Flowmind2Digital

This paper proposes the Flowmind2digital method and hdFlowmind dataset to address the convertion of hand-drawn flowchart/mindmap. Flowmind2digital is an comprehensive recognition and conversion method for flowminds, utilizing a neural network architecture and keypoint detection technology to enhance overall recognition accuracy. Our hdFlowmind dataset consists of 1,776 hand-drawn and manually annotated flowminds, covering 22 scenarios and surpassing existing datasets in size. Our experiments showcase the effectiveness of our method, with an accuracy rate of 87.3% on the hdFlowmind dataset, surpassing the previous state-of-the-art work by 11.9%. Additionally, our dataset demonstrates effectiveness, with a 2.9% increase in accuracy after pre-training and fine-tuning on Handwritten-diagram-dataset. We also highlight the importance of simple graphics for sketch recognition, which can improve accuracy by 9.3%.

Download paper here or here

Recommended citation: Cai Jianfeng. (2024). "Flowmind2Digital: The First Comprehensive Flowmind Recognition and Conversion Approach." arXiv preprint arXiv: 2401.03742, 2024.

BibTeX formatted citation:

@misc{liu2024flowmind2digital,
      title={Flowmind2Digital: The First Comprehensive Flowmind Recognition and Conversion Approach}, 
      author={Huanyu Liu and Jianfeng Cai and Tingjia Zhang and Hongsheng Li and Siyuan Wang and Guangming Zhu and Syed Afaq Ali Shah and Mohammed Bennamoun and Liang Zhang},
      year={2024},
      eprint={2401.03742},
      archivePrefix={arXiv},
      primaryClass={cs.CV}
}