Kannada language is a famous ancient languages and it is a formal reading, writing and speaking language in Karnataka state. Karnataka has one of the heritage states in India. The ancient Kannada scripts provides us the related information about tradition and culture. We can enhance our historical knowledge by retrieving these kinds of information from carving the stone, palm leaves and also from the documents written on the paper. Retrieving and recognizing ancient Kannada scripts is a tedious task because of occluded characters, less quality, less contrast scripts and differentiation of various periods. To lessen the above difficulty, in this work, an automated recognition technique is proposed using advanced and recent deep learning classifier with SIFT feature extraction method. The CNN model is adopted for classification and recognition of the ancient Kannada script. For experimentation, the various periods such as Satavahanaru, Chulukyaru, and Vijayanagar ancient Kannada datasets are used. As a result of this work, an average recognition rate of 92.45% is achieved.