COMPREHENSIVE STUDY OF DEEP LEARNING BASED TELUGU OCR

The aim of the project is to understand offline One most popular and difficult pattern recognition subjects use optical character (OCR) read handwritten Telugu letters. This study suggests a three-stage OCR solution for documents that includes pre-processing, feature extraction, classification. For extraction boundary edge pixel points during preprocessing, we used median filtering on input characters as well normalisation skeletonization techniques. Each initially divided into three 3x3 grids stage, associated centroid each nine zones assessed. allows us recognise in various styles. Following that, drew projection angel's horizontal vertical symmetry character's closest pixel.