Poster fine production pre-screening method, system and terminal

Abstract

The application discloses a poster fine production pre-screening method and system and a terminal, relates to the poster processing technical field, and comprises the following steps: periodically extracting and downloading poster material images according to task batches; performing basic quality detection, removing images that do not meet preset requirements, and obtaining poster material images that pass the basic screening; performing person recognition and analysis, using a face recognition model to frame all faces, calculating face feature vectors for all faces, clustering all faces, classifying different people, and obtaining classified poster material images; performing comprehensive evaluation on the classified and grouped poster material images based on preset multidimensional features, calculating the quality scores of each image, and outputting the quality score results; and according to the score results, fine selection marking is performed on high-quality images with quality scores reaching a predetermined value. The application solves the problems of low poster material screening efficiency, inaccurate quality evaluation and resource waste in the prior art.

Description

Technical Field

[0001] This invention relates to the field of poster processing technology, and in particular to a method, system, smart terminal, and storage medium for pre-screening of posters for refined production. Background Technology

[0002] Current film and television content platforms require a large number of high-quality posters for user display and recommendation. However, the number of original video footage images is enormous, and their quality varies greatly. If all image materials were used in the poster production process, it would not only consume a lot of computing resources, but the resulting low-quality posters would also be unusable on the operating platform, causing serious waste of resources.

[0003] The common practice in the industry is to manually select high-quality materials, but this method is inefficient and inconsistent in standards, failing to meet the needs of processing massive amounts of content. Simple automated selection methods struggle to accurately assess multi-dimensional quality factors such as image composition and character performance, leading to discrepancies between the selection results and actual requirements. Therefore, there is an urgent need for a technical solution that can automatically select and accurately evaluate high-quality materials before poster production, addressing the problems of low efficiency in poster material selection, inaccurate quality assessment, and resource waste in existing technologies. Summary of the Invention

[0004] The technical problem to be solved by the present invention is to provide a method, system, smart terminal and storage medium for pre-production screening of posters to address the problems and defects of the prior art. The present invention provides a method that can automatically screen and accurately evaluate high-quality materials before poster production, solving the problems of low efficiency in poster material screening, inaccurate quality evaluation and waste of resources in the prior art.

[0005] The technical solution adopted by this invention to solve the problem is as follows: A method for pre-screening in refined poster production, comprising: Periodically extract poster materials to be processed and download poster material images in batches according to tasks; Perform basic quality checks on the downloaded poster material images, remove images that do not meet the preset requirements, and obtain poster material images that pass the basic screening; The poster material images that have passed the basic screening are subjected to person recognition and analysis. The face recognition model is used to select all faces, calculate the face feature vectors for all faces, and cluster all faces to classify different people and obtain the classified poster material images. The poster material images after classification and grouping are comprehensively evaluated based on preset multi-dimensional features, and the quality score of each image is calculated and output. Based on the scoring results, high-quality images that reach a predetermined quality score are selected and marked.

[0006] The aforementioned poster refinement production pre-screening method includes the following steps: performing basic quality checks on downloaded poster material images, removing images that do not meet preset requirements, and obtaining poster material images that pass the basic screening. The poster material image is subjected to dynamic image judgment to detect whether it is an image from an inappropriate platform; whether the image length exceeds a predetermined size; whether the image color is monochromatic; whether no face is found or whether the number of faces in the image exceeds a predetermined number; whether the face is vertical; whether the face is within the bounding box; face gap check; face point-to-head ratio check; main character face check; back of head check; background tilt check; pose violation check; and posture detection check. Exclude animated images and grayscale images with monochromatic colors; check the aspect ratio of the images and filter out images with distorted proportions; exclude images from specified sources that do not meet the requirements; obtain poster material images that pass the basic screening.

[0007] The aforementioned poster refinement production pre-screening method includes the following steps: performing character recognition and analysis on the poster material images that have passed the basic screening; using a face recognition model to select all faces; calculating facial feature vectors for all faces; and clustering all faces to classify different people and obtain the classified poster material images. For poster images that have passed the basic screening, a face detection algorithm is used to identify people in the images and extract key features of the people, including: face pose, number, position and proportion. Clustering algorithms are used to identify and classify different people. Based on the number of people, posture, and facial orientation, the images are grouped according to the way people are grouped, resulting in classified and grouped poster material images. The identification and classification filtering rules include: detecting facial pose and excluding images with severe side profiles or abnormal angles; limiting the number of faces in the image; analyzing the position of faces in the image and excluding images with faces located at the edge or cropped; evaluating the proportion of the head in the overall image and excluding images with too small a proportion or severe occlusion; and ensuring that the image contains a clear subject.

[0008] The aforementioned poster refinement production pre-screening method includes the following steps: comprehensively evaluating the categorized and grouped poster material images based on preset multi-dimensional features, calculating the quality score of each image, and outputting the quality score results. The poster images after being categorized and grouped are initially scored based on the basic characteristics of the number of people, their postures, and their expressions. The poster images are evaluated to determine whether the figures are looking directly at the camera. Images that do look directly at the camera receive bonus points. For poster images containing multiple characters, bonus points are awarded based on the evaluation of the interactions between the characters. Bonus points are awarded based on the analysis of the visual appeal, background complexity, and lighting effects. The various indicators are weighted and calculated to obtain a quality score for each image, and the final quality score result is output as the basis for selection.

[0009] The aforementioned poster refinement production pre-screening method, wherein the step of selecting and marking high-quality images that reach a predetermined quality score based on the scoring results includes: Based on the scoring results, high-quality images that reach the predetermined quality score are selected and marked, and the database status is updated for priority use in subsequent poster production processes.

[0010] The aforementioned poster refinement production pre-screening method, wherein the steps of periodically extracting poster materials to be processed and downloading poster material images according to task batches include: The system obtains information on the content type, promotional theme, target audience, and specific style preferences of the current poster production task from the operational target input, which is used as the basis for dynamically adjusting filtering preferences. We continuously collect user feedback data and operational performance data for published posters; and conduct data analysis, combining operational goal inputs with rule-based expert systems and statistical analysis methods to identify different content types, promotional themes, and style preferences.

[0011] The aforementioned poster refinement production pre-screening method further includes the following steps: comprehensively evaluating the categorized and grouped poster material images based on preset multi-dimensional features, calculating a quality score for each image, and outputting the quality score results. Based on the input of operational goals, and the identification of different content types, promotional themes and style preferences, the parameters and weights of basic quality detection, character identification and analysis, and multi-dimensional feature comprehensive evaluation are dynamically adjusted to generate a personalized screening rule and scoring weight for the current task. In the basic quality screening stage, personalized rules generated by a dynamic screening preference generator are used for preliminary screening. In the person recognition and analysis stage, the parameters for face detection and analysis are adjusted according to individual preferences; In the multi-dimensional quality scoring stage, the weights of each scoring indicator are adjusted according to individual preferences to ensure that the final score accurately reflects the task's specified requirements for image quality. The process involves a multi-stage, dynamically adjusted selection process where high-quality images that meet the requirements are labeled and added to the database. This process, along with continuous optimization of the learning model of the dynamic selection preference generator, forms a closed-loop adaptive optimization process.

[0012] A poster refinement production pre-screening system, wherein the system includes: The poster material extraction and download module is used to periodically extract poster materials to be processed and download poster material images in batches according to tasks. The basic quality inspection module is used to perform basic quality inspection on the downloaded poster material images, remove images that do not meet the preset requirements, and obtain poster material images that pass the basic screening. The face classification processing module is used to identify and analyze people in the poster material images that have passed the basic screening. It uses a face recognition model to select all faces, calculates face feature vectors for all faces, and clusters all faces to classify different people and obtain the classified poster material images. The comprehensive scoring module is used to comprehensively evaluate the classified and grouped poster material images based on preset multi-dimensional features, calculate the quality score of each image, and output the quality score results. The Featured Labeling module is used to select and label high-quality images that have reached a predetermined quality score based on the scoring results.

[0013] A smart terminal includes a memory and one or more programs, wherein one or more programs are stored in the memory and configured to be executed by one or more processors, the one or more programs comprising the method for performing any one of the methods.

[0014] A computer-readable storage medium, wherein, when instructions in the storage medium are executed by a processor of an electronic device, the electronic device is enabled to perform any of the methods described above.

[0015] The beneficial effects of this invention are as follows: This invention provides a method, system, intelligent terminal, and storage medium for pre-screening in refined poster production. By performing multi-stage automated quality assessment on original image materials, this invention pre-screens high-quality images that meet requirements, significantly improving subsequent poster production efficiency and reducing resource consumption. Furthermore, this invention also has the following advantages: 1) This invention adopts a pre-screening mechanism, introducing multi-stage fine screening before poster production, which effectively reduces the entry of low-quality materials into the production process and significantly reduces system resource consumption.

[0016] 2) This invention adopts a multi-dimensional evaluation system, combining multiple indicators such as basic image quality, human characteristics, and composition aesthetics to establish a comprehensive evaluation system, ensuring that the screening results meet actual operational needs.

[0017] 3) This invention uses automated grouping processing. Through person recognition and clustering technology, it automatically classifies images by person groups, providing targeted materials for poster production in different scenarios.

[0018] 4) This invention adopts a quantitative scoring mechanism and establishes a standardized scoring system, making the material selection process quantifiable and traceable, thereby improving operational decision-making efficiency.

[0019] By implementing this solution, resource consumption in the poster production process can be reduced, while ensuring that the quality of the produced posters meets operational standards, thereby significantly improving content production efficiency and user experience. Attached Figure Description

[0020] To more clearly illustrate the technical solutions in the embodiments of the present invention or the prior art, the drawings used in the description of the embodiments or the prior art will be briefly introduced below. Obviously, the drawings described below are only some embodiments recorded in the present invention. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0021] Figure 1 This is a flowchart illustrating the poster refinement production pre-screening method provided in Embodiment 1 of the present invention.

[0022] Figure 2 This is a flowchart illustrating the poster refinement production pre-screening method provided in Embodiment 2 of the present invention.

[0023] Figure 3 The principle block diagram of the poster fine production pre-screening system provided by the present invention.

[0024] Figure 4 This is a block diagram illustrating the internal structure of a smart terminal provided in an embodiment of the present invention. Detailed Implementation

[0025] To make the objectives, technical solutions, and advantages of this invention clearer and more explicit, the invention will be further described in detail below with reference to the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are merely illustrative of the invention and are not intended to limit the invention.

[0026] It should be noted that if the embodiments of the present invention involve directional indicators (such as up, down, left, right, front, back, etc.), the directional indicators are only used to explain the relative positional relationship and movement of the components in a certain specific posture (as shown in the figure). If the specific posture changes, the directional indicators will also change accordingly.

[0027] Current film and television content platforms require a large number of high-quality posters for user display and recommendation. However, the raw video footage is vast in quantity and quality. Introducing all images into the poster production process would not only consume significant computing resources but also result in low-quality posters that cannot be displayed on the platform, leading to severe resource waste. Current technology often involves manual selection of high-quality footage, which is inefficient and inconsistent in standards, failing to meet the demands of processing massive amounts of content. Simple automated selection methods, on the other hand, struggle to accurately assess the composition, character performance, and other multi-dimensional quality factors of images, resulting in discrepancies between the selection results and actual needs. Therefore, a technological solution is urgently needed that can automatically select and accurately evaluate high-quality footage before poster production.

[0028] like Figure 1 As shown, an embodiment 1 of the present invention provides a method for pre-screening of posters for refined production, comprising the following steps: Step S100: Periodically extract the poster materials to be processed and download the poster material images according to task batches; The poster materials to be processed refer to the original poster materials that have been planned and require subsequent design (such as modification, layout, optimization) but have not yet been processed, including poster image files (JPG, PNG, etc.), basic templates, and material elements (LOGO, background image, text material, etc.).

[0029] The task batch refers to the grouping of poster materials to be processed according to certain rules (such as project category, processing priority, and delivery time). Each batch contains several poster material tasks of the same type or with the same requirements, which facilitates centralized management and efficient processing.

[0030] The poster material images refer to the visual image files related to the poster. They are the core carrier of poster design and include image files that can be directly downloaded and used, such as blank poster templates, finished poster drafts, and material illustrations.

[0031] In this embodiment of the invention, poster materials to be processed can be periodically extracted from the database and downloaded to the processing environment in batches according to the task, in preparation for subsequent screening and evaluation.

[0032] Specifically, poster materials that require further design processing can be selected from sources such as material libraries and requirement submission channels according to fixed cycles (such as daily or weekly). Then, according to the pre-defined task batches, all poster material images in the corresponding batch can be downloaded to the designated working path to prepare for subsequent design, modification and other work.

[0033] This invention avoids material accumulation and reduces the need for emergency handling by periodically extracting materials; centralized downloading in batches saves operation time and improves pre-preparation efficiency compared to downloading individual images in a scattered manner, thus improving overall work efficiency; furthermore, each batch of this invention corresponds to a clear task requirement, and subsequent design tasks can be assigned and processing progress tracked according to batches, ensuring that poster design work proceeds in an orderly manner, avoiding omissions, and facilitating process control.

[0034] Step S200: Perform basic quality checks on the downloaded poster material images, remove images that do not meet the preset requirements, and obtain poster material images that pass the basic screening. In this embodiment of the invention, the downloaded poster material images undergo basic quality checks, including checks on format compliance, image integrity, and basic composition rules, to eliminate low-quality images that clearly do not meet preset requirements. The preset requirement filtering rules include: excluding animated formats and grayscale images with monochromatic colors; checking the image aspect ratio and filtering images with severely distorted proportions; and excluding images from specific sources that do not meet the requirements.

[0035] In this embodiment of the invention, all downloaded poster material images undergo preliminary and rapid quality checks. This stage primarily focuses on the basic attributes of the images, such as file format, color mode, aspect ratio, and source compliance. These basic rules allow for the rapid elimination of clearly non-compliant low-quality images, such as animated GIFs, grayscale images, severely disproportionate images, or images from non-compliant sources. This step significantly reduces the amount of data required for subsequent processing and lowers computational resource consumption.

[0036] This invention can ensure the final quality of posters from the source through basic screening, ensuring that the posters meet the usage requirements.

[0037] Step S300: Perform character recognition and analysis on the poster material images that have passed the basic screening. Use a face recognition model to select all faces, calculate face feature vectors for all faces, and cluster all faces to classify different people and obtain the classified poster material images. This step primarily involves character recognition and analysis of the poster images. Images that have passed the initial screening are analyzed to extract key character features, and then categorized and grouped based on factors such as the number of people, their posture, and facial orientation. Specific screening rules include: detecting facial posture and excluding images with severely abnormal profiles or angles; limiting the number of faces in the image; analyzing the position of faces within the image and excluding images with faces at the edges or cropped; assessing the proportion of heads in the overall image and excluding images that are too small or severely obscured; and ensuring that the image contains a clearly defined subject.

[0038] In this embodiment of the invention, for images that have passed the basic screening, a face detection algorithm is used to identify people in the images and extract their key features, such as facial pose, number, position, and proportion. Simultaneously, a clustering algorithm is used to identify and classify different people, grouping the images according to the way people are grouped. This stage aims to ensure that the images contain a clear main subject and to exclude images with poorly presented figures or flawed composition, such as images with severely slanted faces, too many or too few people, or images with faces located at the edge or occluded.

[0039] This invention uses person recognition and clustering technology to automatically classify images by person groups, providing targeted materials for poster production in different scenarios.

[0040] Step S400: After classifying and grouping the poster material images, perform a comprehensive evaluation based on preset multi-dimensional features, calculate the quality score of each image, and output the quality score results. In this embodiment, the preset multi-dimensional features include: the character looking directly at the camera, the character's facial expression, the character's interaction, the visual aesthetics, and the background.

[0041] In this step of the embodiment, the poster material images after classification and grouping are comprehensively evaluated based on multiple dimensions such as image composition, character performance, and visual appeal, and a quality score is calculated for each image. Specific scoring rules include: scoring based on basic features such as the number of characters, posture, and expression; considering the degree of interaction between characters and the camera (e.g., looking directly at the camera); evaluating the interactive relationships between characters (e.g., hugging, eye contact); analyzing visual appeal, background complexity, and lighting effects; and calculating a final score based on all indicators, which serves as the basis for selection.

[0042] Specifically, this invention comprehensively evaluates each poster image after classification and grouping, based on multiple dimensions such as composition, character performance, aesthetic appeal, background complexity, and lighting effects. This stage calculates a standardized quality score that comprehensively considers factors such as the number of characters, their posture, expressions, interaction with the camera, interpersonal relationships, and overall aesthetics. This quantitative scoring mechanism allows for a more objective and accurate assessment of image quality, ensuring that the selection results meet actual operational needs.

[0043] Step S500: Based on the scoring results, select and mark high-quality images that have reached a predetermined quality score.

[0044] In this step, high-quality images are selected and marked based on the scoring results, and the database status is updated for priority use in subsequent poster production processes.

[0045] Specifically, in the selection, labeling, and result entry stage, this embodiment of the invention selects and labels high-quality images that meet a preset threshold based on the results of multi-dimensional quality scoring, and updates the status in the database. These images labeled as high-quality will be given priority for use in subsequent poster production processes, thereby significantly improving poster production efficiency, reducing resource consumption, and ensuring that the quality of the produced posters meets operational standards.

[0046] For example, a film and television content platform needs to create a large number of promotional posters for a new drama, "Starry Sea." The original video footage of the drama contains tens of thousands of stills and screenshots, including both high-quality, beautiful images and a large number of low-quality images unsuitable for poster use. The specific implementation process using this invention is as follows: S1. Task Scheduling and Material Acquisition: In this embodiment, the system first identifies image material tasks related to "Starry Sea" from the database. Subsequently, the system batch downloads tens of thousands of original images related to the series to temporary storage space.

[0047] S2, Basic Quality Screening (Coarse Screening Process), including: First, exclude animated images and grayscale images: In this embodiment, the system detected that about 5,000 of them were animated images in GIF format or black and white grayscale images with a single color, and these images were directly excluded.

[0048] Then the aspect ratio was checked: the system in this embodiment further detected that about 3,000 images had a severely distorted aspect ratio (e.g., extremely narrow or extremely wide), and these images were also filtered out.

[0049] Further exclusion of specific sources: In this embodiment, the system identified approximately 1,000 images from unofficial screenshot tools, which were either watermarked or low-resolution, and these images were also marked as unacceptable.

[0050] After basic quality screening, approximately 41,000 images remain to proceed to the next stage, S3.

[0051] S3. Perform character content analysis for face classification, including: First, face pose detection is performed. In this embodiment, the system performs face detection on the remaining images and finds that the faces of people in about 2,000 images are seriously lateral or have abnormal angles, making them unsuitable as the main subject of the poster and thus excluded.

[0052] Then, the number of faces is limited. In this embodiment, the system detects that there are too many faces in about 1,500 images (such as group photos, which are not suitable for highlighting the main character) or too few faces (such as pure landscape photos, which do not meet the requirements of a character poster), and excludes them.

[0053] Next, the position and proportion of the face are evaluated. In this embodiment, the system analyzes the remaining images and finds that in about 1,000 images, the face is located at the edge of the image or is cropped, or the proportion of the head is too small / too large, and these images are excluded.

[0054] To ensure the main subject is present, the system ultimately selected approximately 36,500 images, all of which contained a clearly defined main subject and whose depiction met the basic requirements.

[0055] Meanwhile, the system in this embodiment identifies and clusters the figures in these images, classifying them into different combinations such as "single protagonist," "two protagonists," and "protagonist and supporting character," providing materials for the subsequent production of different types of posters. Then proceed to S4.

[0056] S4. Conduct a multi-dimensional quality scoring process, specifically including: First, a basic feature score is performed. In this embodiment, the system scores each image based on basic features such as the number of people, posture, and expression. For example, images with vivid expressions and relaxed postures of the main characters receive high scores.

[0057] The system also scores the interaction between the person and the camera. In this embodiment, the system evaluates whether the person is looking directly at the camera. Images that look directly at the camera are generally more attractive and receive bonus points.

[0058] The system also scores the interaction between characters. In this embodiment, for images containing multiple characters, the system evaluates the interaction between characters (such as hugging or eye contact). Images with natural interaction and rich emotions receive bonus points.

[0059] In addition, the system scores the image's aesthetics and lighting effects. This embodiment analyzes the image's composition, color saturation, background complexity, and lighting effects. For example, images with a golden ratio composition, vibrant colors, a simple background, and excellent lighting effects receive high scores.

[0060] Then, combining the scores above, the system in this embodiment performs a weighted calculation of each indicator to obtain a final standardized quality score for each image. For example, an image with a subject looking directly at the camera, a vivid expression, perfect composition, and excellent lighting might receive 95 points. Then proceed to S5.

[0061] S5. Selected images are marked and stored in the database. In this embodiment, the system sets a threshold, for example, 80 points. All images scoring above 80 points are marked as "selected materials." Ultimately, approximately 15,000 high-quality images are marked as selected, and the database status is updated. These images will be used preferentially in subsequent poster production processes. Simultaneously, the system frees up temporary storage space, reducing resource consumption.

[0062] As can be seen from the above, through this refined production process, the platform can efficiently select high-quality images from massive amounts of raw materials, ensuring the quality of the promotional posters for "Starry Sea" and significantly improving content production efficiency while reducing the waste of computing resources.

[0063] The present invention will be further described in detail below through another specific application embodiment; like Figure 2 As shown in a further embodiment of the present invention, the poster refinement production pre-screening method includes the following steps: S10. Download all materials corresponding to the current film / TV series to temporary storage space; S20. Perform animation judgment on the poster material image, detect whether it is an image from an inappropriate platform; detect whether the image length exceeds a predetermined size; detect whether the image color is monochromatic; determine whether no face is found or whether the number of faces in the image exceeds a predetermined number; detect whether the face is vertical; detect whether the face is within the edge frame; perform face gap check; perform face-to-head ratio check; perform main character face check; perform back of head check; perform background tilt check; perform posture violation check and pose detection check; exclude grayscale images with monochromatic animation format and color; check the image aspect ratio and filter images with distorted aspect ratio; exclude images from specified sources that do not meet the requirements; obtain the poster material image that passes the basic screening.

[0064] In this embodiment, the "animated image judgment" refers to detecting whether the poster material image is a dynamic image (such as GIF format). The core is to distinguish between static and dynamic images to prevent animated images from entering the subsequent design stage (if the design only requires static materials). "Unsuitable platform images" refers to poster material images obtained from unauthorized, illegal, low-quality, or non-compliant platforms (such as pirated material sites or infringing websites). Using such images may pose copyright risks.

[0065] The predetermined size refers to the standard for the length (or file size) of the poster material image set in advance based on design requirements, storage requirements, and usage scenarios (such as online publishing or printing). If it exceeds the standard, it is considered unqualified.

[0066] The edge frame refers to a pre-defined virtual frame (such as a reserved frame for faces in the main area of ​​the poster) used to regulate the position of faces, ensuring that faces are in a suitable visual position on the poster, improving aesthetics and information transmission efficiency. The face gap check refers to detecting the distance between multiple faces in the image to ensure that the face spacing is reasonable, not crowded or overlapping, and avoiding affecting the visual effect of the poster.

[0067] The facial proportion check refers to detecting the ratio of key facial features (such as eyes, nose, and mouth) to the entire head in an image, ensuring that the facial proportions are harmonious, without distortion, and conform to normal visual aesthetics. The protagonist's face check refers to identifying the protagonist's face (such as a product spokesperson or key figure) in the poster image, confirming that the protagonist's face is clear and prominent, meeting the core requirements of the poster design. The back of the head check refers to detecting whether there are faces in the image that only show the back of the head and do not reveal the front / side view; such faces cannot convey the character's features and do not meet the requirements for poster character display.

[0068] The background tilt check refers to detecting whether the background of the poster material image is horizontal or vertical, so as to avoid the overall visual imbalance and unattractiveness of the poster caused by a tilted background.

[0069] The posture violation check refers to detecting whether the posture of people in the image conforms to company regulations and public order and good morals (such as avoiding indecent postures and violations) to ensure that the poster conveys a positive image.

[0070] The posture detection check refers to detecting the overall posture (such as standing or sitting posture) of people in the image to ensure that the posture is natural, coordinated, and in line with the style and requirements of the poster design.

[0071] The grayscale image refers to an image containing only black, white, and gray colors, without any other colored elements. It is a typical type of image with a single color, and most poster designs should exclude such images (unless there are special design requirements). The aspect ratio mismatch refers to the length and width ratio of the poster material image not conforming to the pre-set standards (such as the commonly used 16:9, 4:3, 1:1), resulting in the image being stretched, distorted, or unable to fit the design template.

[0072] In this embodiment, the poster material images that pass the basic screening refer to poster material images that, after passing all the above-mentioned detection items, meet all preset requirements, have no violations or quality issues, and can be directly entered into subsequent design stages (such as layout, optimization, and modification).

[0073] In this embodiment of the invention, 13 specific tests are carried out on each of the downloaded poster material images, namely: whether it is an animated image, whether it is an image from an inappropriate platform, whether the length exceeds the predetermined size, whether the color is uniform, whether the number of faces is compliant (no faces or too many faces are unqualified), whether the faces are vertical, whether the faces are within the edge frame, whether the gaps between faces are reasonable, whether the ratio of faces to heads is coordinated, whether the main character's face is clear and prominent, whether there are faces on the back of heads, whether the background is tilted, and whether the character's posture and pose are compliant. Then, filter and remove unqualified images: Based on the above detection results, uniformly exclude materials that do not meet the requirements, including: animated images, grayscale images with single color, images with unbalanced aspect ratios, unqualified images from sources other than those specified, and all images that fail the detection of faces, backgrounds, poses, etc. Then, qualified materials are retained. All poster material images that have passed the above tests and eliminations meet the preset standards and are thus "poster material images that have passed the basic screening". They can be directly entered into the subsequent design stage, which ensures the quality of the finished poster and the efficiency of the design.

[0074] In this embodiment, by setting clear and quantifiable basic screening rules, a large number of obviously unacceptable low-quality images can be quickly and efficiently eliminated in the initial stage of the poster production process. This significantly reduces the amount of data in subsequent complex processing stages, thereby greatly reducing the consumption of computing resources and improving overall screening efficiency. At the same time, these rules ensure that images entering subsequent processes at least meet basic visual and format requirements, laying the foundation for subsequent refined evaluation.

[0075] An example of basic quality screening (coarse screening process): Suppose a film and television content platform needs to create a large number of promotional posters for a new drama, "Starry Sea." The original video footage of the drama contains tens of thousands of stills and screenshots, including: Excluding GIFs and grayscale images: The system detected approximately 5,000 images that were either GIFs or monochrome grayscale images, which were directly excluded. Checking aspect ratio: The system further detected approximately 3,000 images with severely distorted aspect ratios (e.g., extremely narrow or extremely wide), which were also filtered out. Excluding specific sources: The system identified approximately 1,000 images from unofficial screenshot tools, with watermarks or low resolution, which were also marked as unacceptable. After basic quality screening, approximately 41,000 images remain for the next stage.

[0076] S30. Perform face classification on the poster material images that have passed the basic screening. Use face detection algorithms to identify people in the images and extract key features of the people, including: face pose, number, position and proportion; use clustering algorithms to identify and classify different people. Group the images according to the number of people, pose and face information, and obtain the classified and grouped poster material images; the identification and classification screening rules include: detecting face pose and excluding images with severe side profiles or abnormal angles; limiting the number of faces in the image; analyzing the position of faces in the image and excluding images with faces on the edge or cropped; evaluating the proportion of the head in the overall image and excluding images with too small a proportion or severe occlusion; ensuring that the image contains a clear subject.

[0077] In this step of the embodiment, face classification is performed on the poster material images that have passed the basic screening. A face detection algorithm is then used to identify people in the images and extract key features of the people. These key features include: Posture: front view, side view, head up, head down, etc.; Quantity: How many people are in one picture; Location: The face is in the middle, on the side, or in the corner of the image; Proportion: The size of the face in the entire image; In this embodiment, the specific rules for character content analysis include: 1) Detect facial posture and exclude images with severe side profile or abnormal angle: That is, the system of this invention will recognize the posture of the face and remove those images of people with severe side profile or abnormal angle that are not suitable as the main body of the poster.

[0078] 2) Limit the number of faces in the image. In this embodiment, the system will set a threshold for the number of faces in the image to exclude images with too many faces (such as group photos, which are not suitable for highlighting the main character) or too few faces (such as pure landscape photos, which do not meet the requirements of character posters).

[0079] 3) Analyze the position of the face in the image and exclude images where the face is located at the edge or is cropped. This implementation system will evaluate the position of the face in the image and remove images where the face is located at the edge or is cropped.

[0080] 4) Evaluate the proportion of the human head in the overall image and exclude images that are too small or severely obscured: The system will calculate the proportion of the human head in the overall image and exclude images that are too small or too large, as well as those that are severely obscured.

[0081] 5) Ensure that the images contain a clear subject: By following the above rules, ensure that the selected images contain a clear and well-presented subject.

[0082] This step, through refined and multi-dimensional analysis and screening of the figures in the images, ensures that the selected images meet the high requirements of poster production in terms of the figures' performance. This solves the problem that simple automated screening methods struggle to accurately assess the figures' performance, avoiding low-quality posters caused by poor poses, inappropriate numbers, unreasonable positions, or unbalanced proportions of the figures. This significantly improves the usability of the poster materials and the visual appeal of the final poster.

[0083] In this embodiment, based on character recognition and analysis, the images are further classified and grouped. Using character recognition and clustering technology, images are automatically categorized by character combinations, providing targeted materials for poster production in different scenarios. For example, the system performs character recognition and clustering on images that have passed basic screening, classifying them into different combination categories such as "single protagonist," "two protagonists," and "protagonist and supporting character," providing materials for the production of different types of posters. For instance, when the platform needs to create a poster highlighting the male protagonist's personal charm, it can directly select high-scoring images from the "single protagonist" category; when it needs to create a poster expressing the emotions between the male and female protagonists, it can select from the "two protagonists" category.

[0084] This invention introduces character recognition and clustering technologies to achieve automated and intelligent image classification and grouping. This transforms the selected high-quality materials from a random collection into a structured organization based on character combinations (such as "single protagonist," "two protagonists," "protagonist and supporting character," etc.). This pre-classification significantly improves the efficiency and flexibility of the subsequent poster production process. Operators can quickly find corresponding high-quality materials based on different promotional needs (e.g., creating solo posters, couple posters, or group posters), eliminating the tedious work of manual screening and classification, further reducing resource consumption and improving the targeting of content production.

[0085] S40. After classifying and grouping the poster images, perform image quality scoring. Initial scoring is based on the number of people, their postures, and facial expressions in the images. Assess whether any figures in the poster images are looking directly at the camera; images with such direct eye contact receive bonus points. For poster images containing multiple figures, bonus points are awarded based on the interaction between the figures. Bonus points are also awarded based on the aesthetic appeal, background complexity, and lighting effects. All indicators are weighted and calculated to obtain a quality score for each image, and the final quality score is output as a selection criterion.

[0086] In this embodiment, the poster material images after classification and grouping are subjected to multi-dimensional image quality scoring. The specific rules for multi-dimensional quality scoring in this embodiment include: 1) Scoring based on basic features such as the number of people, posture, and expression: In this embodiment, the system makes a preliminary score based on the basic features such as the number of people, posture, and expression in the image.

[0087] 2) Consider the degree of interaction between the person and the camera (e.g., looking directly at the camera): In this embodiment, the system evaluates whether the person is looking directly at the camera. Images that look directly at the camera are usually more attractive and will receive a bonus.

[0088] 3) Evaluate the interaction between people (such as hugging, eye contact, etc.): In this embodiment, for images containing multiple people, the system will evaluate the interaction between people (such as hugging, eye contact). Images with natural interaction and rich emotions will receive bonus points.

[0089] 4) Analyze the aesthetics, background complexity, and lighting effects of the image: In this embodiment, the system analyzes the composition, color saturation, background complexity, and lighting effects of the image. For example, images with a golden ratio composition, vivid colors, a simple background, and excellent lighting effects will receive high scores.

[0090] 5) Calculate the final score based on all indicators, which will be used as the basis for selection: that is, the system of this invention will calculate the weighted average of the above indicators to obtain the final standardized quality score of each image.

[0091] For example, following the above embodiments, the system of the present invention performs multi-dimensional quality scoring on 36,500 images analyzed through character content, including: Basic Feature Scoring: The system scores each image based on basic features such as the number of people, their posture, and their expressions. For example, images with vivid expressions and relaxed postures of the main characters receive high scores.

[0092] Interaction between person and camera: The system evaluates whether the person is looking directly at the camera. Images that show the person looking directly at the camera are generally more attractive and receive a bonus.

[0093] Interactions between characters: For images containing multiple characters, the system evaluates the interactions between them (such as hugging, eye contact), and images with natural and emotionally rich interactions receive bonus points.

[0094] Image aesthetics and lighting effects: The system analyzes the image's composition, color saturation, background complexity, and lighting effects. For example, images with a golden ratio composition, vibrant colors, a simple background, and excellent lighting effects receive high scores.

[0095] Overall Score: The system calculates a weighted average of the various indicators to arrive at a final standardized quality score for each image. For example, an image with a subject looking directly at the camera, a vivid expression, perfect composition, and excellent lighting might receive a score of 95.

[0096] As can be seen, this embodiment of the invention, by establishing a comprehensive, objective, and quantitative multi-dimensional scoring system, can more accurately evaluate the overall quality of images, ensuring that the selection results highly align with actual operational needs. It not only considers the basic performance of the individuals but also deeply evaluates the interaction between the individuals and the camera, the relationships between the individuals, and the overall aesthetics of the image, thus overcoming the difficulty of accurately evaluating multi-dimensional quality factors using simple automated methods. This refined scoring mechanism ensures that the final selected poster materials are not only technically qualified but also possess higher artistic merit and appeal, significantly improving the quality of the produced posters and the user experience.

[0097] In a further embodiment of the present invention, the poster refinement production pre-screening method, wherein the step of selecting and marking high-quality images that reach a predetermined quality score based on the scoring results includes: Based on the scoring results, high-quality images that reach the predetermined quality score are selected and marked, and the database status is updated for priority use in subsequent poster production processes.

[0098] In this embodiment of the invention, during the result processing stage, high-quality images that achieve a predetermined quality score are selected and marked based on the scoring results, and the database status is updated for priority use in subsequent poster production processes. Furthermore, after updating the database with the scoring results and selection status, temporary storage space is released simultaneously, effectively optimizing the utilization of system resources.

[0099] When processing massive amounts of raw image data, the demand for temporary storage space is enormous. If this space is not released in a timely manner, it will lead to continuous occupation and waste of storage resources. The mechanism of timely release of temporary storage space in this invention ensures that the system can quickly reclaim and reuse resources after completing the task, reducing operating costs and improving the overall operating efficiency and stability of the system. Its optimization effect on resource management is particularly significant in large-scale, high-concurrency content processing scenarios.

[0100] In another embodiment of the present invention, the poster refinement production pre-screening method, in order to further solve the problem of material screening preference drift in the multi-stage screening process, adopts the following steps: S90. Obtain the content type, promotional theme, target audience, and specific style preference information of the current poster production task inputted by the operational goals, and use it as the basis for dynamically adjusting the filtering preferences; In a specific implementation of this invention, an operational target input interface can be added, allowing operators or content creators to input information such as the content type of the current poster production task (e.g., romance, action, documentary), the promotional theme (e.g., suspense, heartwarming, epic), the target audience (e.g., young people, families), and specific style preferences (e.g., retro, futuristic, minimalist). This information will serve as an important basis for dynamically adjusting filtering preferences.

[0101] When a new poster production task is initiated, information such as the content type, promotional theme, target audience, and style preferences can be submitted through the operational target input interface.

[0102] S91. Continuously collect user feedback data and operational performance data of published posters; and conduct data analysis, combining operational goal inputs, and using rule-based expert systems and statistical analysis methods to identify different content types, promotional themes and style preferences.

[0103] In this step of the embodiment, user feedback data (such as click-through rate, dwell time, number of shares, and sentiment of comments) and operational performance data (such as conversion rate and exposure) of published posters will be continuously collected. By analyzing this data, combined with user / operational goal input, and using rule-based expert systems and simple statistical analysis methods, the system identifies which image features (such as composition, facial expressions, color, and lighting) are more popular under different content types, promotional themes, and style preferences, and which adjustments to the filtering rules can bring better results. For example, if images of characters with slightly shadowed faces or blurred expressions have a higher click-through rate in a poster for a suspense film, the system will learn that when processing suspense film materials, the filtering requirements for "facial clarity" or "expression clarity" can be appropriately relaxed.

[0104] S92. Based on the input of operational goals, and the identification of different content types, promotional themes and style preferences, dynamically adjust the parameters and weights of basic quality detection, character identification and analysis, and multi-dimensional feature comprehensive evaluation to generate a personalized screening rule and scoring weight for the current task. The operational objectives in this step refer to the business goals to be achieved by this poster production, such as brand promotion, event traffic generation, product promotion, holiday marketing, and customer conversion. The content type refers to the classification of the poster's purpose, such as main visual poster, detail image, cover image, event image, product image, and character poster. The promotional theme refers to the core theme around which this poster revolves, such as new product launch, holiday promotion, brand image, public service announcement, and conference / event promotion.

[0105] The style preference refers to design style requirements, such as minimalist, high-end, lively, business, Chinese trend, cartoon, formal, youthful, etc. The dynamically adjusted parameters and weights refer to changing the strictness and importance of the screening criteria according to different tasks. For example, character posters place more emphasis on the face, while product posters place more emphasis on clarity and provenance.

[0106] The comprehensive evaluation refers to scoring multiple dimensions such as format, source, color, face, posture, and background together to determine whether the material is qualified. The personalized filtering rules are not a fixed set of rules, but rather filtering standards specifically generated for the current task, which are more in line with the needs of this operation.

[0107] The scoring weights represent the importance of different detection items. For example: posters featuring people → high proportion of faces; product posters → high proportion of clarity.

[0108] In this embodiment, the system does not use a fixed set of rules to filter all poster materials. Instead, it first considers the operational goals, content type, promotional theme, and style preferences, and automatically adjusts the standards and scoring weights of basic quality detection, character recognition, and multi-dimensional features to generate a set of "exclusive filtering rules + scoring weights" that are only applicable to the current task, making the filtering more accurate and more in line with the needs of this poster.

[0109] S93. In the basic quality screening stage, personalized rules generated by the dynamic screening preference generator are used for preliminary screening. In this embodiment of the invention, during the basic quality screening stage, the system no longer uses fixed general rules, but instead applies personalized rules generated by a dynamic screening preference generator for preliminary screening. For example, for art films, the restriction on "single color" can be appropriately relaxed, allowing more grayscale or monochromatic images to pass; for specific style requirements, the acceptable range of "aspect ratio" can be adjusted.

[0110] S94. In the person recognition and analysis stage, adjust the parameters of face detection and analysis according to personalized preferences; In this embodiment of the invention, during the character content analysis stage, the system adjusts the parameters for face detection and analysis based on personalized preferences. For example, for tasks that require creating a sense of mystery, the system relaxes the requirements for facial clarity and integrity. For suspense films, the detection threshold for "facial pose" can be adjusted to allow more side profiles or partially obscured images; for ensemble dramas, the upper limit of the "number of faces limit" can be increased.

[0111] S95. In the multi-dimensional quality scoring stage, the weights of each scoring indicator are adjusted according to individual preferences to ensure that the final score accurately reflects the task's specified requirements for image quality. In this embodiment of the invention, during the multi-dimensional quality scoring stage, the weights of each scoring indicator are adjusted according to personalized preferences to ensure that the final score can more accurately reflect the specific requirements of the task for image quality.

[0112] Specifically, the weighting of various scoring indicators such as "composition," "character performance," and "visual appeal" can be adjusted based on the content type and promotional theme. For example, "character posture" and "lighting effects" may have higher weightings in action films, while "interactions between characters" and "expressions" may have higher weightings in romance films.

[0113] S96, and after dynamic adjustment and multi-stage screening, high-quality images that meet the requirements are labeled and stored in the database, and the learning model of the dynamic screening preference generator is continuously optimized to form a closed-loop adaptive optimization process.

[0114] In this embodiment, during the selection, labeling, and feedback learning phase, high-quality images, after undergoing dynamic and multi-stage screening, are labeled and added to the database. Simultaneously, user feedback and operational performance data from these images after subsequent releases are collected by the "Historical Data and Feedback Analysis Module" to continuously optimize the learning model of the "Dynamic Screening Preference Generator," forming a closed-loop adaptive optimization process.

[0115] As can be seen, the optimized embodiment of the present invention, by introducing a dynamic adaptive filtering preference adjustment module, makes the filtering rules and scoring criteria for poster materials no longer fixed, but can be intelligently and personally adjusted according to the specific content type, promotional theme and user feedback.

[0116] This embodiment addresses the preference drift in basic quality screening. By allowing operators to input specific style preferences and learning from historical data, the system can dynamically adjust its tolerance for basic features such as color and composition proportions, avoiding the misscreening of images with artistic value or specific style potential due to overly rigid general rules.

[0117] This embodiment addresses the preference drift in character content analysis by dynamically adjusting the detection thresholds for facial pose, quantity, position, and proportion. This allows the system to better adapt to the specific needs of different content types for character portrayal. For example, it can retain more mysterious character images for suspense films or more richly detailed scenes for ensemble dramas.

[0118] This embodiment addresses the preference drift in multi-dimensional quality scoring by dynamically adjusting the weights of various scoring indicators. This allows the system to more accurately match the emphasis on image quality for specific promotional goals. For example, an action film emphasizing visual impact might focus more on lighting effects and character movement in its poster materials, while a romance film emphasizing emotional resonance might focus more on character expressions and interactions.

[0119] In summary, the dynamic adaptive mechanism of this invention enables the screening results to more accurately match actual operational needs, solving the problems of low material utilization and insufficient poster innovation caused by fixed screening preferences, thereby improving the attractiveness and promotional effect of the posters. At the same time, it avoids overly complex artificial intelligence algorithms, and has a high degree of rationality and practicality.

[0120] Exemplary device like Figure 3 As shown, this embodiment of the invention provides a poster fine-tuning pre-screening system, which includes: The poster material extraction and download module 310 is used to periodically extract poster materials to be processed and download poster material images in batches according to tasks. The basic quality inspection module 320 is used to perform basic quality inspection on the downloaded poster material images, remove images that do not meet the preset requirements, and obtain poster material images that pass the basic screening. The face classification processing module 330 is used to identify and analyze people in the poster material images that have passed the basic screening. It uses a face recognition model to select all faces, calculates face feature vectors for all faces, clusters all faces, classifies different people, and obtains the classified poster material images. The comprehensive scoring module 340 is used to comprehensively evaluate the poster material images after classification and grouping based on preset multi-dimensional features, calculate the quality score of each image, and output the quality score results. The selective labeling module 350 is used to selectively label high-quality images that have reached a predetermined quality score based on the scoring results, as described above.

[0121] Based on the above embodiments, the present invention also provides a smart terminal, the principle block diagram of which can be as follows: Figure 4 As shown. The intelligent terminal includes a processor, memory, network interface, display screen, and database connected via a system bus. The processor provides computing and control capabilities. The memory includes a non-volatile storage medium and internal memory. The non-volatile storage medium stores the operating system and computer programs. The internal memory provides an environment for the operation of the operating system and computer programs in the non-volatile storage medium. The network interface is used to communicate with external terminals via a network connection. When the computer program is executed by the processor, it implements a poster refinement production pre-screening method. The database of the intelligent terminal stores the poster refinement production pre-screening program.

[0122] Those skilled in the art will understand that Figure 4 The block diagram shown is merely a partial structural diagram related to the present invention and does not constitute a limitation on the smart terminal to which the present invention is applied. A specific smart terminal may include more or fewer components than those shown in the figure, or combine certain components, or have different component arrangements.

[0123] In one embodiment, a smart terminal is provided, including a memory and one or more programs, wherein the one or more programs are stored in the memory and configured to be executed by one or more processors. The one or more programs include instructions for performing the following operations: Periodically extract poster materials to be processed and download poster material images in batches according to tasks; Perform basic quality checks on the downloaded poster material images, remove images that do not meet the preset requirements, and obtain poster material images that pass the basic screening; The poster material images that have passed the basic screening are subjected to person recognition and analysis. The face recognition model is used to select all faces, calculate the face feature vectors for all faces, and cluster all faces to classify different people and obtain the classified poster material images. The poster material images after classification and grouping are comprehensively evaluated based on preset multi-dimensional features, and the quality score of each image is calculated and output. Based on the scoring results, high-quality images that reach a predetermined quality score are selected and marked, as described above.

[0124] The step of performing basic quality checks on the downloaded poster material images, removing images that do not meet preset requirements, and obtaining poster material images that pass the basic screening includes: The poster material image is subjected to dynamic image judgment to detect whether it is an image from an inappropriate platform; whether the image length exceeds a predetermined size; whether the image color is monochromatic; whether no face is found or whether the number of faces in the image exceeds a predetermined number; whether the face is vertical; whether the face is within the bounding box; face gap check; face point-to-head ratio check; main character face check; back of head check; background tilt check; pose violation check; and posture detection check. Exclude animated images and grayscale images with monochromatic colors; check the aspect ratio of the images and filter out images with distorted proportions; exclude images from specified sources that do not meet the requirements; obtain poster material images that pass the basic screening.

[0125] The steps of performing character recognition and analysis on the poster material images that have passed the basic screening, using a face recognition model to select all faces, calculating face feature vectors for all faces, clustering all faces, classifying different people, and obtaining classified poster material images include: For poster images that have passed the basic screening, a face detection algorithm is used to identify people in the images and extract key features of the people, including: face pose, number, position and proportion. Clustering algorithms are used to identify and classify different people. Based on the number of people, posture, and facial orientation, the images are grouped according to the way people are grouped, resulting in classified and grouped poster material images. The identification and classification filtering rules include: detecting facial pose and excluding images with severe side profiles or abnormal angles; limiting the number of faces in the image; analyzing the position of faces in the image and excluding images with faces located at the edge or cropped; evaluating the proportion of the head in the overall image and excluding images with too small a proportion or severe occlusion; and ensuring that the image contains a clear subject.

[0126] The step of comprehensively evaluating the classified and grouped poster material images based on preset multi-dimensional features, calculating the quality score of each image, and outputting the quality score results includes: The poster images after being categorized and grouped are initially scored based on the basic characteristics of the number of people, their postures, and their expressions. The poster images are evaluated to determine whether the figures are looking directly at the camera. Images that do look directly at the camera receive bonus points. For poster images containing multiple characters, bonus points are awarded based on the evaluation of the interactions between the characters. Bonus points are awarded based on the analysis of the visual appeal, background complexity, and lighting effects. The various indicators are weighted and calculated to obtain a quality score for each image, and the final quality score result is output as the basis for selection.

[0127] The step of selecting and marking high-quality images that reach a predetermined quality score based on the scoring results includes: Based on the scoring results, high-quality images that reach the predetermined quality score are selected and marked, and the database status is updated for priority use in subsequent poster production processes.

[0128] The aforementioned poster refinement production pre-screening method, wherein the steps of periodically extracting poster materials to be processed and downloading poster material images according to task batches include: The system obtains information on the content type, promotional theme, target audience, and specific style preferences of the current poster production task from the operational target input, which is used as the basis for dynamically adjusting filtering preferences. We continuously collect user feedback data and operational performance data for published posters; and conduct data analysis, combining operational goal inputs with rule-based expert systems and statistical analysis methods to identify different content types, promotional themes, and style preferences.

[0129] The step of comprehensively evaluating the classified and grouped poster material images based on preset multi-dimensional features, calculating the quality score of each image, and outputting the quality score results further includes: Based on the input of operational goals, and the identification of different content types, promotional themes and style preferences, the parameters and weights of basic quality detection, character identification and analysis, and multi-dimensional feature comprehensive evaluation are dynamically adjusted to generate a personalized screening rule and scoring weight for the current task. In the basic quality screening stage, personalized rules generated by a dynamic screening preference generator are used for preliminary screening. In the person recognition and analysis stage, the parameters for face detection and analysis are adjusted according to individual preferences; In the multi-dimensional quality scoring stage, the weights of each scoring indicator are adjusted according to individual preferences to ensure that the final score accurately reflects the task's specified requirements for image quality. The process involves a multi-stage, dynamically adjusted selection process where high-quality images that meet the requirements are labeled and added to the database. This process, along with continuous optimization of the learning model of the dynamic selection preference generator, forms a closed-loop adaptive optimization process.

[0130] In other embodiments, this application proposes a computer-readable storage medium that, when the instructions in the storage medium are executed by the processor of an electronic device, enables the electronic device to perform the above-described method.

[0131] Those skilled in the art will understand that all or part of the processes in the methods of the above embodiments can be implemented by a computer program instructing related hardware. The computer program can be stored in a non-volatile computer-readable storage medium. When executed, the computer program can include the processes of the embodiments of the above methods. Any references to memory, storage, databases, or other media used in the embodiments provided by this invention can include non-volatile and / or volatile memory. Non-volatile memory can include read-only memory (ROM), programmable ROM (PROM), electrically programmable ROM (EPROM), electrically erasable programmable ROM (EEPROM), or flash memory. Volatile memory can include random access memory (RAM) or external cache memory. By way of illustration and not limitation, RAM is available in various forms, such as static RAM (SRAM), dynamic RAM (DRAM), synchronous DRAM (SDRAM), dual data rate SDRAM (DDRSDRAM), enhanced SDRAM (ESDRAM), synchronous link DRAM (SLDRAM), Rambus direct RAM (RDRAM), direct memory bus dynamic RAM (DRDRAM), and memory bus dynamic RAM (RDRAM), etc.

[0132] The above description is merely an embodiment of this application and is not intended to limit the scope of protection of this application. Various modifications and variations can be made to this application by those skilled in the art. Any modifications, equivalent substitutions, improvements, etc., made within the spirit and principles of this application should be included within the scope of protection of this application.

Claims

1. A method for pre-screening in refined poster production, characterized in that, include: Periodically extract poster materials to be processed and download poster material images in batches according to tasks; Perform basic quality checks on the downloaded poster material images, remove images that do not meet the preset requirements, and obtain poster material images that pass the basic screening; The poster material images that have passed the basic screening are subjected to person recognition and analysis. The face recognition model is used to select all faces, calculate the face feature vectors for all faces, and cluster all faces to classify different people and obtain the classified poster material images. The poster material images after classification and grouping are comprehensively evaluated based on preset multi-dimensional features, and the quality score of each image is calculated and output. Based on the scoring results, high-quality images that reach a predetermined quality score are selected and marked.

2. The poster refinement production pre-screening method according to claim 1, characterized in that, The step of performing basic quality checks on the downloaded poster material images, removing images that do not meet preset requirements, and obtaining poster material images that pass the basic screening includes: The poster material image is subjected to dynamic image judgment to detect whether it is an image from an inappropriate platform; whether the image length exceeds a predetermined size; whether the image color is monochromatic; whether no face is found or whether the number of faces in the image exceeds a predetermined number; whether the face is vertical; whether the face is within the bounding box; face gap check; face point-to-head ratio check; main character face check; back of head check; background tilt check; pose violation check; and posture detection check. Exclude animated images and grayscale images with monochromatic colors; check the aspect ratio of the images and filter out images with distorted proportions; exclude images from specified sources that do not meet the requirements; obtain poster material images that pass the basic screening.

3. The poster refinement production pre-screening method according to claim 1, characterized in that, The steps of performing character recognition and analysis on the poster material images that have passed the basic screening, using a face recognition model to select all faces, calculating facial feature vectors for all faces, clustering all faces, and classifying different people to obtain the classified poster material images include: For poster images that have passed the basic screening, a face detection algorithm is used to identify people in the images and extract key features of the people, including: face pose, number, position and proportion. Clustering algorithms are used to identify and classify different people. Based on the number of people, posture, and facial orientation, the images are grouped according to the way people are grouped, resulting in classified and grouped poster material images. The identification and classification filtering rules include: detecting facial pose and excluding images with severe side profiles or abnormal angles; limiting the number of faces in the image; analyzing the position of faces in the image and excluding images with faces located at the edge or cropped; evaluating the proportion of the head in the overall image and excluding images with too small a proportion or severe occlusion; and ensuring that the image contains a clear subject.

4. The poster refinement production pre-screening method according to claim 1, characterized in that, The steps of comprehensively evaluating the classified and grouped poster material images based on preset multi-dimensional features, calculating the quality score of each image, and outputting the quality score results include: The poster images after being categorized and grouped are initially scored based on the basic characteristics of the number of people, their postures, and their expressions. The poster images are evaluated to determine whether the figures are looking directly at the camera. Images that do look directly at the camera receive bonus points. For poster images containing multiple characters, bonus points are awarded based on the evaluation of the interactions between the characters. Bonus points are awarded based on the analysis of the visual appeal, background complexity, and lighting effects. The various indicators are weighted and calculated to obtain a quality score for each image, and the final quality score result is output as the basis for selection.

5. The poster refinement production pre-screening method according to claim 1, characterized in that, The step of selecting and marking high-quality images that reach a predetermined quality score based on the scoring results includes: Based on the scoring results, high-quality images that reach the predetermined quality score are selected and marked, and the database status is updated for priority use in subsequent poster production processes.

6. The poster refinement production pre-screening method according to claim 1, characterized in that, The steps of periodically extracting poster materials to be processed and downloading poster material images in batches according to the task include: The system obtains information on the content type, promotional theme, target audience, and specific style preferences of the current poster production task from the operational target input, which is used as the basis for dynamically adjusting filtering preferences. We continuously collect user feedback data and operational performance data for published posters; and conduct data analysis, combining operational goal inputs with rule-based expert systems and statistical analysis methods to identify different content types, promotional themes, and style preferences.

7. The poster refinement production pre-screening method according to claim 6, characterized in that, The step of comprehensively evaluating the classified and grouped poster material images based on preset multi-dimensional features, calculating the quality score of each image, and outputting the quality score results further includes: Based on the input of operational goals, and the identification of different content types, promotional themes and style preferences, the parameters and weights of basic quality detection, character identification and analysis, and multi-dimensional feature comprehensive evaluation are dynamically adjusted to generate a personalized screening rule and scoring weight for the current task. In the basic quality screening stage, personalized rules generated by a dynamic screening preference generator are used for preliminary screening. In the person recognition and analysis stage, the parameters for face detection and analysis are adjusted according to individual preferences; In the multi-dimensional quality scoring stage, the weights of each scoring indicator are adjusted according to individual preferences to ensure that the final score accurately reflects the task's specified requirements for image quality. The process involves a multi-stage, dynamically adjusted selection process where high-quality images that meet the requirements are labeled and added to the database. This process, along with continuous optimization of the learning model of the dynamic selection preference generator, forms a closed-loop adaptive optimization process.

8. A poster refinement production pre-screening system, characterized in that, The system includes: The poster material extraction and download module is used to periodically extract poster materials to be processed and download poster material images in batches according to tasks. The basic quality inspection module is used to perform basic quality inspection on the downloaded poster material images, remove images that do not meet the preset requirements, and obtain poster material images that pass the basic screening. The face classification processing module is used to identify and analyze people in the poster material images that have passed the basic screening. It uses a face recognition model to select all faces, calculates face feature vectors for all faces, and clusters all faces to classify different people and obtain the classified poster material images. The comprehensive scoring module is used to comprehensively evaluate the classified and grouped poster material images based on preset multi-dimensional features, calculate the quality score of each image, and output the quality score results. The Featured Labeling module is used to select and label high-quality images that have reached a predetermined quality score based on the scoring results.

9. A smart terminal, characterized in that, It includes a memory and one or more programs, wherein one or more programs are stored in the memory and configured to be executed by one or more processors, wherein the one or more programs include methods for performing any one of claims 1-7.

10. A computer-readable storage medium, characterized in that, When the instructions in the storage medium are executed by the processor of the electronic device, the electronic device is able to perform the method as described in any one of claims 1-7.

Images