Handloomed fabrics recognition with deep learning.

Area of Science:

• Textile Science
• Artificial Intelligence
• Computer Vision

Background:

• India's handloom industry is vital for cultural heritage and artisan livelihoods.
• Counterfeit powerloom products threaten the authenticity and market of genuine handloom items.
• Distinguishing authentic handwoven textiles, like the

Purpose of the Study:

• To develop an AI-powered tool for automated detection of authentic handloom products.
• To differentiate genuine handwoven

Main Methods:

• Trained six pre-existing deep learning architectures (VGG16, VGG19, ResNet50, InceptionV3, InceptionResNetV2, DenseNet201) on 17,484 handloom and powerloom towel images.
• Developed and trained a novel deep learning model for the same image dataset.
• Evaluated model performance based on validation accuracy, loss, and computational efficiency.

Main Results:

• The novel deep learning model demonstrated superior performance over pre-trained models.
• The proposed model achieved higher validation accuracy and lower validation loss.
• Pre-trained models struggled with generalization to unseen data and posed computational challenges.

Conclusions:

• A novel AI model offers an efficient and accurate solution for authenticating handloom products.
• The developed methodology shows potential for scalability and broader applications in textile authentication.
• This computer-assisted approach is a groundbreaking method for preserving handloom heritage against imitations.