Detecting Tables In Documents: Computer Vision Resources

Created by inuwamobarak and last updated in September 2023, this repository provides resources for table detection within unstructured documents. It utilizes computer vision architectures like DETR and CNNs to identify table structures across various document types.

Use Cases

• Training DETR models to predict bounding boxes for tables in document images
• Evaluating CNN-based table extraction performance on unstructured layouts
• Fine-tuning Hugging Face transformer models for document layout analysis

Strengths

• Utilizes DETR (Detection Transformer) architectures for table localization
• Focuses on unstructured document formats
• Includes integration with Hugging Face datasets

Limitations

• Unknown record count and specific document sources
• Lack of explicit license information
• No public metadata regarding label accuracy or class distribution

Provenance

Source

inuwamobarak via GitHub

Freshness

Last updated September 2023; no stated update frequency.