thanhtung0512 / vietnamese_handwritten_ocr

1st Kalapa ByteBattles 2023 - Vietnamese Handwritten Text Recognition

1.Installation

- git clone https://github.com/manhdh32/1st_kalapa_ocr.git
- cd 1st_kalapa_ocr
- pip install -r requirements.txt

2. Data

• Kalapa dataset.
• Vietnamese address dataset
• Processed address corpus
• Synthetic data. Download my generated data here

3. Training

Get pretrained model here or generate synthetic data with:

python tools/gen_synthetic_data.py

and training from scratch:

python PaddleOCR/tools/train.py -c configs/pretrained_config.yml

Fine-tune:

python PaddleOCR/tools/train.py -c configs/final_kalapa.yml

4. Inference

Inference notebook here