In the field of pattern recognition, HCR is one of the most intricate and tricky area. Plenty of works were proposed for foreign languages but a few works exists for Indian languages due to the complex shape and varying writing styles of individuals. This paper introduces a review of offline and recognition of different natural languages. In this paper, we describe the procedure of developing dataset for offline handwritten Marathi character & numerals recognition. There is no standard dataset available for handwritten characters. Researcher has to develop own character dataset. This paper will provide a way for researcher to develop a dataset for offline handwritten Marathi character recognition. This paper describes use of Prewitt Operator and Support Vector Machine (SVM) classifier for recognition. This paper describes basics of dataset, challenges associated with it and its processing.