Image matching using hybrid graph neural networks

Image matching is a critical task in computer vision applications such as object recognition, 3D reconstruction, and autonomous navigation. The emergence of Graph Neural Networks (GNNs) has opened new avenues to improve the accuracy and efficiency of image matching algorithms. This thesis investigates the application of GNNs to image matching, aiming to evaluate and enhance these methods using different datasets. The research initially focuses on applying GNN-based algorithms to well-known datasets like PASCAL VOC and WILLOW-object class. These datasets, known for their variety of images and annotations, serve as a foundation for testing and developing image matching algorithmsA hybrid system proposal is presented to obtain better and faster results with the above-mentioned data sets. Additionally, this thesis integrates findings from recent significant papers on GNNs and image matching. The comparative analysis across different datasets is conducted to comprehensively understand the strengths and weaknesses of GNN-based image matching algorithms. The outcome of this research is expected to significantly contribute to the field of computer vision by demonstrating the versatility and robustness of GNNs in image matching tasks. This study not only advances the theoretical understanding of GNNs in image matching but also offers practical insights into their application in various contexts.