Neural Networks for Classifying Indian Folk Motifs

In this paper, I have outlined a hybrid neural network model of automated classification of Indian folk motifs within the different regional traditions, such as Madhubani, Warli, Kalamkari, and Pattachitra. Introduced was a culturally authentic curated collection of 5000 high-resolution images created by ethical digitization and by human experts. The suggested model uses a Convolutional Block Attention Module (CBAM) in addition to a ResNet-50 backbone to promote the discrimination of features in space and channels. The experimental findings show better performance in comparison with the baseline CNN and VGG architectures, with the overall accuracy of 94.6, macro F1-score of 94.0 and Cohen 6. Grad-CAM visualizations show that the activations of the model are consistent with motif-specific areas of art, which verify the cultural interpretability. The framework helps to explain explainable cultural AI because it associates computational properties with heritage aesthetics, allowing them to be used in digital museums, education platforms and art restoration systems. The research provides a methodological basis of combining deep learning with the preservation of cultural heritage with a focus on transparency, reproducibility, and cross-disciplinary applicability

Artificial neural network · Convolutional neural network · Cultural heritage · Digitization · Aesthetic Perception and Analysis · Generative Adversarial Networks and Image Synthesis · Image Processing and 3D Reconstruction

• Array programming with NumPy

  Open Access•Charles R Harris, K Jarrod Millman et al.•Nature•2020

  Add to list