Classification of Terahertz Images via Caustic Lens and Transfer Learning

Our study proposes a novel approach to reduce data processing time and improve classification accuracy in terahertz imaging. By combining a caustic lens with transfer learning, this approach enables faster and more accurate analysis of data obtained from terahertz images. Notably, the method achieved a classification accuracy of $99.38 \%$ for five different letter shapes, demonstrating its effectiveness. This research presents a new pathway to make terahertz imaging systems faster, more accurate, and more cost-effective.