Indian Sign Language Alphabet Recognition Using Transfer Learning with MobileNetV2

Abstract

Indian Sign Language (ISL) recognition plays a vital role in bridging the communication gap between the hearing-impaired community and the general population. This research presents an efficient deep learning-based approach for static ISL alphabet recognition using transfer learning with MobileNetV2. A dataset consisting of 26,000 images representing 26 alphabet classes (A–Z) was used. The proposed model leverages a pre-trained MobileNetV2 backbone for feature extraction, followed by custom classification layers. Experimental results demonstrate a high validation accuracy of 99% and test accuracy 99.89%, indicating the effectiveness of the approach for real-world ISL recognition tasks.