data set.
[0114] In a specific implementation of the embodiment of the present application, the clustering submodule may include:
[0115] an IoU determining unit, configured to determine an IoU between the labeled frame and the predicted frame;
[0116] A clustering unit is configured to cluster the labeled box and the predicted box according to an intersection-over-union ratio between the labeled box and the predicted box to obtain clusters.
[0117] In a specific implementation of an embodiment of the present application, the clustering unit can be specifically used to: when the intersection-and-union (IoU) between the labeled box and the predicted box is greater than a preset IoU threshold, determine the labeled box and the predicted box as the same cluster; when the IoU between the labeled box and the predicted box is less than or equal to the IoU threshold, determine the labeled box and the predicted box as different clusters.
[0118] In a specific implementation of the embodiment of the present application, the noise data set determination submodule may include:
[0119] a noise score determination unit, configured to determine the noise score of each noise data;
[0120] The noise data set determining unit is configured to form the noise data set from the plurality of noise data with the highest noise scores.
[0121] In a specific implementation of an embodiment of the present application, the noise score determination unit can be specifically used to: determine the noise score of each noise data according to the confidence of the prediction frame; wherein the noise score is positively correlated with the confidence.
[0122] In a specific implementation of the embodiment of the present application, the target detection model training device may further include:
[0123] The target detection module is used to obtain an image to be detected; use the second target detection model to perform target detection on the image to be detected, and obtain a target detection result of the image to be detected.
[0124] Those skilled in the art will clearly understand that, for the convenience and brevity of description, the specific working processes of the above-described devices, modules and units can refer to the corresponding processes in the aforementioned method embodiments and will not be repeated here.
[0125] In the above embodiments, the description of each embodiment has its own focus. For parts that are not described or recorded in detail in a certain embodiment, reference can be made to the relevant description of other embodiments.
[0126] Figure 5 A schematic block diagram of an electronic device provided in an embodiment of the present application is shown. For ease of explanation, only the parts related to the embodiment of the present application are shown.
[0127] like Figure 5 As shown, the electronic device 5 of this embodiment includes: a processor 50, a memory 51, and a computer program 52 stored in the memory 51 and executable on the processor 50. When the processor 50 executes the computer program 52, the steps in the above-mentioned target detection model training method embodiments are implemented, such as Figure 1 Alternatively, when the processor 50 executes the computer program 52, the functions of the modules / units in the above-mentioned device embodiments are realized, for example, Figure 4 Functions of modules 401 to 405 are shown.
[0128] Exemplarily, the computer program 52 may be divided into one or more modules / units, which are stored in the memory 51 and executed by the processor 50 to implement the present application. The one or more modules / units may be a series of computer program instruction segments capable of performing specific functions, and the instruction segments are used to describe the execution process of the computer program 52 in the electronic device 5.
[0129] The electronic device 5 may include but is not limited to mobile phones, tablet computers, desktop computers, notebooks, PDAs, sweepers, robots, servers and other computing devices. It will be understood by those skilled in the art that Figure 5 It is only an example of the electronic device 5 and does not constitute a limitation of the electronic device 5. It may include more or fewer components than shown in the figure, or a combination of certain components, or different components. For example, the electronic device 5 may also include input and output devices, network access devices, buses, etc.
[0130] The processor 50 may be a central processing unit (CPU), or other general-purpose processors, digital signal processors (DSP), application-specific integrated circuits (ASIC), field-programmable gate arrays (FPGA), or other programmable logic devices, discrete gate or transistor logic devices, discrete hardware components, etc. The general-purpose processor may be a microprocessor or any conventional processor.
[0131] The memory 51 can be an internal storage unit of the electronic device 5, such as a hard disk or memory of the electronic device 5. The memory 51 can also be an external storage device of the electronic device 5, such as a plug-in hard disk, a smart media card (SMC), a secure digital (SD) card, a flash card, etc. equipped on the electronic device 5. Furthermore, the memory 51 can also include both an internal storage unit of the electronic device 5 and an external storage device. The memory 51 is used to store the computer program and other programs and data required by the electronic device 5. The memory 51 can also be used to temporarily store data that has been output or is about to be output.
[0132] Those skilled in the art can clearly understand that, for the convenience and brevity of description, only the division of the above-mentioned functional units and modules is used as an example for illustration. In actual applications, the above-mentioned functions can be distributed and completed by different functional units and modules as needed, that is, the internal structure of the device can be divided into different functional units or modules to complete all or part of the functions described above. The functional units and modules in the embodiment can be integrated into one processing unit, or each unit can exist physically alone, or two or more units can be integrated into one unit. The above-mentioned integrated unit can be implemented in the form of hardware or in the form of software functional units. In addition, the specific names of the functional units and modules are only for the convenience of distinguishing each other, and are not used to limit the scope of protection of this application. The specific working process of the units and modules in the above-mentioned system can refer to the corresponding process in the aforementioned method embodiment, and will not be repeated here.
[0133] In the above embodiments, the description of each embodiment has its own focus. For parts that are not described or recorded in detail in a certain embodiment, reference can be made to the relevant description of other embodiments.
[0134] Those skilled in the art will appreciate that the units and algorithm steps of each example described in conjunction with the embodiments disclosed herein can be implemented in electronic hardware, or a combination of computer software and electronic hardware. Whether these functions are performed in hardware or software depends on the specific application and design constraints of the technical solution. Professional and technical personnel can use different methods to implement the described functions for each specific application, but such implementation should not be considered beyond the scope of this application.
[0135] In the embodiments provided in the present application, it should be understood that the disclosed devices / electronic devices and methods can be implemented in other ways. For example, the device / electronic device embodiments described above are merely schematic. For example, the division of the modules or units is merely a logical function division. In actual implementation, there may be other division methods, such as multiple units or components can be combined or integrated into another system, or some features can be ignored or not executed. Another point is that the mutual coupling or direct coupling or communication connection shown or discussed can be through some interfaces, indirect coupling or communication connection of devices or units, which can be electrical, mechanical or other forms.
[0136] The units described as separate components may or may not be physically separate, and the components shown as units may or may not be physical units, that is, they may be located in one place or distributed across multiple network units. Some or all of these units may be selected to achieve the purpose of this embodiment according to actual needs.
[0137] In addition, the functional units in the various embodiments of the present application may be integrated into a single processing unit, or each unit may exist physically separately, or two or more units may be integrated into a single unit. The aforementioned integrated units may be implemented in the form of hardware or software functional units.
[0138] If the integrated module / unit is implemented in the form of a software functional unit and sold or used as an independent product, it can be stored in a computer-readable storage medium. Based on this understanding, the present application implements all or part of the process in the above-mentioned embodiment method, and can also be completed by instructing the relevant hardware through a computer program. The computer program can be stored in a computer-readable storage medium, and when the computer program is executed by the processor, it can implement the steps of the above-mentioned various method embodiments. Among them, the computer program includes computer program code, and the computer program code can be in source code form, object code form, executable file or some intermediate form. The computer-readable storage medium may include: any entity or device capable of carrying the computer program code, recording medium, USB flash drive, mobile hard disk, magnetic disk, optical disk, computer memory, read-only memory (ROM), random access memory (RAM), electric carrier signal, telecommunication signal and software distribution medium. It should be noted that the content contained in the computer-readable storage medium can be appropriately increased or decreased according to the requirements of legislation and patent practice in the jurisdiction. For example, in some jurisdictions, according to legislation and patent practice, computer-readable storage media do not include electric carrier signals and telecommunication signals.
[0139] The above-described embodiments are only used to illustrate the technical solutions of the present application, rather than to limit them. Although the present application has been described in detail with reference to the aforementioned embodiments, those skilled in the art should understand that they can still modify the technical solutions described in the aforementioned embodiments, or make equivalent replacements for some of the technical features therein. These modifications or replacements do not deviate the essence of the corresponding technical solutions from the spirit and scope of the technical solutions of the various embodiments of the present application, and should all be included in the scope of protection of the present application.
Claims
1. A target detection model training method, characterized in that: include: Obtain a first target detection model; wherein the first target detection model is a target detection model trained based on an existing labeled dataset; Obtaining a reflow dataset and determining labeling information for each sample in the reflow dataset; Performing target detection on each sample in the reflow dataset using the first target detection model to obtain model prediction information for each sample in the reflow dataset; Screening noise data in the reflux dataset according to the annotation information and the model prediction information to obtain a noise dataset; The first target detection model is trained based on the filtered reflow dataset to obtain a second target detection model; wherein the filtered reflow dataset is a dataset obtained after removing the noise dataset from the reflow dataset.
2. The target detection model training method according to claim 1, characterized in that The filtering of noise data in the reflow dataset according to the annotation information and the model prediction information to obtain a noise dataset includes: Clustering the annotation boxes in the annotation information and the prediction boxes in the model prediction information to obtain clusters; Comparing the annotation information and the model prediction information in each cluster to obtain a cluster comparison result; Determining each noise data in the reflow dataset according to the cluster comparison result; The noise data set is determined according to each noise data in the reflow data set.
3. The target detection model training method according to claim 2, characterized in that The clustering of the annotation boxes in the annotation information and the prediction boxes in the model prediction information to obtain clusters includes: Determining an intersection-over-union ratio between the labeled box and the predicted box; The labeled box and the predicted box are clustered according to an intersection-over-union ratio between the labeled box and the predicted box to obtain clusters.
4. The target detection model training method according to claim 3, characterized in that The clustering of the labeled box and the predicted box according to the intersection-over-union ratio between the labeled box and the predicted box to obtain clusters includes: When an intersection-over-union (IoU) ratio between the marked frame and the predicted frame is greater than a preset IoU threshold, determining the marked frame and the predicted frame as belonging to the same cluster; When an intersection-over-union (IoU) ratio between the labeled box and the predicted box is less than or equal to the IoU threshold, the labeled box and the predicted box are determined to be in different clusters.
5. The target detection model training method according to claim 2, characterized in that The determining the noise data set according to each noise data in the reflow data set includes: Determine the noise score of each noise data separately; The noise data set is composed of several noise data with the highest noise scores.
6. The target detection model training method according to claim 5, characterized in that The step of respectively determining the noise score of each noise data comprises: A noise score of each noise data is determined according to the confidence of the prediction box; wherein the noise score is positively correlated with the confidence.
7. The target detection model training method according to any one of claims 1 to 6, characterized in that: After training the first object detection model based on the filtered reflow dataset to obtain a second object detection model, the method further includes: Obtain the image to be detected; Use the second target detection model to perform target detection on the image to be detected to obtain a target detection result of the image to be detected.
8. A target detection model training device, characterized in that: include: A model acquisition module is used to acquire a first target detection model; wherein the first target detection model is a target detection model trained based on an existing labeled data set; a labeling information determination module, configured to obtain a reflow dataset and determine labeling information for each sample in the reflow dataset; a model prediction information determination module, configured to perform target detection on each sample in the reflow dataset using the first target detection model to obtain model prediction information for each sample in the reflow dataset; A noise data screening module, configured to screen noise data in the return data set according to the annotation information and the model prediction information to obtain a noise data set; A model training module is used to perform model training on the first target detection model based on the filtered reflow data set to obtain a second target detection model; wherein the filtered reflow data set is a data set obtained after removing the noise data set from the reflow data set.
9. A computer-readable storage medium storing a computer program, characterized in that: When the computer program is executed by a processor, the steps of the target detection model training method according to any one of claims 1 to 7 are implemented.
10. An electronic device comprising a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein: When the processor executes the computer program, the steps of the target detection model training method according to any one of claims 1 to 7 are implemented.