A Novel Approach for Monument Identification Using a Modified ResNet-101 Encoder-Decoder Architecture

Identification and categorization of monuments play an integral part in the preservation and documentation of cultural heritage, with research greatly improved thanks to advances in computer vision and deep learning methods. This work presents an innovative methodology for the identification of monuments using an altered ResNet-101 encoder-decoder architecture. This research employs ResNet-101’s transfer learning capabilities to extract relevant information from images of monuments. Subsequently, these extracted characteristics are fed into a decoder model, which produces a segmentation mask to accurately recognize monuments featured in any given image. Our experimental results illustrate that our proposed methodology achieves 95.8% accuracy compared to existing methods when applied to a standardized dataset of monument images.