Transformer OCR

An end-to-end Transformer model for printed, handwritten, and scene-text recognition.

  • Duration: January 2024 – March 2024
  • Technologies: Python, OpenCV, PyTorch

View code on GitHub

Highlights

  • Created a 62-million-parameter, end-to-end Transformer-based OCR model using pretrained computer vision and NLP models.
  • Outperformed contemporary state-of-the-art models on printed, handwritten, and scene-text recognition tasks.