Video detection method and adult product

Abstract

The invention discloses a video detection method and an adult product, and the method comprises the steps: inputting a target video image into a convolutional neural network model, carrying out the feature extraction, and obtaining a feature image of each frame; the method comprises the following steps: firstly, extracting a feature image, then classifying the extracted feature image, scoring the classified feature image by a system, generating an auxiliary control signal according to a scoring result, sending the auxiliary control signal to external adult product equipment, and controlling a motion mode of the adult product equipment, thereby realizing synchronization or interaction between video content and equipment action.

Description

Technical Field

[0001] The present application relates to the field of online video detection, and in particular to a video detection method and adult products. Background Art

[0002] With the rapid development of society, adult products are also used by the general public. Adult products are also called sex toys, such as vibrators, massage sticks and simulated vaginas, etc. They are mainly used to enhance or assist adults' sexual experience, so as to facilitate self-stimulation or use with a partner to help achieve sexual pleasure.

[0003] When single people use adult products, they are usually used together with adult videos to enhance the sexual experience. At present, when using adult products, adult products cannot adjust the control signal according to the pictures in adult videos, resulting in a weak sense of interaction when single people use adult products while watching adult videos, and cannot adjust the adult products according to the content of the pictures and the sexual movements of the characters, thus affecting the user's experience of using adult products. Summary of the invention

[0004] In view of this, it is necessary to provide a video detection method and adult products to solve the above problems.

[0005] The present invention provides a video detection method, which comprises:

[0006] Acquire target video image;

[0007] Inputting the target video image into a convolutional neural network model, the convolutional neural network model performs feature extraction on each frame image in the target video image to obtain feature images of all frame images;

[0008] The convolutional neural network model detects the feature images of all the frame images to obtain classified feature images;

[0009] The classified feature images are scored to obtain scored feature images, an auxiliary control signal is obtained based on the scored feature images, and the auxiliary control signal is sent to an external adult product to control the movement of the external adult product.

[0010] In at least one embodiment of the present application, the convolutional neural network model detects the feature images of all the frame images to obtain the classified feature images, including:

[0011] Calculating the confidence score of each feature image in the feature images of all the frame images to obtain the confidence scores corresponding to all the frame images;

[0012] The feature images of all the frame images are classified according to the confidence scores corresponding to all the frame images to obtain classified feature images.

[0013] In at least one embodiment of the present application, the step of scoring the classified feature image to obtain a scored feature image, and obtaining an auxiliary control signal according to the scored feature image includes:

[0014] The confidence score corresponding to the classified feature image is compared with the confidence threshold one by one. If the confidence score corresponding to the classified feature image is greater than the confidence threshold, it is judged as an auxiliary image.

[0015] In at least one embodiment of the present application, the step of scoring the classified feature image to obtain a scored feature image, and obtaining an auxiliary control signal according to the scored feature image further includes:

[0016] The maximum confidence value is screened out from the confidence scores corresponding to the classified feature images, and the original image corresponding to the maximum confidence value is recorded as an enhanced auxiliary image, and an auxiliary control signal is generated according to the enhanced auxiliary image.

[0017] In at least one embodiment of the present application, the method further includes:

[0018] The accuracy is calculated based on the confidence score corresponding to the classified feature image and the confidence scores corresponding to all frame images. If the accuracy is greater than the accuracy threshold, an auxiliary control signal is output.

[0019] In at least one embodiment of the present application, the method further includes:

[0020] If the accuracy rate is less than the accuracy rate threshold, performing part feature extraction on the classified feature image to obtain a feature image after part extraction;

[0021] The feature image after the part extraction is scored to obtain a part classification video-assisted score.

[0022] In at least one embodiment of the present application, the method further includes:

[0023] The recall rate is calculated according to the part classification video-assisted score, the recall rate is used as the final score, and the final score is output.

[0024] In at least one embodiment of the present application, if the accuracy rate is less than the accuracy rate threshold, performing part feature extraction on the classified feature image, and obtaining the feature image after part extraction includes:

[0025] Obtaining a feature image corresponding to a maximum confidence score in the classified feature image, and performing part extraction to obtain a feature image after part extraction;

[0026] Acquire a part feature image of a previous frame and a part feature image of a subsequent frame of the feature image after the part extraction;

[0027] The confidence scores are performed on the part feature image of the previous frame, the part feature image of the next frame, and the feature image after the part extraction to obtain the confidence scores after the frame image verification.

[0028] In at least one embodiment of the present application, the step of calculating the recall rate according to the part classification video-assisted score, taking the recall rate as the final score, and outputting the final score comprises:

[0029] The recall rate is calculated based on the confidence score of the frame image after review and the confidence scores corresponding to all frame images. If the recall rate is greater than the recall threshold, a final control signal is generated based on the feature image corresponding to the maximum confidence score in the classified feature image, the part feature image of the previous frame, and the part feature image of the next frame, and the final control signal is sent to the external adult product.

[0030] An adult product, applied in any one of the above-mentioned video detection methods, the adult product comprising:

[0031] A receiving module, used for receiving an auxiliary control signal or a final control signal;

[0032] The motion module executes the auxiliary control signal or the final control signal to adjust the motion mode.

[0033] Implementing the video detection method of this embodiment will have at least the following beneficial effects:

[0034] The video detection method and adult products provided above, the system obtains the target video image, and then inputs the target video image into the convolutional neural network model to perform deep feature extraction to obtain the feature image of each frame.

[0035] Subsequently, the extracted feature images are classified through the CNN model to distinguish different types of scenes or actions.

[0036] After the classification is completed, the system scores the classified feature images and generates auxiliary control signals based on the scoring results. The auxiliary control signals are sent to external adult product equipment to control its movement mode, thereby achieving synchronization or interaction between video content and device actions.

[0037] By synchronizing the video content with the movements of adult products, the user's interactive experience is enhanced, especially when watching adult videos, users can feel a more realistic interactive effect.

[0038] Using convolutional neural networks to extract and classify image features can automatically generate control signals based on video content, making the use of adult products more intelligent and personalized.

[0039] By analyzing video content and controlling corresponding device actions, users can be provided with a more immersive experience and overall satisfaction can be improved. BRIEF DESCRIPTION OF THE DRAWINGS

[0040] Figure 1 is a flow chart of an embodiment of a video detection method of the present invention;

[0041] Figure 2 is a flow chart of another embodiment of the video detection method of the present invention;

[0042] Figure 3 A flowchart of another embodiment of the video detection method of the present invention;

[0043] Figure 4 It is a structural block diagram of the adult product of the present invention.

[0044] Reference Numbers

[0045] 10. Adult products; 20. Receiving module; 30. Sports module. DETAILED DESCRIPTION

[0046] The embodiments of the present application will be described below in conjunction with the drawings in the embodiments of the present application. Obviously, the described embodiments are only part of the embodiments of the present application, rather than all the embodiments.

[0047] It should be noted that when a component is considered to be "connected" to another component, it may be directly connected to the other component or there may be a central component at the same time. When a component is considered to be "located on" another component, it may be directly located on the other component or there may be a central component at the same time. The terms "top", "bottom", "upper", "lower", "left", "right", "front", "back", and similar expressions used herein are for illustrative purposes only.

[0048] Some embodiments of the present application are described in detail below in conjunction with the accompanying drawings. In the absence of conflict, the following embodiments and features in the embodiments can be combined with each other.

[0049] An embodiment of the present application provides a video detection method, the method comprising:

[0050] S101, acquiring a target video image;

[0051] S102, inputting the target video image into a convolutional neural network model, wherein the convolutional neural network model performs feature extraction on each frame image in the target video image to obtain feature images of all frame images;

[0052] S103, the convolutional neural network model detects the feature images of all the frame images to obtain classified feature images;

[0053] S104, scoring the classified feature images to obtain scored feature images, obtaining auxiliary control signals according to the scored feature images, and sending the auxiliary control signals to the external adult product to control the movement of the external adult product.

[0054] Please refer to Figure 1-Figure 3 In this embodiment, the system acquires a target video image, and then inputs the target video image into a convolutional neural network model to perform deep feature extraction to obtain a feature image of each frame.

[0055] Subsequently, the extracted feature images are classified through the CNN model (convolutional neural network model) to distinguish different types of scenes or actions.

[0056] After the classification is completed, the system scores the classified feature images and generates auxiliary control signals based on the scoring results. The auxiliary control signals are sent to external adult product equipment to control its movement mode, thereby achieving synchronization or interaction between video content and device actions.

[0057] By synchronizing the video content with the movements of adult products, the user's interactive experience is enhanced, especially when watching adult videos, users can feel a more realistic interactive effect.

[0058] Using convolutional neural networks to extract and classify image features can automatically generate control signals based on video content, making the use of adult products more intelligent and personalized.

[0059] By analyzing video content and controlling corresponding device actions, users can be provided with a more immersive experience and overall satisfaction can be improved.

[0060] It should be noted that the target video image is a video that requires interaction.

[0061] Obtain target video images: Obtain video image data from a data source. These video images are usually the contents of adult films.

[0062] The target video image is input into a convolutional neural network model, and the video image data is input into a pre-trained convolutional neural network (CNN) model.

[0063] The convolutional neural network model performs feature extraction on each frame image in the target video image to obtain feature images of all frame images. Feature extraction is to extract key information that can represent the image content from the image.

[0064] It should be noted that the auxiliary control signal is generated according to the time and score of the scored feature image. For example, if the score of the scored feature image at the 10th second is 70, an auxiliary control signal is output, and the auxiliary control signal controls the movement speed of the adult product to be adjusted to 70%, and the movement time is adjusted to the 10th second to synchronize with the video.

[0065] In at least one embodiment of the present application, the convolutional neural network model detects the feature images of all the frame images to obtain the classified feature images, including:

[0066] S201, calculating the confidence score of each feature image in the feature images of all the frame images to obtain the confidence scores corresponding to all the frame images;

[0067] S202: Classify the feature images of all the frame images according to the confidence scores corresponding to all the frame images to obtain classified feature images.

[0068] In at least one embodiment of the present application, the step of scoring the classified feature image to obtain a scored feature image, and obtaining an auxiliary control signal according to the scored feature image includes:

[0069] S203, comparing the confidence scores corresponding to the classified feature images with the confidence thresholds one by one;

[0070] S204: If the confidence score corresponding to the classified feature image is greater than the confidence threshold, it is determined to be an auxiliary image.

[0071] In at least one embodiment of the present application, the step of scoring the classified feature image to obtain a scored feature image, and obtaining an auxiliary control signal according to the scored feature image further includes:

[0072] S205 , selecting a maximum confidence value from the confidence scores corresponding to the classified feature images, recording the original image corresponding to the maximum confidence value as an enhanced auxiliary image, and generating an auxiliary control signal according to the enhanced auxiliary image.

[0073] Please refer to Figure 1-Figure 3In this embodiment, the system first extracts features from each frame image in the video, and then calculates the confidence score of each frame feature image.

[0074] Based on the confidence score of each frame, the system classifies the image. The classification criterion is whether the image meets the preset category characteristics.

[0075] The system compares the confidence scores of the classified images with a preset threshold, and images exceeding the threshold are marked as “auxiliary images”.

[0076] Among all the auxiliary images, the system selects the image with the highest confidence score and uses its confidence score as the final auxiliary control signal, which is used to control the movement of the external adult product.

[0077] By carefully calculating and classifying the confidence score of each frame of the image, the system can generate more accurate control signals, so that the actions of adult products are more in line with the interactive effects that users expect.

[0078] The system can generate control signals in real time according to changes in video content, so that the movement mode of adult products can be dynamically adjusted, enhancing the synchronization and interactivity with video content.

[0079] By selecting the most representative image (with the highest confidence), the system ensures that the generated control signal can most effectively guide the device action, improving the user's overall experience.

[0080] Auxiliary control signals include the time when the picture appears, the strength of the movement mode, the duration of the movement process, etc.

[0081] The confidence threshold is a manually set threshold.

[0082] The CNN model extracts features from each frame of the target video image to extract specific objects and scenes.

[0083] The CNN model uses two prediction heads during training, one using one-to-many assignment and the other using one-to-one assignment.

[0084] During CNN model training, rich supervisory signals of one-to-many assignment are utilized, while during inference, the prediction results of one-to-one assignment are used, thereby achieving efficient inference without NMS to improve detection accuracy.

[0085] In at least one embodiment of the present application, the method further includes:

[0086] S301, calculating the accuracy rate according to the confidence score corresponding to the classified feature image and the confidence scores corresponding to all frame images;

[0087] S302: If the accuracy rate is greater than the accuracy rate threshold, an auxiliary control signal is output.

[0088] In at least one embodiment of the present application, the method further includes:

[0089] S303, if the accuracy rate is less than the accuracy rate threshold, performing part feature extraction on the classified feature image to obtain a feature image after part extraction;

[0090] S304: Score the feature image after the part extraction to obtain a part classification video-assisted score.

[0091] In at least one embodiment of the present application, the method further includes:

[0092] S305 , calculating a recall rate according to the part classification video-assisted score, taking the recall rate as a final score, and outputting the final score.

[0093] Please refer to Figure 1-Figure 3 In this embodiment, the system first calculates the accuracy based on the confidence scores of the classified feature images and all frame images. If the accuracy exceeds the accuracy threshold, it indicates that the classification result has sufficient credibility, and the system will directly output the auxiliary control signal.

[0094] If the accuracy is lower than the preset threshold, the system will not immediately output a control signal, but will perform a deeper analysis of the classified feature image, namely, part feature extraction. By scoring specific parts, the system generates a part classification video-assisted score.

[0095] Based on the part classification video-assisted scoring, the system further calculates the recall rate. As a comprehensive evaluation indicator, the recall rate is used as the final score. If the final score meets the standard, the system will output the final control signal.

[0096] It should be noted that the accuracy threshold is a manually set threshold.

[0097] Part feature extraction is to extract features of exposed parts of the human body in the classified feature images.

[0098] By introducing the calculation steps of accuracy and recall, the system can more accurately evaluate the reliability of the classification results, ensuring that control signals are generated only when the classification is accurate enough to avoid false actions.

[0099] When the accuracy of the initial classification does not meet the standard, the system performs further analysis through part feature extraction and detailed scoring, so that the system can still maintain a high recognition accuracy when processing complex or blurred images.

[0100] By generating more accurate control signals, the system can better synchronize with the video content and provide a richer and more precise interactive experience. This precise control and feedback greatly improves the overall user satisfaction.

[0101] The CNN model calculates the quotient of the confidence score corresponding to the classified feature image and the confidence score corresponding to all frame images to obtain the accuracy, which is calculated according to the following formula:

[0102] Precision=(True Positives) / (True Positives+False Positives)

[0103] Among them, Precision is the accuracy; True Positives is the confidence score corresponding to the feature image after classification of each frame.

[0104] True Positives+False Positives is the confidence score corresponding to all frame images.

[0105] In at least one embodiment of the present application, if the accuracy rate is less than the accuracy rate threshold, performing part feature extraction on the classified feature image, and obtaining the feature image after part extraction includes:

[0106] Obtaining a feature image corresponding to a maximum confidence score in the classified feature image, and performing part extraction to obtain a feature image after part extraction;

[0107] Acquire a part feature image of a previous frame and a part feature image of a subsequent frame of the feature image after the part extraction;

[0108] The confidence scores are performed on the part feature image of the previous frame, the part feature image of the next frame, and the feature image after the part extraction to obtain the confidence scores after the frame image verification.

[0109] In at least one embodiment of the present application, the step of calculating the recall rate according to the part classification video-assisted score, taking the recall rate as the final score, and outputting the final score comprises:

[0110] The recall rate is calculated based on the confidence score of the frame image after review and the confidence scores corresponding to all frame images. If the recall rate is greater than the recall threshold, a final control signal is generated based on the feature image corresponding to the maximum confidence score in the classified feature image, the part feature image of the previous frame, and the part feature image of the next frame, and the final control signal is sent to the external adult product.

[0111] Please refer to Figure 1-Figure 3In this embodiment, the system selects the image with the highest confidence score from the classified feature images, and then extracts features of its key parts to obtain the feature image after part extraction.

[0112] The part feature images of the previous frame and the next frame adjacent to the current frame are obtained. Subsequently, the system performs confidence scores on the part features of these three frames of images and reviews these scores.

[0113] By comparing the verified confidence score with the confidence scores of all frame images, the system calculates the recall rate, which is used to verify the comprehensiveness and reliability of the classification results.

[0114] If the recall rate is higher than the set threshold, the system generates a final control signal and sends it to the external adult product. The control signal is generated based on the image with the highest confidence score and the feature images of its adjacent frames.

[0115] By extracting part features from the image with the highest confidence and reviewing them in combination with the features of adjacent frames, the system can capture key features more accurately and reduce the possibility of misjudgment.

[0116] By analyzing the feature images of adjacent frames, the system can capture dynamic changes, improve the continuity of the overall analysis, and help generate more accurate final control signals.

[0117] Since the generation of the final control signal is based on the review of multiple frames of images and accurate recall rate calculation, the possibility of equipment malfunction due to misjudgment is greatly reduced, and the safety and reliability of equipment use are improved.

[0118] The CNN model calculates the recall rate based on the confidence score of the frame image after review and the confidence score corresponding to all frame images according to the following formula:

[0119] Recall=(True Positives) / (True Positives+False Negatives);

[0120] Among them, Recall is the recall rate; True Positives is the confidence score after frame image review; True Positives+False Negatives is the confidence score corresponding to all frame images.

[0121] An adult product 10, applied in any one of the above-mentioned video detection methods, the adult product comprising:

[0122] A receiving module 20, used for receiving an auxiliary control signal or a final control signal;

[0123] The motion module 30 executes the auxiliary control signal or the final control signal to adjust the motion mode.

[0124] Please refer to Figure 4 In this embodiment, when the user watches a video, the video detection method starts to work and generates corresponding auxiliary control signals or final control signals. These signals are then received by the receiving module of the adult product 10.

[0125] The receiving module 20 transmits the received control signal to the motion module 30. The motion module 30 makes corresponding motion adjustments according to the signal content (such as intensity, frequency, time, etc.).

[0126] The motion module 30 executes the instructions in the signal by adjusting the motion pattern. For example, if the signal indicates that the device needs to increase the vibration frequency, the motion module 30 will adjust its motion to meet this requirement.

[0127] When the user is using the adult product 10, the motion module 30 of the device is synchronized or associated with the video content, providing an interactive experience with the video content and enhancing the user's sense of immersion and satisfaction.

[0128] By synchronizing the video content with the movement patterns of adult products, the device can provide users with a more realistic interactive experience. Users can not only watch the video, but also feel the situation in the video through the movement of the device, which enhances the pleasure of use.

[0129] The cooperation between the receiving module 20 and the motion module 30 enables the device to intelligently adjust its actions according to the signals generated by the video detection method.

[0130] The receiving module 20 receives the signal in time, and the motion module 30 executes quickly, so that the device can quickly respond to changes in video content and provide instant feedback. This high responsiveness and precision control greatly improves the user experience.

[0131] The above is only an implementation method of the present application. It should be pointed out that a person skilled in the art can make improvements without departing from the inventive concept of the present application, but these improvements are within the scope of protection of the present application.

Claims

1. A video detection method, characterized in that: The method comprises: Acquire target video image; Inputting the target video image into a convolutional neural network model, the convolutional neural network model performs feature extraction on each frame image in the target video image to obtain feature images of all frame images; The convolutional neural network model detects the feature images of all the frame images to obtain classified feature images; The classified feature images are scored to obtain scored feature images, an auxiliary control signal is obtained based on the scored feature images, and the auxiliary control signal is sent to an external adult product to control the movement of the external adult product.

2. The video detection method according to claim 1, characterized in that: The convolutional neural network model detects the feature images of all the frame images to obtain the classified feature images, which includes: Calculating the confidence score of each feature image in the feature images of all the frame images to obtain the confidence scores corresponding to all the frame images; The feature images of all the frame images are classified according to the confidence scores corresponding to all the frame images to obtain classified feature images.

3. The video detection method according to claim 1, characterized in that: The step of scoring the classified feature images to obtain scored feature images, and obtaining auxiliary control signals according to the scored feature images comprises: The confidence score corresponding to the classified feature image is compared with the confidence threshold one by one. If the confidence score corresponding to the classified feature image is greater than the confidence threshold, it is judged as an auxiliary image.

4. The video detection method according to claim 3, characterized in that: The step of scoring the classified feature images to obtain scored feature images, and obtaining auxiliary control signals according to the scored feature images further comprises: The maximum confidence value is screened out from the confidence scores corresponding to the classified feature images, and the original image corresponding to the maximum confidence value is recorded as an enhanced auxiliary image, and an auxiliary control signal is generated according to the enhanced auxiliary image.

5. The video detection method according to claim 1, characterized in that: The method further comprises: The accuracy is calculated based on the confidence score corresponding to the classified feature image and the confidence scores corresponding to all frame images. If the accuracy is greater than the accuracy threshold, an auxiliary control signal is output.

6. The video detection method according to claim 5, characterized in that: The method further comprises: If the accuracy rate is less than the accuracy rate threshold, performing part feature extraction on the classified feature image to obtain a feature image after part extraction; The feature image after the part extraction is scored to obtain a part classification video-assisted score.

7. The video detection method according to claim 6, characterized in that: The method further comprises: The recall rate is calculated according to the part classification video-assisted score, the recall rate is used as the final score, and the final score is output.

8. The video detection method according to claim 6, characterized in that: If the accuracy rate is less than the accuracy rate threshold, the step of performing part feature extraction on the classified feature image to obtain the feature image after part extraction includes: Obtaining a feature image corresponding to a maximum confidence score in the classified feature image, and performing part extraction to obtain a feature image after part extraction; Acquire a part feature image of a previous frame and a part feature image of a subsequent frame of the feature image after the part extraction; The confidence scores are performed on the part feature image of the previous frame, the part feature image of the next frame, and the feature image after the part extraction to obtain the confidence scores after the frame image verification.

9. The video detection method according to claim 8, characterized in that: The steps of calculating the recall rate according to the part classification video-assisted score, taking the recall rate as the final score, and outputting the final score include: The recall rate is calculated based on the confidence score of the frame image after review and the confidence scores corresponding to all frame images. If the recall rate is greater than the recall threshold, a final control signal is generated based on the feature image corresponding to the maximum confidence score in the classified feature image, the part feature image of the previous frame, and the part feature image of the next frame, and the final control signal is sent to the external adult product.

10. An adult product, using the video detection method according to any one of claims 1 to 9, characterized in that: The adult products include: A receiving module, used for receiving an auxiliary control signal or a final control signal; The motion module executes the auxiliary control signal or the final control signal to adjust the motion mode.

Images