Uploading the cied model

[AndreasZhukov]

This is the pull request to start the process of integration of cardiac device identification model with MHub, as discussed with my supervisor, Keno Bressem. The submission contains standalone demo application that allows to classify the cardiac devices based on smarpthone photos (or DICOM files converted to standard image formats) of chest X-rays.

The code that is relevant for MHub is located in the backend application file (app.py), namely, the functions predict_class, fix_bbox and resize_img. The requirements for the package can be seen in requirements_backend.txt. The models are hosted on Zenodo, links are available in Dockerfile-backend.

[LennyN95]

Hi Andreas,

I appreciate your interest in submitting a model to MHub.

Please follow the steps outlined in our documentation to prepare your contribution: https://github.com/MHubAI/documentation/blob/main/documentation/mhub_contribution/contributing_a_model.md

If you have any specific questions regarding the setup or packaging of your model in MHub, don't hesitate to send me an e-mail (@kbressem, feel free to share my contact details with Andreas) or comment below.

Some comments on the shared code:

• Ideally, you'll have your inference code in a separate repository that is cloned into the Dockerfile at build-time (you can see this is the standard for all MHub models, e.g., for the CaSuSt model)
• MHub operates all models as a stand-alone executable container operating on the file level (in/out folders); thus, there is no need for the frontend/backend