Method and device for generating a video based on artificial intelligence to build a rare bird gallery

Abstract

The application provides a method and device for constructing a rare bird image library based on artificial intelligence generated video, and belongs to the technical field of rare bird image augmentation. The method generates dynamic video of rare birds according to sample images of the rare birds through an artificial intelligence video generation model, and then constructs generated images of the rare birds through the dynamic video, so that the rare bird images can be augmented, and the research on the rare birds is facilitated.

Description

Technical Field

[0001] This invention relates to the field of rare bird image augmentation technology, and in particular to a method and apparatus for constructing a rare bird image library based on artificial intelligence-generated videos. Background Technology

[0002] Rare birds are generally those bird species that are few in number in nature, have a limited distribution range, or face a high risk of extinction.

[0003] Because rare birds are few in number and depend on special habitats, they are difficult to observe, resulting in a limited number of images and making research on them challenging. Summary of the Invention

[0004] This invention proposes a method and apparatus for constructing a rare bird image library based on artificial intelligence-generated videos. By using an artificial intelligence video generation model to generate dynamic videos of rare birds based on sample images of rare birds, and then constructing generated images of rare birds from the dynamic videos, it is possible to expand the images of rare birds, which is beneficial to the research of rare birds.

[0005] To achieve the above objectives, the present invention adopts the following technical solution: In a first aspect, this invention provides a method for constructing a rare bird image library based on artificial intelligence-generated videos, comprising: acquiring an initial image set of rare birds; the initial image set including multiple sample images of each rare bird species; for each rare bird species, the multiple sample images all feature the rare bird as the main subject, showcasing multiple key parts, eye features, and feather features of the rare bird in a standing or flying posture from multiple real-life perspectives. Based on the initial image set of rare birds and an artificial intelligence video generation model, dynamic videos of each rare bird species are determined; the artificial intelligence video generation model is constructed based on a diffusion model. Using the dynamic videos of each rare bird species, a rare bird image database is determined; the rare bird image database includes multiple generated images obtained frame-by-frame from the dynamic videos of each rare bird species.

[0006] In one implementation of the first aspect, based on an initial image set of rare birds and an artificial intelligence video generation model, dynamic videos of each rare bird species are determined, including: Step 1: For each rare bird species, based on multiple sample images of the rare bird and dynamic video generation prompts, the artificial intelligence video generation model is used to analyze the multiple sample images of the rare bird to generate candidate dynamic videos of the rare bird; the dynamic video generation prompts are used to guide the artificial intelligence video generation model to generate candidate dynamic videos of the rare bird. Step 2: Judge the candidate dynamic videos of rare birds: When the candidate dynamic videos of rare birds have flaws, edit the videos and determine the candidate dynamic videos of rare birds that are flaw-free after editing as the real videos of rare birds; flaws include incorrect eye features, incorrect feather features, and / or blurred bird outlines; when the candidate dynamic videos of rare birds have no flaws, determine the candidate dynamic videos of rare birds as the real videos of rare birds; when the species and / or behavior of the birds in the candidate dynamic videos of rare birds contradicts the species and / or behavior of rare birds, repeat Step 1.

[0007] In one implementation of the first aspect, the dynamic video generation prompt includes the species name of the rare bird, eye features, feather features, and the actions of the rare bird in the generated candidate dynamic video; the actions include flight actions, walking actions, running actions, foraging actions, grooming actions, courtship actions, and / or attack actions.

[0008] In one implementation of the first aspect, the AI ​​video generation model is the Veo model, the Sora model, the Gen-2 model, or MagicVideo V2.

[0009] In one implementation of the first aspect, multiple key parts include the crest, beak, neck, chest, abdomen, legs, claws, and tail feathers; feather characteristics include feather color, feather pattern, feather layering, and barb arrangement; eye characteristics include eye color and eye structure; and multiple shooting perspectives include a side view, a quarter-side level view, and an obliquely upward view.

[0010] Secondly, this invention provides an apparatus for constructing a rare bird image library based on artificial intelligence-generated videos, including an image set acquisition module, a dynamic video generation module, and a database construction module. The image set acquisition module is used to acquire an initial image set of rare birds; the initial image set includes multiple sample images of each rare bird species; for each rare bird species, the multiple sample images all feature the rare bird as the main subject, showcasing multiple key parts, eye features, and feather features of the rare bird in a standing or flying posture from various real-life perspectives. The dynamic video generation module is used to determine a dynamic video of each rare bird species based on the initial image set and the artificial intelligence video generation model; the artificial intelligence video generation model is constructed based on a diffusion model. The database construction module is used to determine a rare bird image database using the dynamic videos of each rare bird species; the rare bird image database includes multiple generated images obtained by frame-by-frame extraction from the dynamic videos of each rare bird species.

[0011] In one implementation of the second aspect, the dynamic video generation module is specifically used for: Step 1: For each rare bird species among multiple rare bird species, based on multiple sample images of the rare bird and dynamic video generation prompts, an artificial intelligence video generation model is used to analyze the multiple sample images of the rare bird to generate candidate dynamic videos of the rare bird; the dynamic video generation prompts are used to guide the artificial intelligence video generation model to generate candidate dynamic videos of the rare bird. Step 2: Judging the candidate dynamic videos of rare birds: When the candidate dynamic videos of rare birds have flaws, the candidate dynamic videos of rare birds are edited, and the candidate dynamic videos of rare birds without flaws after editing are determined as the dynamic videos of rare birds; flaws include incorrect eye features, incorrect feather features, and / or blurred bird outlines; when the candidate dynamic videos of rare birds have no flaws, the candidate dynamic videos of rare birds are determined as the dynamic videos of rare birds; when the species and / or behavior of the bird in the candidate dynamic videos of rare birds contradicts the species and / or behavior of rare birds, step 1 is repeated.

[0012] Thirdly, the present invention provides an electronic device including a processor and a memory coupled to the processor; the memory is used to store computer instructions, and when the electronic device is running, the processor executes the computer instructions stored in the memory to cause the electronic device to perform the method described in the first aspect above or any implementation thereof.

[0013] Fourthly, the present invention provides a computer-readable storage medium including computer program instructions that, when executed by a computer, cause the computer to perform the method described in the first aspect above or any implementation thereof.

[0014] Fifthly, the present invention provides a computer program product, including computer program instructions, which, when executed on a computer, cause the computer to perform the method described in the first aspect above or any implementation thereof.

[0015] The technical effects corresponding to the second to fifth aspects and their possible implementations can be referred to the above description of the technical effects of the first aspect and its possible implementations, and will not be repeated here.

[0016] Compared with the prior art, the present invention has the following beneficial effects.

[0017] The method for constructing a rare bird image library based on artificial intelligence-generated videos provided by this invention employs an artificial intelligence video generation model based on a diffusion model. It generates dynamic videos of rare birds based on an initial image set of rare birds with a small sample size. Then, it dynamically extracts multiple generated images of rare birds frame by frame from the dynamic videos to construct a rare bird image database containing a large number of rare bird images. This expands the image database of rare birds and is beneficial for the research of rare birds. Attached Figure Description

[0018] Figure 1 This is a schematic diagram of a method for constructing a rare bird image library based on artificial intelligence-generated videos, provided in an embodiment of this application. Figure 2 This is a schematic diagram of the device for constructing a rare bird image library based on artificial intelligence-generated videos, as provided in an embodiment of this application. Detailed Implementation

[0019] In the embodiments of this application, "and / or" indicates a relationship between objects. For example, A and / or B can represent the following three situations: A exists alone, B exists alone, and A and B exist simultaneously.

[0020] In the embodiments of this application, the terms "exemplary" or "for example" are used to indicate that something is an example, illustration, or description. Any embodiment or design that is described as "exemplary" or "for example" in the embodiments of this application should not be construed as being more preferred or advantageous than other embodiments or design. Specifically, the use of the terms "exemplary" or "for example" is intended to present the relevant concepts in a specific manner.

[0021] In the description of this invention, unless otherwise stated, "multiple" means two or more. For example, "multiple rare birds" means two or more rare birds.

[0022] The method and apparatus provided in this application relate to the construction of a rare bird image database. They can generate a large number of rare bird images in various specific poses from a small number of rare bird images in specific poses, thereby constructing a rare bird image database.

[0023] Understandably, rare birds, as important indicator species of ecosystems and key components of biodiversity, possess multidimensional research value. This value extends beyond the field of biology, reaching into areas such as ecological policy, technological innovation, and social ethics.

[0024] However, existing images of rare birds are scarce, and there are problems such as: 1. Insufficient sample size due to limited observation conditions in field shooting; 2. Difficulty in directly observing the habitat preferences and breeding behaviors of rare birds; 3. Low image clarity and accuracy of rare birds; 4. Difficulty in directly obtaining images of rare birds with specific movements and angles through field shooting; 5. Lack of images of certain bird species at specific growth stages or in different regions.

[0025] To address the problem in the background art that the limited number of rare birds and their dependence on specific habitats make observation of rare birds difficult, resulting in a small number of rare bird images and consequently hindering research on rare birds, this application provides a method and apparatus for constructing a rare bird image library based on artificial intelligence-generated videos. This method uses an artificial intelligence video generation model to generate dynamic videos of rare birds based on sample images, and then constructs generated images of rare birds from these dynamic videos. This expands the image library of rare birds, thereby facilitating research on them.

[0026] For example, the method for constructing a rare bird image library based on artificial intelligence-generated videos provided in this embodiment of the invention can be executed by an electronic device with processing capabilities, such as a computer or server. Taking a computer as an example, the hardware components of the computer may include: a processor, memory, a network interface, a user interface, a communication bus, etc.

[0027] The processor is used to control electronic devices to perform related processing and calculation tasks. The processor may include a central processing unit (CPU) or other processors. The processor may be single-core or multi-core. For example, the processor may include multiple CPUs.

[0028] Memory is used to store computer instructions and related data, such as initial image sets of rare birds, dynamic videos, and rare bird image databases. Memory can be random access memory (RAM), read-only memory (ROM), erasable programmable read-only memory (EPROM), flash memory, optical storage, magnetic disk storage media, or other magnetic storage devices, or any other medium capable of storing program code or data accessible by a computer. Optionally, memory can be integrated into the processor, or it can be independent of the processor.

[0029] A network interface is used for communication between a computer and other devices or communication networks. A network interface can be a transceiver with transmit and receive capabilities. Optionally, a network interface may include standard wired interfaces or wireless interfaces (such as Wi-Fi interfaces, Bluetooth interfaces, and 5G interfaces).

[0030] The communication bus is used to enable communication between different components. For example, the processor, memory, network interface and user interface mentioned above can be interconnected through the communication bus.

[0031] The user interface may include a display screen and an input unit (such as a keyboard). Optionally, the user interface may also include a standard wired interface or a wireless interface.

[0032] Those skilled in the art will understand that the computer described above may include more or fewer components, or combine certain components, or have different component arrangements; the embodiments of this application do not limit this.

[0033] like Figure 1 As shown, the method for constructing a rare bird image library based on artificial intelligence-generated videos provided in this application includes steps S101-S103.

[0034] S101. Obtain the initial image set of rare birds.

[0035] In this embodiment, the aforementioned rare birds may include the kakapo, Chinese merganser, heron, Sumatran ground cuckoo, and Philippine eagle, among others. The initial image set includes multiple sample images of each rare bird species. For each rare bird species, the sample images all feature the bird as the main subject, showcasing key features, eye characteristics, and feather characteristics of the bird in standing or flying postures from various real-life perspectives. These sample images can be either animated or static; this embodiment does not limit the format of the sample images.

[0036] Optionally, the aforementioned key body parts may include the crest, beak, neck, chest, abdomen, legs, claws, and tail feathers. The aforementioned feather characteristics may include feather color, feather pattern, feather layering, and barb arrangement. The aforementioned eye characteristics may include eye color and eye structure. The aforementioned various shooting perspectives may include a side view, a quarter-side eye-level view, and a slightly upward angled view.

[0037] Specifically, the selection principles for sample images in the initial image set of the above rare birds are given below.

[0038] (1) From the perspective of image elements: ① Subject: The main subject is a bird, showing the bird in flight or standing posture. The individual bird should occupy more than half of the picture to highlight its body shape, so that the AI ​​can restore the bird's actual body shape and detailed structure to the greatest extent.

[0039] ② Background: Choose bird images with blank, solid color, or transparent backgrounds whenever possible. If no such images are available, select images with relatively simple backgrounds and manually or using AI to extract the bird subjects. If the bird is a raptor, the background should ideally be a sky or open ground scene; if the bird is a forest bird, the foreground should be free of obstructing elements such as branches, and the background branches should minimize the area obscured by the bird. Overall, the background should be free of complex or unnecessary elements that interfere with the presentation of the subject, and the contrast between the bird and the background should be high (e.g., a dark bird against a light sky, or a light bird against a dark forest background). (2) From the perspective of visual features: ① Clarity: Key parts (beak, claws, flight feathers) are clearly detailed, and feather textures (such as barb arrangement and spots) are discernible. This provides AI models with more accurate and detailed features (such as feather layers and eye structure), helping the models learn more realistic bird forms and generate higher-quality new images. Even if the image clarity is low or blurry (such as images of some birds of prey), it can still provide information such as the basic shape, color, and basic features of the bird, with a complete main outline.

[0040] ② Viewpoint selection: mainly from the side, with a few from a slightly overhead angle (showing the full view of the wings when in flight).

[0041] (3) From the perspective of specific morphological characteristics of birds: ① Posture: Images that correspond to the most common behaviors of the bird species (such as standing or flying) should be selected. Images with small movements (such as scratching, flapping wings, etc.) are difficult for AI to identify key subject features and movements when used as modeling reference images, and the generated models are more likely to have distorted or incorrect shapes.

[0042] ②Crown: The crest features are obvious (if the crest is not obvious due to the shooting angle or bird species, it can be ignored).

[0043] ③Beak: The base of a bird's beak has a clear outline where it connects to its head.

[0044] ④ Neck (throat + front of neck), chest, abdomen.

[0045] Neck: From a side or frontal view, the neck feathers are arranged in a straight line, and the throat and foreneck feathers are of a single color or have simple stripes.

[0046] Chest and abdomen: The chest is densely feathered, showing a full muscle outline (especially the chest of birds of prey is broad), while the abdominal feathers are smooth.

[0047] ⑤ Tail feathers Tail feathers: All are in a spread or semi-spread state (such as forked tail feathers during flight, or horizontal tail feathers during gliding), with clear edges on the feather vanes and no breaks or abnormal missing parts.

[0048] ⑥ Legs and claws Legs: In images of a standing posture, the legs are mostly bare (without feathers covering the feet), and the scale patterns on the shanks are visible (such as reticulated or transverse scales). The leg color is mostly gray, yellow, or orange (contrasting with the environment).

[0049] Claws: The claws are all in a clenched or grasping state (such as grasping a branch or prey), with sharp, curved tips. Some pictures show webbed feet (such as waterbirds).

[0050] S102. Based on the initial image set of rare birds and the artificial intelligence video generation model, determine the dynamic video of each rare bird among a variety of rare birds.

[0051] For example, S102 above includes the following steps 1 to 2.

[0052] Step 1: For each rare bird species among the various rare bird species, generate prompts based on multiple sample images and dynamic videos of the rare bird species. Use an artificial intelligence video generation model to analyze multiple sample images of rare birds to generate candidate dynamic videos of rare birds.

[0053] Specifically, for each rare bird species among a variety of rare birds, multiple sample images of the aforementioned rare birds and dynamic video generation prompts are input into an artificial intelligence video generation model. Guided by the dynamic video generation prompts, the artificial intelligence video generation model generates dynamic videos of the aforementioned rare birds with corresponding actions based on the multiple sample images of the rare birds.

[0054] In this embodiment, the aforementioned AI video generation model is constructed based on a diffusion model. Optionally, the AI ​​video generation model can be a Veo model, a Sora model, a Gen-2 model, or a MagicVideo V2 model, etc. Specifically, the Veo and Sora models are existing diffusion models developed by OpenAI, the Veo model is an existing diffusion model developed by Google DeepMind, and the Gen-2 model is an existing diffusion model developed by Runway. Since these models are all prior art, this embodiment will not further elaborate on the model structure of the aforementioned AI video generation model.

[0055] The aforementioned dynamic video generation prompts are used to guide the AI ​​video generation model in generating candidate dynamic videos of rare birds. In one implementation, the dynamic video generation prompts include the species name of the rare bird, its eye features, feather features, and the actions of the rare bird in the generated candidate dynamic videos. These actions include flight actions, walking actions, running actions, foraging actions, preening actions, courtship actions, and / or attack actions.

[0056] In one application scenario, the specific requirements for generating prompts in the aforementioned dynamic video are as follows.

[0057] (1) The number of key verbs should be no more than 3.

[0058] When generating videos, the AI ​​video generation model primarily relies on the key verbs in the aforementioned dynamic video generation prompts. If the number of key verbs is too large and exceeds its comprehension capabilities, it will be unable to generate a video that meets the requirements. Therefore, the number of key verbs provided should not exceed three.

[0059] (2) The number of instructions should be controlled within 3.

[0060] When the number of commands is too large, the generated video is likely to differ significantly from the desired effect. To ensure the quality of the generated video, especially when performing large operations such as dynamic changes and angle variations, the number of commands given at one time should be limited to three or less.

[0061] (3) Detailed classification information such as order, family, and genus must be provided.

[0062] When generating dynamic videos, if only the description of "birds on the ground" is given, the AI ​​video generation model may easily misidentify them as thrushes. Therefore, it is necessary to supplement them with detailed classification information such as order, family, and genus. For example, the word "eagle" should be avoided in the instruction to prevent the AI ​​video generation model from automatically adding the characteristics of an eagle.

[0063] In another application scenario, the aforementioned flight maneuvers can include flapping flight (providing lift and thrust by flapping the wings up and down), gliding (spreading the wings and moving forward using updrafts or inertia without flapping), soaring (a special type of gliding, mainly hovering and ascending in updrafts, commonly seen in birds of prey such as eagles and vultures), hovering (maintaining an almost stationary position in the air, requiring high-speed flapping of the wings, such as hummingbirds and kingfishers), and diving (folding the wings from high altitude and flying vertically downwards at high speed, used for hunting or courtship).

[0064] The aforementioned walking and running actions can include walking (alternating between the left and right feet), jumping (jumping forward on the ground with both feet simultaneously), and running (the running of flightless birds at extremely high speeds).

[0065] The foraging actions mentioned above include pecking (using the beak to quickly and accurately peck at food), probing (inserting the beak into mud or soft soil to search for worms and crustaceans), tearing (birds of prey use their sharp claws to hold onto their prey and tear the meat with their hook-shaped beaks), filtering (filtering water and mud through a special beak structure, leaving behind food such as small shrimp and algae), tapping, and storing food.

[0066] The aforementioned grooming actions include feather preparation, bathing in water / sand, shaking feathers, applying oil, stretching, and sharpening beaks / claws.

[0067] The aforementioned courtship behaviors include calls and songs, posture displays (such as standing tall and proud, crouching / curling up), courtship dances, and feather displays.

[0068] The aforementioned aggressive actions include intimidation (spreading wings and beak, raising feathers), accompaniment (pretending to be injured to lure predators away from the nest or chicks), pecking, and scratching.

[0069] Optionally, the actions in the aforementioned dynamic video generation prompts may also include nest building and brooding actions. These nest building and brooding actions may include actions such as gathering materials, turning eggs, feeding, and cleaning the nest.

[0070] Step 2: Evaluate the candidate dynamic videos of rare birds: When there are flaws in the candidate dynamic videos of rare birds, the candidate dynamic videos of rare birds are edited, and the candidate dynamic videos of rare birds that are free of flaws after editing are determined as the dynamic videos of rare birds; flaws include incorrect eye features, incorrect feather features, and / or blurred bird outlines. If the candidate dynamic video of a rare bird is free of defects, the candidate dynamic video of the rare bird is determined as the dynamic video of the rare bird. When the species and / or behavior of a bird in a candidate dynamic video of a rare bird contradicts the species and / or behavior of a rare bird, step 1 is repeated.

[0071] Understandably, the judgment process in step 2 above can be implemented manually or through a large language model.

[0072] When the judgment is made manually, the specific judgment process for step 2 is as follows.

[0073] Judgment Criterion 1: Manually selected videos are considered acceptable if they are accurate and conform to the basic characteristics of birds, the behavioral characteristics of rare birds, and the differences between the generated videos are significant. In this case, the videos are imported into the database frame by frame as images. Specific criteria for good quality generation are as follows: (1) The resolution is moderate, the main subject occupies a reasonable proportion (e.g., occupies 1 / 5-1 / 3 of the image), and the background is simple.

[0074] (2) The bird’s posture, background and other features in the images generated from the same image are diverse, such as several images with different postures such as flying and standing, generated from the same image.

[0075] (3) The bird’s specific morphological characteristics are correct and there are no obvious errors (such as errors in feather texture or body structure); the bird’s posture conforms to the behavioral characteristics of rare birds and is consistent with reality (such as errors in flight posture or take-off posture).

[0076] (4) The generated habitat is consistent with the actual habitat of the bird.

[0077] (5) If the video has defects, the defective frames are removed using image processing tools and then imported into the database frame by frame. The criteria for defining defective videos are as follows.

[0078] (5.1) The color areas are unnatural and the feather details are distorted.

[0079] (5.2) The specific morphological characteristics of the bird are incorrect or have undergone relatively minor changes.

[0080] (5.3) Blurred boundaries and mixed colors.

[0081] (5.4) The bird’s behavior and posture do not conform to the behavioral characteristics of the bird.

[0082] Judgment condition 2: If the video has significant problems, such as a marked difference between the generated species and the original image, the video will be regenerated and its quality will be checked again. The criteria for defining a video with significant problems are as follows: Blurry or low resolution. This means that key features of the bird, such as feather texture, eye details, and beak shape, are not clear.

[0083] The content is incorrect or distorted, meaning that the bird's body structure (such as the number of wings and the proportion of legs) and posture (such as the reasonable angle of the wings when flying) are deformed or violate the laws of nature.

[0084] The labeling is incorrect. The bird species names labeled in the video do not match the actual species.

[0085] Over-editing or chaotic compositing. This means that the bird's natural form has been destroyed through exaggerated deformation (such as abnormally enlarging parts of the bird's body) or chaotic compositing (such as splicing the bird's head with other animal bodies).

[0086] The background obscures the bird significantly, meaning it obscures more than one-third of the bird's body or obscures key parts that affect species identification, such as the beak and neck.

[0087] The content is not realistic, meaning it does not reflect the behavioral characteristics of the target bird species (such as landfowl fishing).

[0088] When judging using a large language model, the specific judgment process for step 2 is as follows.

[0089] The candidate dynamic videos of the aforementioned rare birds and the judgment prompts are input into a large language model. Guided by the judgment prompts, the large language model judges whether the candidate dynamic videos of the rare birds have flaws or / and contradictory behaviors and outputs the results. It is understood that the judgment prompts may include the judgment conditions 1 and 2 obtained manually.

[0090] S103. Use dynamic videos of each rare bird species from a variety of rare bird species to establish a rare bird image database.

[0091] The aforementioned rare bird image database comprises multiple generated images obtained by frame-by-frame extraction from dynamic videos of each of a variety of rare bird species. Specifically, for each rare bird species, frame-by-frame images are extracted from the dynamic videos of these rare birds, and the extracted images are stored in the rare bird image database.

[0092] In summary, the method for constructing a rare bird image library based on AI-generated video provided in this application employs an AI video generation model based on a diffusion model. It generates dynamic videos of rare birds from an initial image set with a small sample size, and then dynamically extracts multiple generated images of rare birds frame by frame from the dynamic video to construct a rare bird image database containing a large number of rare bird images. This expands the number of rare bird images and is beneficial for rare bird research.

[0093] Accordingly, embodiments of this application provide an apparatus for constructing a rare bird image library based on artificial intelligence-generated videos, such as... Figure 2 As shown, it includes an image set acquisition module 501, a dynamic video generation module 502, and a database construction module 503.

[0094] The image set acquisition module 501 is used to acquire an initial image set of rare birds. The initial image set includes multiple sample images of each rare bird species. For each rare bird species, the multiple sample images all feature the rare bird as the main subject, showcasing multiple key parts, eye features, and feather features of the rare bird in a standing or flying posture from various real-world perspectives. For example, the image set acquisition module 501 is used to implement step S101 of the above method.

[0095] The dynamic video generation module 502 is used to determine the dynamic video of each rare bird species among a variety of rare birds based on an initial image set of rare birds and an artificial intelligence video generation model; the artificial intelligence video generation model is constructed based on a diffusion model. For example, the dynamic video generation module 502 is used to implement S102 of the above method.

[0096] The database construction module 503 is used to determine a rare bird image database by using dynamic videos of each rare bird species from a variety of rare bird species; the rare bird image database includes multiple generated images obtained by frame-by-frame extraction of dynamic videos of each rare bird species from a variety of rare bird species. For example, the database construction module 503 is used to implement S103 of the above method.

[0097] Optionally, the dynamic video generation module 502 is specifically used to: for each rare bird species among a variety of rare birds, based on multiple sample images of the rare bird species and dynamic video generation prompts, use an artificial intelligence video generation model to analyze the multiple sample images of the rare bird species to generate candidate dynamic videos of the rare bird species; the dynamic video generation prompts are used to guide the artificial intelligence video generation model to generate candidate dynamic videos of the rare bird species. The process involves judging candidate dynamic videos of rare birds: when a candidate dynamic video of a rare bird has flaws, it is edited, and the edited video without flaws is identified as the actual dynamic video of the rare bird. Flaws include incorrect eye features, incorrect feather features, and / or blurred bird outlines. When a candidate dynamic video of a rare bird has no flaws, it is identified as the actual dynamic video of the rare bird. If the species and / or behavior of a bird in a candidate dynamic video of a rare bird contradicts the species and / or behavior of a rare bird, the process is repeated for each rare bird species, generating prompts based on multiple sample images and dynamic videos of the rare bird, and using an artificial intelligence video generation model to analyze multiple sample images of the rare bird to generate candidate dynamic videos of the rare bird. For example, the dynamic video generation module 502 is specifically used to implement steps 1 to 2 of the above method.

[0098] The various modules of the above-mentioned device for constructing a rare bird image library based on artificial intelligence-generated videos can also be used to perform other steps in the above method embodiments. All relevant content involved in the above method embodiments can be referred to in the functional description of the corresponding functional module, and will not be repeated here.

[0099] This application also provides an electronic device, including: a processor and a memory coupled to the processor; the memory is used to store computer instructions, and when the electronic device is running, the processor executes the computer instructions stored in the memory to cause the electronic device to perform the methods described in the above embodiments. The processor can implement the image set acquisition module 501, the dynamic video generation module 502, and the database construction module 503; the memory can also be used to store an initial image set of rare birds, dynamic videos, and a rare bird image database, etc.

[0100] This application also provides a computer-readable storage medium including a computer program that, when run on a computer, performs the methods described in the above embodiments.

[0101] This application also provides a computer program product, which includes computer program instructions that, when run on a computer, execute the methods described in the above embodiments.

[0102] The various embodiments in this specification are described in a progressive manner. The same or similar parts between the various embodiments can be referred to each other. Each embodiment focuses on describing the differences from other embodiments.

[0103] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention, and not to limit them; although the present invention has been described in detail with reference to the foregoing embodiments, those skilled in the art should understand that modifications can still be made to the technical solutions described in the foregoing embodiments, or equivalent substitutions can be made to some of the technical features; and these modifications or substitutions do not cause the essence of the corresponding technical solutions to deviate from the spirit and scope of the technical solutions of the embodiments of this application.

Claims

1. A method for constructing a rare bird image library based on artificial intelligence-generated videos, characterized in that, include: Obtain an initial image set of rare birds; The initial image set includes multiple sample images of each rare bird species from a variety of rare bird species; For each of the various rare bird species, multiple sample images of the rare bird species are taken with the rare bird species as the main subject of the image, and multiple real-shot perspectives are used to show multiple key parts, eye features and feather features of the rare bird species in standing or flying postures. Based on the initial image set of the rare birds and the artificial intelligence video generation model, dynamic videos of each rare bird species among the various rare bird species are determined; the artificial intelligence video generation model is constructed based on a diffusion model. A rare bird image database is established using dynamic videos of each of the aforementioned rare bird species; the rare bird image database includes multiple generated images obtained by frame-by-frame extraction of dynamic videos of each of the aforementioned rare bird species.

2. The method as described in claim 1, characterized in that, The step of determining the dynamic video of each rare bird species among the various rare bird species based on the initial image set of the rare birds and the artificial intelligence video generation model includes: Step 1: For each of the various rare bird species, based on multiple sample images and dynamic video generation prompts of the rare bird species, the artificial intelligence video generation model is used to analyze the multiple sample images of the rare bird species to generate candidate dynamic videos of the rare bird species; the dynamic video generation prompts are used to guide the artificial intelligence video generation model to generate candidate dynamic videos of the rare bird species. Step 2: Evaluate the candidate dynamic videos of the rare birds: When there are defects in the candidate dynamic video of the rare bird, the candidate dynamic video of the rare bird is edited, and the candidate dynamic video of the rare bird that is free of defects after editing is determined as the dynamic video of the rare bird; the defects include incorrect eye features, incorrect feather features, and / or blurred bird outline; If the candidate dynamic video of the rare bird is free of defects, the candidate dynamic video of the rare bird is determined as the dynamic video of the rare bird. If the species and / or behavior of a bird in a candidate dynamic video of the rare bird contradicts the species and / or behavior of the rare bird, step 1 is repeated.

3. The method as described in claim 1 or 2, characterized in that, The dynamic video generation prompts include the species name of the rare bird, eye features, feather features, and the actions of the rare bird in the generated candidate dynamic videos; the actions include flight actions, walking actions, running actions, foraging actions, grooming actions, courtship actions, and / or attack actions.

4. The method as described in claim 1, characterized in that, The AI ​​video generation model is either the Veo model, the Sora model, the Gen-2 model, or MagicVideo V2.

5. The method as described in claim 1, characterized in that, The key parts include the crest, beak, neck, chest, abdomen, legs, claws, and tail feathers; The feather characteristics include feather color, feather pattern, feather layers, and barb arrangement; The eye features include eye color and eye structure; The various shooting perspectives include a side view, a quarter-side eye-level view, and an oblique overhead view.

6. A device for constructing a rare bird image library based on artificial intelligence-generated videos, characterized in that, It includes an image set acquisition module, a dynamic video generation module, and a database construction module; The image set acquisition module is used to acquire an initial image set of rare birds; the initial image set includes multiple sample images of each rare bird among a variety of rare birds; for each rare bird among the variety of rare birds, the multiple sample images of the rare bird all take the rare bird as the main subject of the picture and use multiple real-shot perspectives to show multiple key parts, eye features and feather features of the rare bird in a standing or flying posture. The dynamic video generation module is used to determine the dynamic video of each rare bird among the various rare bird species based on the initial image set of the rare birds and the artificial intelligence video generation model; the artificial intelligence video generation model is constructed based on the diffusion model. The database construction module is used to determine a rare bird image database by using dynamic videos of each of the various rare bird species; the rare bird image database includes multiple generated images obtained by frame-by-frame extraction of the dynamic videos of each of the various rare bird species.

7. An electronic device, characterized in that, The device includes a processor and a memory coupled to the processor; the memory is used to store computer instructions, which, when the electronic device is running, are executed by the processor to cause the electronic device to perform the method as described in any one of claims 1 to 5.

8. A computer-readable storage medium, characterized in that, It includes computer program instructions that, when executed by a computer, cause the computer to perform the method as described in any one of claims 1 to 5.

9. A computer program product, characterized in that, It includes computer program instructions that, when executed on a computer, cause the computer to perform the method as described in any one of claims 1 to 5.

Images