Project summary

I want to build an open evaluation kit for checking how visual localization systems fail when GNSS is unavailable and the scene provides weak, missing, or misleading evidence. This grows out of my peer-reviewed NaviLoc research (https://www.mdpi.com/2504-446X/10/2/97), but the grant work will be new code and public-data tooling that I can release independently of materials covered by my current contractual and NDA obligations.

What are this project's goals? How will you achieve them?

Over 12 weeks I will release a tested Python toolkit, dataset manifests using only public or permissively licensed data, reproducible baselines, and metrics for odometry drift, false visual matches, ambiguous landmarks, missing evidence, uncertainty calibration, and safe fallback. I will also document an edge-device demonstration on Raspberry Pi 5 and publish clear reproduction instructions.

How will this funding be used?

At the $10,000 minimum: $6,000 for engineering time, $2,000 for compute and storage, $1,200 for data preparation and annotation, and $800 for documentation and release. Funding above the minimum, up to $20,000, would expand the benchmark, pay for external replication, and support additional edge-device testing. Funds will not be used for university tuition.

Who is on your team? What's your track record on similar projects?

I will lead the project. I am a Ukrainian computer-vision researcher and the first author of NaviLoc. In the published system, the method reduced visual-inertial odometry drift by 32.1 times, achieved 19.5 m mean localization error, and ran at 9 FPS on a Raspberry Pi 5. I am an incoming MSc Advanced Computer Science student at Oxford.

What are the most likely causes and outcomes if this project fails?

The main risks are dataset licensing, a benchmark that is too narrow, and reproduction failures across hardware. I will use only clearly licensed sources, start with a small core benchmark, pin and test environments, and invite independent reproduction early. If funding is limited to the minimum, I will ship the core benchmark and one edge demonstration before expanding scope.

How much money have you raised in the last 12 months, and from where?

I have raised $0 for this independent open-source project in the last 12 months. Other grant applications are pending, but no award has been received.