CN116521917A - Picture screening method and device - Google Patents
Picture screening method and device Download PDFInfo
- Publication number
- CN116521917A CN116521917A CN202310464508.6A CN202310464508A CN116521917A CN 116521917 A CN116521917 A CN 116521917A CN 202310464508 A CN202310464508 A CN 202310464508A CN 116521917 A CN116521917 A CN 116521917A
- Authority
- CN
- China
- Prior art keywords
- picture
- face
- pictures
- similarity
- screening
- 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.)
- Pending
Links
- 238000012216 screening Methods 0.000 title claims abstract description 79
- 238000000034 method Methods 0.000 title claims abstract description 35
- 230000014509 gene expression Effects 0.000 claims abstract description 63
- 238000001514 detection method Methods 0.000 claims abstract description 58
- 238000012545 processing Methods 0.000 claims abstract description 13
- 238000012549 training Methods 0.000 claims description 21
- 238000013136 deep learning model Methods 0.000 claims description 11
- 230000000007 visual effect Effects 0.000 claims description 6
- 238000010606 normalization Methods 0.000 claims description 4
- 238000001914 filtration Methods 0.000 claims description 2
- 210000000887 face Anatomy 0.000 description 14
- 238000011156 evaluation Methods 0.000 description 5
- 238000013527 convolutional neural network Methods 0.000 description 4
- 238000012360 testing method Methods 0.000 description 4
- 238000013135 deep learning Methods 0.000 description 2
- 238000005457 optimization Methods 0.000 description 2
- 230000009286 beneficial effect Effects 0.000 description 1
- 230000036461 convulsion Effects 0.000 description 1
- 238000010586 diagram Methods 0.000 description 1
- 238000005516 engineering process Methods 0.000 description 1
- 238000000605 extraction Methods 0.000 description 1
- 210000004709 eyebrow Anatomy 0.000 description 1
- 230000008921 facial expression Effects 0.000 description 1
- 230000006870 function Effects 0.000 description 1
- 238000007781 pre-processing Methods 0.000 description 1
Classifications
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F16/00—Information retrieval; Database structures therefor; File system structures therefor
- G06F16/50—Information retrieval; Database structures therefor; File system structures therefor of still image data
- G06F16/53—Querying
- G06F16/535—Filtering based on additional data, e.g. user or group profiles
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F16/00—Information retrieval; Database structures therefor; File system structures therefor
- G06F16/50—Information retrieval; Database structures therefor; File system structures therefor of still image data
- G06F16/55—Clustering; Classification
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F16/00—Information retrieval; Database structures therefor; File system structures therefor
- G06F16/50—Information retrieval; Database structures therefor; File system structures therefor of still image data
- G06F16/58—Retrieval characterised by using metadata, e.g. metadata not derived from the content or metadata generated manually
- G06F16/583—Retrieval characterised by using metadata, e.g. metadata not derived from the content or metadata generated manually using metadata automatically derived from the content
-
- 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
- G06V10/74—Image or video pattern matching; Proximity measures in feature spaces
- G06V10/761—Proximity, similarity or dissimilarity measures
-
- 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
- G06V10/82—Arrangements for image or video recognition or understanding using pattern recognition or machine learning using neural networks
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06V—IMAGE OR VIDEO RECOGNITION OR UNDERSTANDING
- G06V40/00—Recognition of biometric, human-related or animal-related patterns in image or video data
- G06V40/10—Human or animal bodies, e.g. vehicle occupants or pedestrians; Body parts, e.g. hands
- G06V40/16—Human faces, e.g. facial parts, sketches or expressions
- G06V40/172—Classification, e.g. identification
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06V—IMAGE OR VIDEO RECOGNITION OR UNDERSTANDING
- G06V40/00—Recognition of biometric, human-related or animal-related patterns in image or video data
- G06V40/10—Human or animal bodies, e.g. vehicle occupants or pedestrians; Body parts, e.g. hands
- G06V40/16—Human faces, e.g. facial parts, sketches or expressions
- G06V40/174—Facial expression recognition
Landscapes
- Engineering & Computer Science (AREA)
- Theoretical Computer Science (AREA)
- Physics & Mathematics (AREA)
- General Physics & Mathematics (AREA)
- Databases & Information Systems (AREA)
- Multimedia (AREA)
- Health & Medical Sciences (AREA)
- General Health & Medical Sciences (AREA)
- Data Mining & Analysis (AREA)
- General Engineering & Computer Science (AREA)
- Evolutionary Computation (AREA)
- Computer Vision & Pattern Recognition (AREA)
- Oral & Maxillofacial Surgery (AREA)
- Human Computer Interaction (AREA)
- Artificial Intelligence (AREA)
- Computing Systems (AREA)
- Library & Information Science (AREA)
- Medical Informatics (AREA)
- Software Systems (AREA)
- Image Analysis (AREA)
Abstract
The invention relates to the technical field of image processing, discloses a picture screening method and a picture screening device, and aims to solve the problems of low efficiency, low accuracy and poor picture quality of the existing picture screening mode, wherein the scheme mainly comprises the following steps: determining the picture similarity among the pictures in the picture library, and dividing the pictures into a plurality of similar picture groups according to the picture similarity; determining the picture quality of each picture, and determining a representative picture corresponding to each similar picture group according to the picture quality; carrying out face recognition on each representative picture, and screening face pictures containing preset faces from the representative pictures according to a face recognition result; performing motion blur detection on each face picture, and removing face pictures with motion blur confidence scores greater than a second threshold; and carrying out expression detection on the rest face pictures, and screening the face pictures with the expressions meeting preset conditions from the rest face pictures. The invention improves the efficiency, accuracy and quality of picture screening, and is suitable for batch picture screening.
Description
Technical Field
The invention relates to the technical field of image processing, in particular to a picture screening method and device.
Background
In order to screen out face pictures meeting requirements from a picture library, the following two modes are generally adopted in the prior art: the first is manual screening from the picture library, which consumes very much manpower resources and time cost and has lower efficiency. The second method is to analyze and process the images by using technologies such as computer vision and deep learning, and automatically display similar pictures or face pictures in a picture library, so that the mode of analyzing and processing the images is single, the screening requirement of the face pictures with diversity cannot be met, manual further screening is still needed when the screening requirement is high, human resources and time cost are very consumed, and the screening accuracy and the quality of the pictures are poor.
Disclosure of Invention
The invention aims to solve the problems that the existing picture screening mode cannot meet the requirement of screening various face pictures and has low efficiency, low accuracy and poor picture quality, and provides a picture screening method and device.
The technical scheme adopted by the invention for solving the technical problems is as follows:
in one aspect, a method for screening pictures is provided, the method comprising:
determining the picture similarity among the pictures in a picture library, dividing the pictures into a plurality of similar picture groups according to the picture similarity, wherein the picture similarity of each picture in each similar picture group and at least one other picture in the same similar picture group is larger than a first threshold value;
determining the picture quality of each picture, and determining a representative picture corresponding to each similar picture group according to the picture quality;
carrying out face recognition on each representative picture, and screening face pictures containing preset faces from the representative pictures according to a face recognition result;
performing motion blur detection on each face picture, and removing face pictures with motion blur confidence scores greater than a second threshold;
and carrying out expression detection on the rest face pictures, and screening the face pictures with the expressions meeting preset conditions from the rest face pictures.
Further, the determining the similarity of the pictures in the picture library specifically includes:
and respectively determining the structural similarity and the color similarity of each picture and other pictures in the picture library based on a picture similarity detection algorithm, determining the picture similarity according to the structural similarity and the color similarity, and adding the pictures with the picture similarity larger than a first threshold value into the same similar picture group.
Further, the similarity detection algorithm is a structural similarity algorithm or a mean square error algorithm.
Further, the picture similarity is obtained according to a weighted sum of the structural similarity and the color similarity.
Further, the determining the picture quality of each picture specifically includes:
and determining the picture quality of each picture according to one or more indexes of peak signal-to-noise ratio, structural similarity index, visual information fidelity and image naturalness.
Further, the performing face recognition on each representative picture specifically includes:
and creating a face recognition model for detecting whether the preset face is included or not, and inputting each representative picture into the face recognition model to obtain a face recognition result of the representative picture.
Further, the creating a face recognition model for detecting whether the preset face is included specifically includes:
acquiring a face picture set, determining face information of each picture in the face picture set, and performing image enhancement, normalization, cutting and rotation processing on each picture in the face picture set;
and training the processed face picture set and face information corresponding to each picture to obtain a deep learning model for identifying the face information, and creating a face recognition model according to the deep learning model.
Further, the motion blur detection for each face picture specifically includes:
acquiring a motion blur picture set, and determining a motion blur confidence score of each picture in the motion blur picture set;
and taking the motion blur picture set and the motion blur confidence scores corresponding to the pictures as training samples, training a motion blur detection model for motion blur detection, and inputting the face pictures into the motion blur detection model to obtain a motion blur detection result of the face pictures.
Further, the performing expression detection on the remaining face pictures specifically includes:
acquiring an expression picture set, and determining the expression type of each picture in the expression picture set;
and taking the expression picture set and the expression type corresponding to each picture as training samples, training an expression recognition model for expression detection, and inputting the rest face pictures into the expression recognition model to obtain the expression detection result of the face pictures.
In another aspect, a picture screening apparatus is provided, the apparatus comprising:
the similarity detection unit is used for determining the picture similarity among the pictures in the picture library, dividing the pictures into a plurality of similar picture groups according to the picture similarity, and enabling the picture similarity of each picture in each similar picture group to be at least greater than a first threshold value with the picture similarity of another picture in the same similar picture group;
the quality detection unit is used for determining the picture quality of each picture and determining the representative picture corresponding to each similar picture group according to the picture quality;
the face recognition unit is used for recognizing the face of each representative picture, and screening face pictures containing preset faces from the representative pictures according to the face recognition result;
the motion blur detection unit is used for performing motion blur detection on each face picture and removing face pictures with the motion blur confidence scores larger than a second threshold value;
the expression detection unit is used for carrying out expression detection on the remaining face pictures, and screening the face pictures with the expressions meeting preset conditions from the remaining face pictures.
The beneficial effects of the invention are as follows: according to the picture screening method and device, the face pictures meeting the requirements are automatically screened and filtered by adopting various image processing and deep learning models, automatic accurate screening and optimization processing of a large number of pictures are realized, the tedious and time-consuming process of manual screening is avoided, the screening efficiency and quality are improved, meanwhile, customization processing can be carried out according to different picture screening requirements, the flexibility and adaptability are high, and the screening requirements of the face pictures with diversity can be met.
Drawings
Fig. 1 is a schematic flow chart of a picture screening method according to an embodiment of the present invention;
FIG. 2 is a schematic flow chart of another method for screening pictures according to an embodiment of the present invention;
fig. 3 is a schematic structural diagram of a picture screening apparatus according to an embodiment of the present invention.
Detailed Description
Embodiments of the present invention will be described in detail below with reference to the accompanying drawings.
The invention aims to improve the efficiency, accuracy and quality of picture screening, and provides a picture screening method and device, wherein the main technical scheme comprises the following steps: determining the picture similarity among the pictures in a picture library, dividing the pictures into a plurality of similar picture groups according to the picture similarity, wherein the picture similarity of each picture in each similar picture group and at least one other picture in the same similar picture group is larger than a first threshold value; determining the picture quality of each picture, and determining a representative picture corresponding to each similar picture group according to the picture quality; carrying out face recognition on each representative picture, and screening face pictures containing preset faces from the representative pictures according to a face recognition result; performing motion blur detection on each face picture, and removing face pictures with motion blur confidence scores greater than a second threshold; and carrying out expression detection on the rest face pictures, and screening the face pictures with the expressions meeting preset conditions from the rest face pictures.
When the method is actually applied, the screening requirement of the face pictures is determined in advance by a user, the screening requirement can comprise faces and expressions to be screened, and after the screening requirement is determined, the high-quality face pictures meeting the screening requirement can be automatically screened in a picture library. The method comprises the steps of firstly dividing all pictures in a picture library into a plurality of similar picture groups according to picture similarity, judging the pictures in each similar picture group as similar pictures, then determining a representative picture with highest quality in each similar picture group, respectively carrying out face recognition on all the representative pictures, screening out face pictures meeting screening requirements, then eliminating the face pictures which are blurred due to movement aiming at the screened face pictures, further improving the quality of the face pictures, and finally carrying out expression detection on the rest face pictures, thereby screening out the face pictures meeting the screening requirements.
Examples
Referring to fig. 1, the picture screening method according to the embodiment of the invention includes the following steps:
step 1, determining the picture similarity among pictures in a picture library, dividing the pictures into a plurality of similar picture groups according to the picture similarity, wherein the picture similarity of each picture in each similar picture group and another picture in the same similar picture group is at least greater than a first threshold;
referring to fig. 2, in practical application, the present embodiment needs to determine the screening requirement of the face picture in advance by the user, where the screening requirement may include faces and expressions to be screened, and after determining the screening requirement, the face picture can be automatically screened in the picture library to obtain a high-quality face picture meeting the screening requirement.
The pictures in the picture library are obtained in advance, for example, a film is downloaded to the local, a lot of pictures are obtained by locally taking frames (screenshot) of the video, and then the pictures are added into the picture library.
Specifically, the embodiment respectively determines the structural similarity and the color similarity of each picture and each other picture in the picture library based on the picture similarity detection algorithm, takes the weighted sum of the structural similarity and the color similarity as the picture similarity, adds the picture corresponding to the picture similarity larger than the first threshold value into the same similar picture group, and can divide the picture into a plurality of similar picture groups, and the picture in each similar picture group is judged to be a similar picture.
In this embodiment, an open source library, such as OpenCV, may be used to perform image processing and calculate the similarity between the pictures, or a structural similarity algorithm (SSIM) or a mean square error algorithm (MSE) may be used to calculate the similarity between the two pictures, and these algorithms may measure the structural similarity and the color similarity between the two pictures, and then obtain the final picture similarity through the weighted sum of the structural similarity and the color similarity. The higher picture similarity indicates that the contents of two pictures are very similar, possibly different versions of the same picture, while the lower similarity indicates that the contents of two pictures are very different, and when the similarity is greater than a preset first threshold, the two pictures are judged to be similar pictures and are put into the same similar picture group. By the above way, after all the pictures in the picture library are traversed twice, a plurality of similar picture groups can be obtained.
The first threshold may be set according to practical situations, for example, the first threshold is 0.8, that is, when the similarity of the two pictures is greater than 0.8, it indicates that the two pictures are similar, and the two pictures are temporarily placed in a similar picture group.
Step 2, determining the picture quality of each picture, and determining a representative picture corresponding to each similar picture group according to the picture quality;
in this embodiment, the picture quality of each picture may be determined according to the peak signal-to-noise ratio, the structural similarity index, the visual information fidelity, the image naturalness, and other indexes, and for each similar picture group, a picture with the best quality is selected from the similar picture groups to be used as a representative picture of the similar picture group.
Wherein, PSNR (Peak Signal-to-Noise Ratio): peak signal-to-noise ratio is a widely used method of assessing image quality by comparing the signal-to-noise ratio between an original image and a compressed image.
SSIM (Structural Similarity Index): the structural similarity index is an evaluation method based on image structural information, and the quality of an image is evaluated by comparing the structural similarity between an original image and a compressed image.
VIF (Visual Information Fidelity): visual information fidelity is an evaluation method based on perceived quality, and the quality of an image is evaluated by comparing the visual information fidelity between an original image and a compressed image.
NIQE (Natural Image Quality Evaluator): the natural image quality estimator is an estimating method based on the naturalness of an image, which estimates the quality of the image by comparing the naturalness between an original image and a compressed image.
In addition, there are other objective evaluation methods, such as FSIM (Feature SIMilarity ), IFC (Image Fidelity Criterion, image assurance criterion), laplacian gradient method, and the like, each of which has suitable scenes and features, and may be evaluated by selecting one or more suitable methods according to actual requirements in practical application, which is not limited in this embodiment.
Step 3, carrying out face recognition on each representative picture, and screening face pictures containing preset faces from the representative pictures according to a face recognition result;
the embodiment can create a face recognition model for detecting whether the preset face is included or not, and input each representative picture into the face recognition model to obtain the face recognition result of the representative picture.
Wherein, the step of creating the face recognition model may include:
acquiring a face picture set, determining face information of each picture in the face picture set, and performing image enhancement, normalization, cutting and rotation processing on each picture in the face picture set; and training the processed face picture set and face information corresponding to each picture to obtain a deep learning model for identifying the face information, and creating a face recognition model according to the deep learning model.
In practical application, a group of face picture sets with labels can be prepared, and the pictures in the face picture sets can be acquired from the internet or manually acquired. For each picture in the face picture set, face information is marked, the face information comprises the position of the face and the name of the face, and the marking can be manually completed or an automatic tool can be used. And then preprocessing each picture in the face picture set, including operations such as image enhancement, normalization, cutting, rotation and the like. Then dividing the face picture set and the face information corresponding to the face picture set into a training set and a testing set, training the training set to obtain a deep learning model for identifying the face information, verifying the accuracy of the deep learning model through the testing set, and when the accuracy meets the conditions, using the deep learning model to penetrate the face recognition model, wherein the face recognition model can judge whether the input picture contains a preset face or not.
For example, assume that a face recognition model has been trained in advance and data information of a plurality of faces is entered into a database, the face recognition model can recognize the number of qualified faces (faces), and can compare with the data in the database to obtain the most similar face similarity (between 0 and 1). If the filtering requirement is to filter out all the pictures containing the "first", the pictures filtered in the previous step are input into the face recognition model one by one, and the face recognition model returns whether the pictures contain the "first" or not (the probability of returning to be the "first" if the pictures contain the "first") for each picture. If the similarity exceeds the threshold value and the number of faces is less than or equal to 3 (of course, the number of faces is not limited), the face picture is reserved.
Step 4, performing motion blur detection on each face picture, and removing face pictures with the motion blur confidence scores larger than a second threshold;
according to the embodiment, a motion blur detection model can be created, each face picture is input into the motion blur detection model, a motion blur detection result of the face picture can be obtained, further, the faces and the human body in the pictures are judged to be blurred due to motion, the face pictures are removed from the list, and therefore the quality of picture screening is further improved.
Wherein the step of creating a motion blur detection model may comprise:
acquiring a motion blur picture set, and determining a motion blur confidence score of each picture in the motion blur picture set; taking the motion blur picture set and the motion blur confidence scores corresponding to the pictures as training samples to train a motion blur detection model for motion blur detection
In this embodiment, a deep learning framework such as TensorFlow or PyTorch may be used to train the model, and training data may be derived from the faces and motion blurred images of the human body that have been labeled in the motion blurred image set; or by a Convolutional Neural Network (CNN) model, such as the "blurdection" model implemented based on Tensorflow, which can detect motion blur on an input image and return a motion blur confidence score.
The second threshold may be set according to practical situations, which is not limited in this embodiment.
And 5, carrying out expression detection on the rest face pictures, and screening the face pictures with the expressions meeting preset conditions from the rest face pictures.
According to the embodiment, the expression recognition model can be created, the rest face pictures are input into the expression recognition model, the expression detection result of the face pictures is obtained, further, the face pictures with the expressions meeting the screening requirements are screened out, meanwhile, the face pictures with exaggerated expressions can be filtered, and the picture screening quality is further improved.
In this embodiment, the expression recognition model may be implemented by using a HaarCascade classifier model provided in OpenCV, which may detect a face and recognize expression features, or may train a Convolutional Neural Network (CNN) to implement this function, and specifically includes the following steps:
acquiring an expression picture set, and determining the expression type of each picture in the expression picture set; and taking the expression picture set and the expression type corresponding to each picture as training samples, and training an expression recognition model for expression detection.
Specifically, a public expression picture set, such as FER2013, may be searched from the internet first. The picture sets need to contain facial expression images of different people, labels corresponding to each picture and label the types of expressions. Features then need to be extracted from each of the emoticons for subsequent classification or recognition, and some common feature extraction methods, such as Local Binary Pattern (LBP), direction gradient Histogram (HOG), face keypoints, etc., may be used. And then dividing the marked expression picture set into a training sample and a test sample, training the model by using the training sample, evaluating the model by using the test sample, and calculating evaluation indexes such as accuracy, recall rate, F1 value and the like of the model. After the evaluation index meets the requirement, the trained expression recognition model can be applied to an actual scene, expressions in the pictures can be classified through the model, the pictures meeting the requirement can be screened out, meanwhile, the pictures with exaggerated expressions are filtered out, and face pictures containing the exaggerated expressions, such as anger, twitch, squeezing eyebrows and making eyes, can be removed.
In summary, the method and the device for screening pictures provided in this embodiment automatically screen and filter out face pictures meeting requirements by adopting various image processing and deep learning models, so as to realize automatic accurate screening and optimization processing of a large number of pictures, avoid tedious and time-consuming manual screening, improve screening efficiency and quality, and simultaneously, can perform customized processing according to different picture screening requirements, have higher flexibility and adaptability, and can meet the face picture screening requirements of diversity.
Referring to fig. 3, based on the above technical solution, this embodiment further provides a device for screening pictures, where the device includes:
the similarity detection unit is used for determining the picture similarity among the pictures in the picture library, dividing the pictures into a plurality of similar picture groups according to the picture similarity, and enabling the picture similarity of each picture in each similar picture group to be at least greater than a first threshold value with the picture similarity of another picture in the same similar picture group;
the quality detection unit is used for determining the picture quality of each picture and determining the representative picture corresponding to each similar picture group according to the picture quality;
the face recognition unit is used for recognizing the face of each representative picture, and screening face pictures containing preset faces from the representative pictures according to the face recognition result;
the motion blur detection unit is used for performing motion blur detection on each face picture and removing face pictures with the motion blur confidence scores larger than a second threshold value;
the expression detection unit is used for carrying out expression detection on the remaining face pictures, and screening the face pictures with the expressions meeting preset conditions from the remaining face pictures.
It can be understood that, since the picture screening apparatus according to the embodiments of the present invention is an apparatus for implementing the picture screening method according to the embodiments, for the apparatus disclosed in the embodiments, since the apparatus corresponds to the method disclosed in the embodiments, the description is simpler, and the relevant points refer to the part of the description of the method.
Claims (10)
1. A picture screening method, the method comprising:
determining the picture similarity among the pictures in a picture library, dividing the pictures into a plurality of similar picture groups according to the picture similarity, wherein the picture similarity of each picture in each similar picture group and at least one other picture in the same similar picture group is larger than a first threshold value;
determining the picture quality of each picture, and determining a representative picture corresponding to each similar picture group according to the picture quality;
carrying out face recognition on each representative picture, and screening face pictures containing preset faces from the representative pictures according to a face recognition result;
performing motion blur detection on each face picture, and removing face pictures with motion blur confidence scores greater than a second threshold;
and carrying out expression detection on the rest face pictures, and screening the face pictures with the expressions meeting preset conditions from the rest face pictures.
2. The method for screening pictures according to claim 1, wherein the determining the similarity of pictures in the picture library specifically comprises:
and respectively determining the structural similarity and the color similarity of each picture and other pictures in the picture library based on a picture similarity detection algorithm, determining the picture similarity according to the structural similarity and the color similarity, and adding the pictures with the picture similarity larger than a first threshold value into the same similar picture group.
3. The picture screening method of claim 2, wherein the similarity detection algorithm is a structural similarity algorithm or a mean square error algorithm.
4. The picture screening method of claim 2, wherein the picture similarity is derived from a weighted sum of structural similarity and color similarity.
5. The picture screening method as claimed in claim 1, wherein the determining the picture quality of each picture specifically includes:
and determining the picture quality of each picture according to one or more indexes of peak signal-to-noise ratio, structural similarity index, visual information fidelity and image naturalness.
6. The picture screening method as claimed in claim 1, wherein the performing face recognition on each representative picture specifically includes:
and creating a face recognition model for detecting whether the preset face is included or not, and inputting each representative picture into the face recognition model to obtain a face recognition result of the representative picture.
7. The picture filtering method according to claim 6, wherein the creating a face recognition model for detecting whether the preset face is included comprises:
acquiring a face picture set, determining face information of each picture in the face picture set, and performing image enhancement, normalization, cutting and rotation processing on each picture in the face picture set;
and training the processed face picture set and face information corresponding to each picture to obtain a deep learning model for identifying the face information, and creating a face recognition model according to the deep learning model.
8. The picture screening method as claimed in claim 1, wherein the performing motion blur detection on each face picture specifically includes:
acquiring a motion blur picture set, and determining a motion blur confidence score of each picture in the motion blur picture set;
and taking the motion blur picture set and the motion blur confidence scores corresponding to the pictures as training samples, training a motion blur detection model for motion blur detection, and inputting the face pictures into the motion blur detection model to obtain a motion blur detection result of the face pictures.
9. The picture screening method as claimed in claim 1, wherein the performing expression detection on the remaining face picture specifically includes:
acquiring an expression picture set, and determining the expression type of each picture in the expression picture set;
and taking the expression picture set and the expression type corresponding to each picture as training samples, training an expression recognition model for expression detection, and inputting the rest face pictures into the expression recognition model to obtain the expression detection result of the face pictures.
10. A picture screening apparatus, the apparatus comprising:
the similarity detection unit is used for determining the picture similarity among the pictures in the picture library, dividing the pictures into a plurality of similar picture groups according to the picture similarity, and enabling the picture similarity of each picture in each similar picture group to be at least greater than a first threshold value with the picture similarity of another picture in the same similar picture group;
the quality detection unit is used for determining the picture quality of each picture and determining the representative picture corresponding to each similar picture group according to the picture quality;
the face recognition unit is used for recognizing the face of each representative picture, and screening face pictures containing preset faces from the representative pictures according to the face recognition result;
the motion blur detection unit is used for performing motion blur detection on each face picture and removing face pictures with the motion blur confidence scores larger than a second threshold value;
the expression detection unit is used for carrying out expression detection on the remaining face pictures, and screening the face pictures with the expressions meeting preset conditions from the remaining face pictures.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202310464508.6A CN116521917A (en) | 2023-04-26 | 2023-04-26 | Picture screening method and device |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202310464508.6A CN116521917A (en) | 2023-04-26 | 2023-04-26 | Picture screening method and device |
Publications (1)
| Publication Number | Publication Date |
|---|---|
| CN116521917A true CN116521917A (en) | 2023-08-01 |
Family
ID=87404143
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN202310464508.6A Pending CN116521917A (en) | 2023-04-26 | 2023-04-26 | Picture screening method and device |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN116521917A (en) |
Cited By (1)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN117453936A (en) * | 2023-10-19 | 2024-01-26 | 山东三木众合信息科技股份有限公司 | Data arrangement method and system |
-
2023
- 2023-04-26 CN CN202310464508.6A patent/CN116521917A/en active Pending
Cited By (2)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN117453936A (en) * | 2023-10-19 | 2024-01-26 | 山东三木众合信息科技股份有限公司 | Data arrangement method and system |
| CN117453936B (en) * | 2023-10-19 | 2024-03-26 | 山东三木众合信息科技股份有限公司 | Data arrangement method and system |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN108898137B (en) | Natural image character recognition method and system based on deep neural network | |
| CA3100642A1 (en) | Multi-sample whole slide image processing in digital pathology via multi-resolution registration and machine learning | |
| CN108537751B (en) | Thyroid ultrasound image automatic segmentation method based on radial basis function neural network | |
| CN107358141B (en) | Data identification method and device | |
| CN113011357A (en) | Depth fake face video positioning method based on space-time fusion | |
| CN111382766A (en) | Equipment fault detection method based on fast R-CNN | |
| CN111753873A (en) | Image detection method and device | |
| CN113763348A (en) | Image quality determination method and device, electronic equipment and storage medium | |
| CN112528939A (en) | Quality evaluation method and device for face image | |
| CN116521917A (en) | Picture screening method and device | |
| CN108647703B (en) | Saliency-based classification image library type judgment method | |
| CN113269903A (en) | Face recognition class attendance system | |
| CN113076860B (en) | Bird detection system under field scene | |
| CN113128522B (en) | Target identification method, device, computer equipment and storage medium | |
| CN114037868B (en) | Image recognition model generation method and device | |
| CN114529852A (en) | Video data-based carry-over detection and analysis method | |
| CN115457620A (en) | User expression recognition method and device, computer equipment and storage medium | |
| CN112906707A (en) | Semantic segmentation method and device for surface defect image and computer equipment | |
| CN114494103A (en) | Defect detection method and device | |
| CN114419711B (en) | Concentration degree identification method based on AI (artificial intelligence) education system | |
| Bhanumathi et al. | Underwater Fish Species Classification Using Alexnet | |
| CN117037049B (en) | Image content detection method and system based on YOLOv5 deep learning | |
| CN117079255B (en) | Fatigue driving detection method based on face recognition and voice interaction | |
| CN112307908B (en) | Video semantic extraction method and device | |
| CN115170837A (en) | Image recognition method based on artificial intelligence |
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 |