Neural Network Object Detection for AR Video Stream Insertion
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Current object detection techniques in telecommunications require separate training models for individual features, leading to inefficient recognition of new objects and limited integration of graphical elements in video streams.
Innovation Solution
A neural network-based object detection system that generates a virtual bounding box around objects of interest, allowing for real-time detection and tracking, and inserts graphical elements relative to the object within the video stream, enabling augmented-reality experiences without analyzing facial or body features.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Ease of manufacture
If separate training models are used for individual features, then model training for specific features can be conducted independently, but recognition efficiency for new objects deteriorates due to the need to sequentially apply multiple separately trained models
Solution Approach 1:
The patent combines multiple separately trained feature models into a single integrated neural network model. This unified model simultaneously processes multiple features (such as shape, texture, color) and performs object recognition in one pass, eliminating the need to sequentially apply multiple independent models while maintaining the ability to train on diverse features.
Solution Approach 2:
The neural network model is designed with multi-functionality to handle various object features and recognition tasks within a single model architecture. It can process different feature types (geometric, textural, color) and adapt to recognize diverse object categories, replacing the need for multiple specialized separate models.
2Ease of operation
If user selection guides object detection, then detection can be targeted to specific objects, but detection flexibility deteriorates when modeling new objects that require combining multiple features
Solution Approach 1:
The neural network model employs dynamic feature weighting and adaptive feature selection mechanisms. The model can dynamically adjust which features to emphasize based on the input image characteristics and detection requirements, enabling it to adapt to new object types without requiring complete retraining or manual reconfiguration for each new object category.
Solution Approach 2:
The system performs preliminary feature extraction and processing through the neural network's multiple layers before final object classification. This preliminary action of extracting various feature types (geometric, textural, color) and combining them in intermediate processing stages enables the model to handle new object types that require multi-feature integration.
3Adaptability or versatility
If graphical elements are inserted in response to object detection, then augmented-reality features can be provided, but processing complexity increases on mobile devices with limited processing power
Solution Approach 1:
The neural network is organized into segmented functional modules that perform specific tasks in sequence: feature extraction layers, object detection layers, and graphical element insertion layers. This segmentation allows the mobile device to process only the necessary portions of the image stream and perform graphical insertions only when objects are detected, reducing overall processing complexity compared to analyzing every pixel for graphical enhancements.
Data Source
AI summary
Systems, devices, media, and methods are presented for object detection and inserting graphical elements into an image stream in response to detecting the object. The systems and methods detect an object of interest in received frames of a video stream. The systems and methods identify a bounding box for the object of interest and estimate a three-dimensional position of the object of interest based on a scale of the object of interest. The systems and methods generate one or more graphical elements having a size based on the scale of the object of interest and a position based on the three-dimensional position estimated for the object of interest. The one or more graphical elements are generated within the video stream to form a modified video stream. The systems and methods cause presentation of the modified video stream including the object of interest and the one or more graphical elements.