GNN-Transformer Fusion for Enhanced Missing Child Identification

In India, a countless number of children are reported missing every year. Among the missing child cases a large percentage of children remain untraced. This project presents a deep learning-based system for identifying missing children by comparing uploaded images with a repository of known missing child photos. The system employs Convolutional Neural Networks (CNNs) for face recognition, using the VGG-Face pretrained model for face descriptor extraction and an SVM classifier for child recognition tasks. While the existing approach demonstrates robustness to factors like noise, illumination, contrast, occlusion, image pose, and child age, it faces limitations in handling complex relationships, unseen scenarios, and scalability. The model also lacks the ability to capture temporal and relational data between different images or missing reports. To overcome these drawbacks, a hybrid Graph Neural Network (GNN) and Transformer model can be employed, where GNNs capture complex relational patterns and Transformers handle large-scale data and long-range dependencies. This combination enhances the model's ability to generalize to unseen conditions, improves scalability for larger datasets, and incorporates temporal and relational context, leading to more accurate and robust missing child identification.