About TableLabeling

TableLabeling is a graphical tool developed with Python and PyQt5 for creating and editing table annotations in images. The application helps users easily define table structures, making it particularly useful for training table structure recognition models. Notably, it features specialized capabilities to handle challenging layouts, such as borderless tables and complex merged cells.

Recent Update

• 2026.07: Our public VietFinTab dataset includes over 10,000 labeled table images from the financial reports of 13 Vietnamese companies. All data were annotated using the TableLabeling tool.

  VietFinTab dataset: https://huggingface.co/datasets/VietFinTabGroup/VietFinTab

Installation

You can install and run the application with the following steps:

1. Install dependencies

   pip install -r requirements.txt

2. Download model files

The tool works with TATR-based models (Table Transformer) for automatic table structure detection. The application expects the model files to be located inside the model/ directory.

Please download your TATR model file separately (e.g., TATR-v1.1-Fin.pth, TATR-SciTSR.pth, etc.) and place it in the model/ folder as shown below:

project_root/
├── main.py
├── table_editor.py
├── auto_labeling.py
├── model/
│   └── your_tatr_model.pth

3. Run the application

   python main.py

Usage

1. Select folders

After opening the application:

• Go to File → Select image folder
  → Choose a folder containing images (.png, .jpg, .jpeg, .bmp).

• Go to File → Select label folder
  → Choose the folder where XML annotations will be saved.

The left panel will display a list of images with status indicators:

• ❌ Missing label – No annotation found
• ⚠️ Unchecked label – Annotation exists but not yet verified
• ✅ Checked (saved) – Annotation has been saved and confirmed

2. Viewing and navigating images

• Click any image in the list to load it.
• Use the mouse wheel to zoom in/out.
• Click and drag to pan around the image.
• The viewer automatically fits the table editor when resized.

3. Creating and editing table structure labels

To manually create table labels:

1. Click Create table label in the right panel.
2. Enter the number of rows and columns.
3. A grid will appear on top of the image.

You can then:

• Select cells, rows, or columns
• Merge / Unmerge cells
• Drag borders to adjust cell sizes
• View updated cell details in the cell list on the right panel

4. Auto-generate table labels using TATR

The application can automatically generate table annotations using TATR-based models.

Note: The examples provided use our fine-tuned model rather than the original TATR model.

Option A — Auto-create for the current image

• Click Auto Create table label

Option B — Auto-create for all images in the folder

• Go to Tool → Auto Create all table labels

The dialog will automatically close when processing is complete.

Contribution

Pull requests and improvements are welcome! Please send a pull request.

License

LICENSE

Citation: Hong Quan Pham, Hai Dang Bui. TableLabeling. Git code (2026). https://github.com/dangbuiii/TableLabeling