Match a query image against GPS-tagged tiles to infer its ground location.
npx skills add https://github.com/sarocu/tile-search --skill sar-locate
Please help me install this Agent Skill. Skill: sar-locate Source: https://github.com/sarocu/tile-search/tree/main/.claude/skills/sar-locate Command: npx skills add https://github.com/sarocu/tile-search --skill sar-locate
Match a query image against an incident's ingested tile corpus to locate where the image was captured by visual matching to GPS-tagged tiles.
Run extract_context.py with --image and --incident-id to build the candidate context and produce the initial context.json for the locate workflow.
High-intent search queries and answers about installing and using this skill.
To locate a query image, match it visually against a known GPS-tagged tile corpus to infer its ground location within an ingested SAR incident. The Skill uses a dHash pre-ranker to prune the search space and a multimodal verifier to confirm the best match.
The best way to search a large tile corpus is a two-stage visual matching workflow: a fast dHash pre-ranker narrows thousands of tiles to a top-K batch, then a multimodal verifier confirms the final match to infer ground location.
Yes, ImageMagick is a required dependency to run this geospatial image matching workflow. The Skill relies on it to support the underlying dHash and pHash visual matching operations against the GPS-tagged tile corpus.
Run the extract_context.py script with the --image and --incident-id parameters to build the candidate context. This action produces an initial context.json file to initiate the locate workflow against the built phash index.
Yes, you can constrain the geospatial search space by applying optional region filters and incident profiles. These region priors tailor the visual matching to specific sub-areas or mission contexts within the ingested SAR incident.
Matching a query image generates per-run artifacts including a context.json file, a candidates page, and result reports. These outputs support subsequent prompt rendering, reporting, and audit of the visual matching workflow.
sarocuIngests drone imagery into SAR incidents by tiling frames and filtering them into manifests.
camgrimsecAnalyze code to map attack surfaces and trace data flows from source to sink.
curiositechProcess geospatial data with PostGIS, GDAL, and npm packages.
fikriafMatch images and text using CLIP with PyTorch and transformers.
