8 results about "Internet video" patented technology

This invention discloses a subtitle extraction method based on video text merging, filtering, and classification. The method includes extracting frames from the video, using optical character recognition (OCR) to identify text within all video frames, obtaining a set of text boxes within the video; merging and filtering the text boxes based on text content, text box coordinates, and text appearance time; and using a machine learning-based subtitle classification model to predict whether each filtered text box is a subtitle, saving the text identified as subtitles and their location information as the subtitle information for the video. This method, by merging and filtering text boxes, can initially filter out most text that does not belong to the subtitle type. By constructing a machine learning subtitle classification model, the text box type can be further determined. This method does not require specifying the subtitle area and can address the problem of variable subtitle positions in existing internet videos.

The invention discloses a method and system for assisting a green driver in improving driving skills, and relates to the technical field of automobile intelligent assistance, and the method comprises the steps: S1, obtaining driving information: obtaining driving-related multi-source information provided by a vehicle system through a vehicle machine central control system, and enabling the multi-source information to be used for representing a current driving environment and a vehicle state; s2, scene element analysis: performing analysis based on the multi-source information by the vehicle end novice driver auxiliary application; s3, element cloud uploading; s4, prompt content is generated; s5, prompting content presentation; and S6, prompting strategy adjustment. In a real-time driving scene, a cloud skill knowledge base is combined with Internet video resources for matching, skill prompt content matched with the current scene is generated based on core elements, and characters, graphical representation or video demonstration is pushed to a vehicle end, so that assistance is more intuitive and easy to understand during driving, and the practicability and learning effect of the prompt content are improved.

The invention discloses an Internet live video sensitive information monitoring method based on artificial intelligence. The method comprises the following steps: generating a preprocessed frame image; predicting the dominant gradient direction of each pixel in the preprocessed frame image by using a direction prediction sub-network, and outputting a manipulation coefficient matrix; synchronously inputting the control coefficient matrix and the preprocessed frame image into a dynamic controllable differential convolution base module, and generating a direction self-adaptive differential feature map; generating a multi-scale fusion feature map; projecting the multi-scale fusion feature map to a Riemannian manifold space through nonlinear mapping, outputting a Riemannian manifold mapping feature map, and generating a geometric deviation mapping map; generating a topological anomaly saliency map according to the geometric deviation mapping map, and obtaining a sparse candidate region feature map; and when the violation marking result indicates that the violation target exists, generating a violation frame capturing result, and performing evidence chain packaging and outputting on the violation frame capturing result. According to the method, the effective frame capturing capability in a high-noise live broadcast environment is greatly improved.

Embodiments of the present application relate to the technical field of Internet video, and particularly relate to a video bullet screen reading method, an electronic device and a storage medium. The bullet screen reading method provided by the embodiments of the present application can acquire bullet screen attributes of each bullet screen appearing within a preset time in a video playing process, wherein the bullet screen attributes include the number of words, semantic content and / or repetitiveness of the bullet screen; the electronic device inputs the bullet screen attributes of each bullet screen to a preset reading matching model, so that the reading matching model determines a bullet screen that needs to be read among the bullet screens according to the bullet screen attributes of each bullet screen; and reads the bullet screen that needs to be read through an audio output unit. In the embodiments, the electronic device can selectively read the bullet screen according to the bullet screen attributes, without reading all the bullet screens one by one, thereby improving the efficiency of bullet screen reading and enhancing user experience.

The invention provides an internet video intelligent directional delivery method and system based on a user portrait, and relates to the technical field of internet information, and the method comprises the following steps: obtaining identity information and behavior information of a user, constructing the user portrait according to the identity information and the behavior information in combination with a heterogeneous graph propagation analysis method, and sending the user portrait to a server; the method comprises the following steps: acquiring a user portrait, performing anti-collision detection optimization on the user portrait by using a bounding box simulation method to obtain a final user portrait, acquiring an internet video and video metadata, performing multi-modal feature extraction on the internet video, and forming a video portrait in combination with a metadata guide method; and calculating the matching degree of the user portrait and the video portrait through a deep learning matching model, and performing internet video directional delivery to the user according to the matching degree. According to the invention, the integrating degree of video delivery and user requirements can be improved, the delivery accuracy is enhanced, and the user experience and the video delivery effect are improved.

The application provides a world model-based unmanned logistics vehicle intelligent driving model training method and system. The method generates instructions by obtaining historical video data and long-tail scene data, merges the historical video data with a preset amount of Internet video data as sample data, trains a preset architecture generative world model through the sample data, determines a structured generation instruction, controls the first generative world model to generate multi-modal long-tail scene data as an input condition, determines first long-tail scene state data, and trains and evaluates the first generative world model through the first long-tail scene state data to obtain a teacher model, determines a first student model, and deploys the first student model on a vehicle end to make decision control during unmanned logistics vehicle driving. The application effectively solves the problems of difficult long-tail data acquisition and high cost, and improves the accuracy of long-tail scene recognition of the student model.

This disclosure provides for automated techniques to measure full reference-based QoE or VQA-as-a-Service (VQAaaS) for an Internet video stream. Generally, the approach herein involves pre-calculating VMAF scores for given media and then correlating those scores with VMAF scores computed from actual playback segments for the given media. By leveraging the pre-calculated VMAF scores and correlating them with playback data, the system provides for enhanced and accurate video quality analysis (VQA) to enable optimization of viewer Quality of Experience (QoE).