Bharat Scene Text Dataset (BSTD) is large and real Indian language scene-text dataset with coverage across 13 Indian languages and English. It consists of 6,582 scene-text images, with polygon bounding box annotations of 1,20,560 words and ground truth text annotations of 1,00,495 cropped words. This dataset is an effort towards scaling scene-text detection and recognition systems to work on Indian languages. The current version of this dataset can be used for studying scene text detection and cropped scene text word recognition.
- [8/8/24] First Public Release.
| Total images | #Total detection annotations | #Train | #Test |
|---|---|---|---|
| 6,582 | 1,20,560 | 94,128 | 26,432 |
| Language | #Total recognition annotations | #Train | #Test |
|---|---|---|---|
| Assamese | 3,966 | 2,623 | 1,343 |
| Bengali | 6,129 | 4,968 | 1,161 |
| English | 36,891 | 28,778 | 8,113 |
| Gujarati | 2,649 | 1,956 | 693 |
| Hindi | 18,889 | 14,855 | 4,034 |
| Kannada | 2,934 | 2,241 | 693 |
| Malayalam | 2,975 | 2,408 | 567 |
| Marathi | 4,977 | 3,932 | 1,045 |
| Meitei | 494 | 381 | 113 |
| Odia | 4,198 | 3,176 | 1,022 |
| Punjabi | 11,104 | 8,544 | 2,560 |
| Tamil | 2,548 | 2,041 | 507 |
| Telugu | 2,709 | 2,227 | 482 |
| Urdu | 32 | 29 | 3 |
| Total | 1,00,495 | 78,159 | 22,336 |
Download the detection.zip from the link (zip file ~17 GB).
Annotations are in BSTD_release_v1.json
Detection/
│
├── A/
│ ├── image_xx.jpg
│ ├── ...
│ └── image_xx.jpg
├── B/
├── C/
├── ...
├── M/
└── BSTD_release_v1.json
Words in the image are annotated in the polygon format. The annotation file is a json file with the following format:
"folderName_image_id": {
"annotations":
{
"polygon_0":
{
"coordinates":
[
[x1, y1],
[x2, y2],
...,
[xn, yn]
],
"text": "text in the current polygon",
"script_language" : "language of the word present in the polygon."
},
...,
"polygon_n":
{
"coordinates":
[
[x1, y1],
[x2, y2],
...,
[xn, yn]
],
"text": "text in the current polygon",
"script_language" : "language of the word present in the polygon."
}
},
"url": "url of the image",
"image_name": "path to the image",
"split" : "train/test split"
"folderName": "folder of the image"
}
Download the recognition.zip from the link (zip file ~774 MB).
Recognition/
│
├── train/
│ ├── assamese/
│ │ ├── X_image_name_xx_xx.jpg
│ ├── X_image_name_xx_xx.jpg
│ ├── X_image_name_xx_xx.jpg
│ ├── bengali/
│ │ ├── ...
│ ├── ...
│ └── urdu/
├── test/
│ ├── assamese/
│ ├── bengali/
│ ├── ...
│ └── urdu/
├── train.csv
└── test.csv
Files: recognition/train.csv and recognition/test.csv
Each file contains rows (each row has comma seperated values as follows)
path_to_the_cropped_word_image, recogntion_annotation, script_language
To convert the recognition data into lmbd files use utils/fetch_lmdb_format_data.py.
Usage
python fetch_lmdb_format_data.py --recognition_folder_path ~bstd/recognition/ --split train --language hindi --output_directory lmdb/hindi/train/real/hindi
To get more details on arguments
python fetch_lmdb_format_data.py --help
For the task of script identification, a dataset comprising images from three languages—English, Hindi, and a specific regional language—has been created. This setup allows for the evaluation of a model that classifies these three classes. The choice of languages reflects the common occurrence of these three languages in certain regions of India.
| Folder | #Language | #English | #Hindi | |||
|---|---|---|---|---|---|---|
| Train | Test | Train | Test | Train | Test | |
| assamese_ | 2623 | 1343 | 2623 | 1343 | 2623 | 1343 |
| bengali_ | 4968 | 1161 | 4968 | 1161 | 4968 | 1161 |
| gujarati_ | 1956 | 693 | 1956 | 693 | 1956 | 693 |
| kannada_ | 2241 | 693 | 2241 | 693 | 2241 | 693 |
| malayalam_ | 2408 | 567 | 2408 | 567 | 2408 | 567 |
| marathi_ | 3932 | 1045 | 3932 | 1045 | 3932 | 1045 |
| meitei_ | 381 | 113 | 381 | 113 | 381 | 113 |
| odia_ | 3176 | 1022 | 3176 | 1022 | 3176 | 1022 |
| tamil_ | 2041 | 507 | 2041 | 507 | 2041 | 507 |
| telugu_ | 2227 | 482 | 2227 | 482 | 2227 | 482 |
| urdu_ | 29 | 3 | 29 | 3 | 29 | 3 |
| hindi_ | - | - | 14855 | 4034 | 14855 | 4034 |
This dataset can be downloaded from this link. A script utils/make_dataset_for_scriptIdentification.py has also beed added to as to be able to directly create this dataset using the recognition dataset made available the upper section.
Each folder contains images from three language folders. For example, the folder bengali_ includes cropped word images of Hindi, English, and Bengali. For the test/bengali_folder, all image paths are listed in test.csv, which includes the correct language tag for each image. Similarly, all images in the train folder under each language-specific folder are listed in train.csv with their respective language tags.
Note: The hindi_ folder contains only cropped images of Hindi and English, with each image path listed in the CSV files.
Image subset used in (Vaidya et al., ICPR 2024) Preprint
BSTD images split used for Hindi to English scene text to scene text transaltion can be downloaded from the link
Images used for Hindi to English scene text to scene text transaltion can be downloaded directly from the link
To visualise detection annotations, run the following command:
python3 visualise.py <image_path> <path_to_BSTD_release_v1.json>
for e.g.
python3 visualise.py D/image_141.jpg path_to_BSTD_release_v1.json
Some examples are below:
- All the images are collected from Wikimedia commons (under Creative Commons Licence, cc-by-sa-4.0).
- Further detection and recognition annotations are manually annotated.
This work was partly supported by MeitY, Government of India (Project Number: S/MeitY/AM/20210114) under NLTM-Bhashini.
For any queries, please contact us at:
@misc{BSTD,
title = {{B}harat {S}cene {T}ext {D}ataset},
howpublished = {\url{https://github.com/Bhashini-IITJ/BharatSceneTextDataset}},
year = 2024,
}