Notice: This page will be used for Kernels, Starter Codes, and Discussions. For competition rules, evaluation metric description, new team registration, data download and result submission click  [here](https://www.kaggle.com/c/numta). Context

Currently, natural language processing (NLP) research is developing rapidly due to the rise of artificial intelligence (AI). One of the key topics of NLP is optical character recognition (OCR). To build an OCR in Bengali language, digit classification provides a convenient starting point.  We have accumulated a large dataset (85,000+) of Bengali digits (NumtaDB) which can be used by researchers for benchmarking their algorithm.

Content The dataset is a combination of six datasets that were gathered from different sources and at different times. However, each of them was checked rigorously under the same evaluation criterion so that all digits were at least legible to one human being without any prior knowledge. Descriptions of these datasets including collection methodology, image segmentation and extraction and image formats of these datasets are described in [https://bengali.ai/datasets](https://bengali.ai/datasets). The sources are labeled from 'a' to 'f'. The training and testing sets have separate subsets depending on the source of the data (training-a, testing-a, etc.). All the datasets have been partitioned into training and testing sets so that handwriting from the same subject/contributor is not present in both. Dataset-f had no corresponding metadata for contributors for which all of it was added to the testing set (testing-f). The metric for the competition is selected to be the Unweighted Average Accuracy (UAA). Starter codes for the competition are available at [https://github.com/BengaliAI](https://github.com/BengaliAI). *Two augmented datasets (augmented from test images of dataset 'a' and 'c') are appended to the testing set which consists of the following augmentations:* - Spatial Transformations: Rotation, Translation, Shear, Height/Width Shift, Channel Shift, Zoom. - Brightness, Contrast, Saturation, Hue shifts, Noise. - Occlusions. - Superimposition (to simulate the effect of text being visible from the other side of a page). Inspiration If you are a Bengali machine learning enthusiast this is a good starting point to get accustomed to computer vision algorithms using a dataset in your native language. Also, the augmented test images challenge the learners to learn about image augmentation and implement their own image augmentation pipeline. Acknowledgements Numerals in dataset 'e' are collected and curated version of [BanglaLekha-Isolated](https://www.sciencedirect.com/science/article/pii/S2352340917301117). We would like to thank the researcher for allowing us to integrate it into our database. When referencing this material please cite the below paper [NumtaDB - Assembled Bengali Handwritten Digits](https://arxiv.org/abs/1806.02452) **Bibtex** @article{alam2018numtadb,  title={NumtaDB-Assembled Bengali Handwritten Digits},  author={Alam, Samiul and Reasat, Tahsin and Doha, Rashed Mohammad and Humayun, Ahmed Imtiaz},  journal={arXiv preprint arXiv:1806.02452},  year={2018} }