TF-ID
Identification tools for tables and charts in academic literature
- Extract tables and charts from academic papers
- Provide model versions with and without titles
- Fine tuning checkpoints from Microsoft/Florence-2 model
- Support training custom models
- Open source model weights and manually annotated datasets
- Provide detailed training and usage guidelines
Product Details
TF-ID is a series of object detection models created by Yifei Hu for extracting tables and charts from academic papers. These models are fine tuned based on Microsoft/Florence-2 checkpoints, providing versions with or without title text, aimed at improving the accessibility and processing efficiency of academic literature information.
