Want to use Meta’s Segment Anything with #OpenSeadragon?
I finally wrote up a proper README for the #Annotorious plugin: https://github.com/annotorious/plugin-segment-anything Give it a try—and let me know what you build! #SegmentAnything #IIIF #Annotation

Wait... is that a new button in #IMMARKUS? Yes! Click the magic wand for automatic object selection, powered by #SegmentAnything. Just point, click, and let AI do the work! Try it out and let us know what you think. #DigitalHumanities #IIIF

Update: Smart object selection in zoomable images with #OpenSeadragon and #SegmentAnything! The next key challenge: Creating a seamless workflow—picking objects, refining shapes, and manually enhancing the outline. Stay tuned! #IMMARKUS #IIIF #DigitalHumanities #Maps

Next steps: #SegmentAnything in OpenSeadragon Sort of, at least. There's no way around the ~2s wait for viewport encoding. But it still feels bearable—especially given the huge speed-up for annotating complex shapes. Stay tuned!

Progress! Refine and extend areas by adding points to #SegmentAnything queries. Plus, polygon smoothing for cleaner annotation shapes. Next up: making it work in OpenSeadragon… #IMMARKUS #DigitalHumanities #Annotorious

Wow—this is #SegmentAnything 2 running *entirely* in the browser! With WebGPU (Chrome, Edge), generating embeddings took ~2s for a 1024x1024px region on my machine. No server required! #Annotation #DigitalHumanities #Maps #IIIF

I'm proud to have published my first scientific paper as preprint on arXiv! https://arxiv.org/abs/2310.08683

I investigated the use of augmented reality for reinforcement learning agents by integrating Meta Research's SAM (Segment Anything) zero-shot model into the RL training data pipeline. Like equipping a virtual entity (the RL agent) with AR goggles.

Does anyone know of open source projects that use AI to segment orthorectified imagery and try to categorise bicycle paths or other cycling infrastructure, and check if they’re missing from OpenStreetMap?

I’m thinking of whether it would be possible to run Meta’s Segment Anything model over suitably licensed aerial imagery to find infra missing on OSM. Any thoughts/comments very welcome!

DevSeed has now integrated DS-Annotate with the Segment Anything Model. This powerful tool enables swift and efficient annotation of map features, streamlining our processes significantly. This advancement wouldn't have been possible without our collaborative work with Ruben Mendoza, Ryan Avery, Wille Marcel, and Nick Ingalls.

RT @giswqs segment-geospatial v0.6.0 sneak peek - Interactive #segmentation of #remotesensing #imagery without coding. Segmentation results can be saved as GeoTIFF and vector data formats

https://github.com/opengeos/segment-geospatial/pull/44