Eye-Position-Based 3D Effects for Video Streams

The patent replaces conventional 3D hardware systems with a software-based machine learning approach. A neural network model processes video frames to generate 3D effects, substituting physical 3D hardware with computational algorithms that analyze eye movement and adjust background layers accordingly, thereby reducing network bandwidth consumption while maintaining 3D effect quality

The patent creates a virtual copy of the 3D effect through image processing rather than using physical 3D hardware. By separating video frames into foreground and background layers and manipulating the background layer based on detected eye movement, the system generates a perceived 3D effect that mimics conventional 3D hardware output without requiring the actual hardware infrastructure

The patent substitutes complex 3D hardware processing systems with a streamlined machine learning pipeline. The neural network model, once trained, requires minimal processing resources to generate 3D effects by simply analyzing eye movement in video frames and adjusting background layers, eliminating the need for expensive and complex conventional 3D hardware infrastructure

The patent performs preliminary training of the neural network model offline before deployment. During the actual 3D effect generation, the pre-trained model requires minimal processing resources as it has already learned the complex patterns of eye movement and background manipulation. This preliminary action transfers the computational burden from runtime processing to initial model training, reducing ongoing processing resource requirements