Skip to content

Latest commit

 

History

History
16 lines (11 loc) · 525 Bytes

README.md

File metadata and controls

16 lines (11 loc) · 525 Bytes

DeepNumta

A Comprehensive Study on Deep Learning Approaches for Bengali Handwritten Digit Recognition

Approaches

  • A custom built 10 layer Neural Network
    • Code: DeepNumtaCNN.ipynb
    • Leaderboard position: 4th
    • Gets 99.4% accuracy on validation set and 96.68% test accuracy on Kaggle test set of NumtaDB
  • ResNet34
  • ResNeXt

Dataset

NumtaDB - https://www.kaggle.com/BengaliAI/numta