geti

Why we included this project

Geti started out as OTX, the OpenVINO Training Extensions library, and the training engine it grew out of is still around, now published separately on PyPI as getitune. The application itself runs locally as a single container or a native Windows app and moves you from a folder of raw images to a deployed model, bundling PyTorch training, OpenVINO-optimized inference, and model compression into that one box. The built-in transfer, incremental, and semi-supervised learning support is aimed at teams with limited labeled data, so you can often get a usable detection, segmentation, or classification model from less annotation. And because getitune is a standalone Python package, you can call the same training routines from your own code instead of working only through the UI. For small teams on Intel hardware that want fast iteration without stitching together a dozen scripts, this keeps the whole pipeline in one place.