Research

- Create, collect and augment Telugu and Tamil printed & handwritten data

- Text detection and preprocessing

- Word level OCR and handwriting recognition for Telugu and Tamil using Deep Learning models

- Accurate enough to fairly work on unclean real samples

- Overcome the problem of insufficient amount of usable forms for training

- Attention mechanism to search and extract information from key fields

- Non Template-specific processing

- Reliable to deploy for practical uses

- Handwriting recognition for English words

- Text detection from challenging and varied manuscripts

- Work accurately on both standard samples (like IAM dataset) as well as difficult examples (like CVPR dataset)

- Handwriting generation model using loops and oscillators

- Buffer mechanism to maintain constant information about the character until completion

- Smooth transition from one character to other

- Create applications to improve learning process among dyslexic children

- Understand causes and develop solutions using similarities between the learning process of artificial neural networks and children