Datasets:
metadata
task_categories:
- image-classification
pretty_name: doctype
size_categories:
- 10K<n<100K
Document Classification
It works per image input.
Classes
| Class | Processing Pipeline | Training Datasets |
|---|---|---|
document_printed |
Docling pipeline | HuggingFaceFW/finepdfs, RVL-CDIP |
document_handwritten |
Docling pipeline with handwriting OCR | IAM Handwritten Forms, RVL-CDIP |
photo |
VLLM: "Describe this image" | COCO 2017 |
diagram |
VLLM: "Describe this diagram" | AI2D, UML Diagrams |
chart |
VLLM: "Analyze this chart/graph" | PlotQA |
screenshot |
VLLM: "Analyze screenshot" | RICO, WebSRC |
map |
VLLM: "Analyze this map" | Satellite Maps, DeepGlobe Roads |