Computers may dominate our lives but using pen for writing is still mightier than keyboard. Computer vision applications like handwriting recognition model is playing a vital role now-a-days. Handwritten text recognition (HTR) is the most efficient way to digitize handwritten documents. For every handwritten text recognition model, feature extraction is the foremost task. In CNN architecture, feature extraction is done automatically. Lot of attention has received in the past years but research has focused only on Latin, Urdu and English. For Tamil language very fewer studies were done. Instead of predicting by word, first the individual characters from the text should be segregated then the segregated character is given to the trained CNN model to predict handwritten Tamil characters.The dataset used here is an isolated handwritten Tamil character dataset which was developed by HP labs India. Hyper-parameter tuning is performed to tune the model and achieved good accuracy results.
Alan : Fen Bilimleri ve Matematik; Sağlık Bilimleri
Dergi Türü : Uluslararası
Benzer Makaleler | Yazar | # |
---|
Makale | Yazar | # |
---|