Auspex Vision

Auspex Vision trains a single detection model across every label type in your dataset and runs it in your own stack. Nothing runs on this site.

One model, every label type

Most detection stacks make you pick a shape and build around it: a box model here, a segmentation model there, keypoints somewhere else, and glue to hold the three together. Auspex Vision trains bounding boxes, polygons, polylines, keypoints and tags in one model, from one dataset, in one run — and detects all of them.

That means mixed annotation sets stop being a problem to solve before training. If your data has rectangles on some images, polygons on others and keypoints on a handful, you can train on it as it is.

Polylines are a first-class label type

Polylines are the shape most detection stacks cannot train on at all. Lane markings, wires, seams, pipes, cracks and road edges are lines, not regions — flattening them into a bounding box throws away the geometry that made them worth labelling, and a polygon is the wrong shape for something with no interior.

Auspex Vision trains on polylines directly and detects them, alongside every other label type or on their own.

Or train one label type on its own

Mixing is optional. Train bounding boxes only, polygons only, polylines only, keypoints only or tags only — the same package, the same command, no separate model family to learn. Start with one shape and add the others later without changing tools.

Built for datasets that are too big to label twice

Auspex Vision is built to train on 10,000+ datasets, not just 10,000+ images, with no cap on the size of any one of them. It handles the scale you reach when annotation is an ongoing programme rather than a one-off task, and it trains on your data in your environment — the images never leave it.

Free to try, keyed after 30 days

Install it and start training straight away — no key, no account, no card. After 30 days training needs a licence key; prediction, validation, threshold calibration and export keep working either way.

Weights you train on your own data belong to you and keep working if a licence lapses. The key is verified offline, so it also works on air-gapped machines.