Model downloads

Navigation models, built for the edge.

Visual positioning for GNSS-denied navigation, packaged to run on the hardware a real vehicle carries. Every release ships as one versioned, checksummed bundle: the training checkpoint, exports for CPU runtimes, a compiled build for the Hailo-8 accelerator, and the evidence that the deployed model behaves like the one we evaluated.

• Visual positioning for GNSS-denied navigation
• Runs on Hailo-8 NPU, ARM, and x86
• INT8-vs-fp32 evidence in every release
• Versioned, checksummed, documented

What ships in every release.

One directory per release: the training checkpoint (PyTorch), ONNX exports for CPU inference on ARM and x86, a compiled Hailo-8 binary for the NPU, a model card describing intended use and limits, and machine-readable verification reports — with a SHA-256 checksum for every file.

• Deployment parity is measured, not assumed: each exported build is verified against the reference implementation on real flight imagery, and the report ships in the bundle.
• Quantised builds come with the accuracy evidence: the INT8-vs-fp32 comparison is published per release, so you can see the cost of the speed-up before you deploy it.
• Full provenance per release: a semantic version, the exact source checkpoint, per-file checksums, and a model card covering evaluation scope and known limits.

Current release.