---
license: apache-2.0
task_categories:
- image-to-text
language:
- gu
tags:
- handwritten-text-recognition
- htr
- gujarati
- ocr
- iiit-indic-hw-words
pretty_name: Gujarati Handwritten Dataset (IIIT-INDIC-HW-WORDS)
size_categories:
- 10K-100K
dataset_info:
  features:
  - name: file_name
    dtype: string
  - name: text
    dtype: string
  - name: image
    dtype: image
  splits:
  - name: train
    num_bytes: 2867138581.533
    num_examples: 82563
  - name: val
    num_bytes: 632023288.427
    num_examples: 17643
  - name: test
    num_bytes: 597646751.57
    num_examples: 16490
  download_size: 4016768714
  dataset_size: 4096808621.53
configs:
- config_name: default
  data_files:
  - split: train
    path: data/train-*
  - split: val
    path: data/val-*
  - split: test
    path: data/test-*
---

# Gujarati Handwritten Word Dataset

This dataset is a subset of the **IIIT-INDIC-HW-WORDS** collection, specifically focused on the **Gujarati** language. It is designed for training and evaluating **Handwritten Text Recognition (HTR)** models.

## Dataset Summary

The original [IIIT-INDIC-HW-WORDS](https://cvit.iiit.ac.in/usodi/istr.php) is a large-scale benchmark for Indic scripts. This Gujarati subset contains word-level images manually written by multiple annotators to capture natural variations in handwriting styles.

### Key Statistics

| Feature | Count |
| :--- | :--- |
| **Total Word Images** | 82,563 |
| **Train Set** | 48,430 |
| **Validation Set** | 17,643 |
| **Test Set** | 16,490 |

---

## Dataset Structure & Extraction

The dataset consists of image folders and corresponding annotation text files. Follow these instructions to map images to their transcriptions:

### 1. Files Overview
* **Images:** Located in the `train/`, `val/`, and `test/` folders.
* **Labels:** Provided in `train.txt`, `val.txt`, and `test.txt`.
* **Lexicon:** `vocab.txt` contains the full list of Unicode strings used in the dataset.

### 2. Label Format
Each row in the label files (`train.txt`, `val.txt`, `test.txt`) follows this format:
` <FileName>, <VocabId> `

### 3. Mapping Logic
The `<VocabId>` is a **0-indexed** pointer to the line number in `vocab.txt`.
* **Step 1:** Locate the `VocabId` for an image in the split text file.
* **Step 2:** Go to that specific line number in `vocab.txt` to extract the Unicode Gujarati string.

---

## Citation

If you use this dataset in your research, please cite the following paper:

```bibtex
@inproceedings{gongidi2021iiit,
  title={IIIT-Indic-HW-Words: A Dataset for Indic Handwritten Text Recognition},
  author={Gongidi, Santhoshini and Jawahar, CV},
  booktitle={Proceedings of the 16th International Conference on Document Analysis and Recognition (ICDAR)},
  pages={444--459},
  year={2021},
  organization={Springer}
}
```