The deaf and mute community has mainly utilized sign language as a primary communication channel. Lack of general public awareness about sign language obstructs effective communication. The proposed project shall be a real- time implementation of Sign language interpretation using the MobileNet architecture, which is light in weight and fast, thus suitable for mobile and embedded applications. This can accordingly employ the separable convolutions depth wise for MobileNet so that feature extraction on the sign gesture can be performed efficiently to afford greater speed and accuracy on sign recognition. The suggested system will identify various signs through image input of hand gestures for completing recognition and classification of signs within pretty limited time. This work utilizes transfer learning on a pre-trained MobileNet model for the refinement on the static sign language gesture data set. The output is an efficient, low-latency system where hand signs can be classified in real time. Thus, it cradles an open yet practical way to be used for personal and educational needs. It returns strong experimental results showing recognition with very high accuracy.