Graph Neural Network-Based Computer Vision Framework for Real-Time Object Detection and Scene Understanding

Abstract

Graph Neural Networks (GNNs) have emerged as a powerful paradigm for modeling relational and structured data, offering significant advantages in computer vision tasks that require contextual reasoning and spatial understanding. This research proposes a graph neural network-based computer vision framework for real-time object detection and scene understanding. The framework integrates convolutional feature extraction with graph-based relational reasoning to enhance object representation and contextual awareness. The proposed approach constructs a graph representation of visual scenes, where nodes correspond to detected objects or regions and edges encode spatial and semantic relationships. By leveraging graph convolution operations, the model captures interactions between objects, enabling improved recognition and interpretation of complex scenes. Experimental evaluation demonstrates that the GNN-based framework achieves higher detection accuracy and improved scene understanding compared to traditional convolutional models. Furthermore, the framework incorporates optimization techniques such as dynamic graph construction and attention-based message passing to ensure real-time performance. Results indicate that the proposed model effectively balances computational efficiency with high-level reasoning capabilities, making it suitable for applications in autonomous driving, surveillance, and intelligent robotics. This study contributes a scalable and context-aware vision framework for next-generation computer vision systems.