https://solarsystem.video/videos/watch/3bc6626f-1626-4aa5-8dd3-bbd4b22e26cf Presented by Jichao Fang (NIU) ML4PSP Seminar April 2026 Mars orbital archives now contain enough imagery that finding morphologically similar features is bottlenecked by search, not data. We present a planetary-scale similarity search system built on foundation-model embeddings over the full CTX Murray global mosaic (~26.9M indexed locations). A Vision Transformer pretrained via self-supervised learning on millions of CTX patches produces embeddings that capture surface texture and landform semantics without any labels. Deployed as a quantized vector index on a single server, the system supports sub-second instance-level retrieval ("find terrains like this"), geo-filtered search within regions of interest, and interactive relevance feedback for iterative refinement. The system is publicly accessible at findmars.space. Sat, 06 Jun 2026 09:23:21 GMT https://validator.w3.org/feed/docs/rss2.html PeerTube - https://solarsystem.video https://solarsystem.video/client/assets/images/icons/icon-1500x1500.png https://solarsystem.video/videos/watch/3bc6626f-1626-4aa5-8dd3-bbd4b22e26cf All rights reserved, unless otherwise specified in the terms specified at https://solarsystem.video/about and potential licenses granted by each content's rightholder.