Handwritten Devanagari Characters – Vowels, Consonants and Numerals for Devanagari CAPTCHA generation (44,000 Isolated images)
Description
Designing Handwritten Devanagari CAPTCHA, selected only 34 alphabet images and rejected 19 images from character set so that it does not create any confusion for humans for identification and retain its usability. The data set contains handwritten samples of Devanagari Vowels, Consonants and numerals (i.e., 4 Vowels, 30 Consonants and 10 numerals). In total the dataset contains 44 different Devanagari characters. The data is collected on a canvas created using python. More than hundred (100 +) Devanagari language users of all age groups are selected including left handers and right handers. Each user write 440 characters (44 characters * 10 times) on the canvas. In total 44,000 (Forty Four Thousand) character images were collected. Data is pre-processed and resized, and stored in a publicly accessible location. By removing the occluded images and scribbles, the final data set contains a total of 44,000 digitized images where 10,000 Devanagari Numerals (10 numerals * 1000 of each) , 4000 Vowels (4 Vowels * 1000 of each), 30,000 consonants ( 30 Consonants * 1000 of each). Each image is in.jpeg format and has a grayscale data type. Image Resolution is 65 by 65, 1.5 kb of storage were needed. Thus, whereas a total of 50 MB of date storage was needed, 162 MB were actually required. This data separated manually into respective folders. The data is also available in the form of CSV files with training and testing sets with labels attached. The data set also available in .zip format. Numeral's (10) - ० १ २ ३ ४ ५ ६ ७ ८ ९, Vowels(4) - अ इ उ ए , Consonants (30)- क ख ग घ च छ ज झ ट ड ढ ण त थ द ध न प ब भ म य र ल व श ष स ह ळ