PTDR | paper

█████████████████████████████████████████

█ Dataset Download

Real-world PTDR

The real-world images and their annotations can be downloaded via simply cloning or downlading the repo as follow:

git clone https://github.com/zobeirraisi/PTDR.git

The downloaded folder contains both the detection and recognition data.

PTDR/
├── dataset
│   ├── detection       #The real images of detection dataset
│   │   ├── test
│   │   └── train
│   ├── recognition     #The real cropped images of recognition dataset
│   │   ├── test
│   │   └── train
│   └── synth           #Synthetic dataset for recognition pretraining
│       └── link.md     #Download link of synthetic dataset
├── README.md           #This file

Synthetic PTDR

The synthetic dataset that used only for pre-trainin of recognition task can be downloded from:

█DROP-BOX-LINK

█████████████████████████████████████████

█ Training and Evaluation

┌─────────────────────────────────────┐
│             TBD                     │
├─────────────────────────────────────┤
│              │                      │
│              │                      │
└─────────────────────────────────────┘

█████████████████████████████████████████

█ CITATION

@article{raisi2025ptdr,
  title={PTDR: A Real-World and Synthetic Benchmark Dataset for Persian Scene and Document Text Detection and Recognition},
  author={Raisi, Zobeir and Nazarzehi Had, Valimohammad and Sarani, Esmaeil and Damani, Raosul},
  journal={SN Computer Science},
  volume={6},
  number={6},
  pages={1--16},
  year={2025},
  publisher={Springer}
}