Cropilot Community

Open community for sharing models used in Cropilot — an AI-powered tool for automatically cropping scanned documents.

This organization is a shared home for fine-tuned computer-vision models that detect page boundaries, predict rotation, and generate crop instructions for digitized books, newspapers, periodicals, and other printed media. If you have trained a model, you are welcome to publish it here.

What lives here

• Crop / page-detection models — YOLO-based detectors that find page positions (e.g. outer, inner crop).
• Rotation models — ResNet-based models (RotateNET) that predict page rotation angle.

Every model is published as a standalone, reusable artifact with its own model card.

Using a model

Download a weight file directly:

from huggingface_hub import hf_hub_download

path = hf_hub_download(
    repo_id="cropilot-community/<model-name>",
    filename="<model-name>.pt",
)
print(path)

Then point your Cropilot configuration at the downloaded file. See each model's card for its recommended crop/rotation settings and intended use.

Contributing a model

Read CONTRIBUTING.md and copy the model card template — fill in the blanks, upload your weights, done.

Ecosystem

• Cropilot (app + backend): https://github.com/moravianlibrary/cropilot
• Cropilot utils (models, training, API tools): https://github.com/moravianlibrary/cropilot-utils

License

Unless a model card states otherwise, models in this organization are released under the GNU Affero General Public License v3.0 (AGPL-3.0), matching the Cropilot project.