Fine-tune Donut to extract data from chart
Data
The data used to train, validate and test is published by account named TeeA via this link huggingface.co/datasets/TeeA/Vietnamese-Chart-Dataset
Fine-tuning instruction
The model is fine-tuned following the instruction of Niels Rogge - Transformer Tutorial, 2020-09-02 via https://github.com/NielsRogge/Transformers-Tutorials
Load model
The model can be loaded by using DonutProcessor
from transformers import DonutProcessor, VisionEncoderDecoderModel
processor = DonutProcessor.from_pretrained("KhimNguyen/chart2text") model = VisionEncoderDecoderModel.from_pretrained("KhimNguyen/chart2text")