SOCOM D2P2: a vision-language model that searches SAR (SICD) for targets
OSW26BZ06-DV034 asks for a vision-language model that detects and classifies targets in synthetic-aperture radar imagery from sensor-independent complex data (SICD).
Official notice(opens in a new tab)Facts
- SOCOM
- SBIR OSW26BZ06-DV034 — VLM for SAR target search
- Due Oct 21, 12:00 PM ET
- U.S. Small Businesses; Direct-to-Phase-II
- SBIR Proposal on DSIP
What they’re asking for
Official topic text: analyze (detect and classify) SAR imagery from SICD. VLMs are described as combining multimodal imagery (video, SAR, IR) with a language model so the system can explain image content. That is the stated goal — not a general computer-vision kit.
How to respond
A Direct-to-Phase-II proposal on DSIP. Search OSW26BZ06-DV034.
Related
AFRL ADAM: sensing autonomy, algorithms and modeling
Kessel Run wants white papers on how to modernize the air operations center
Army Phase I: agentic AI that turns trade studies into auditable decisions
Army STTR twin of the agentic decision-management topic
Space Force SBIR: AI/ML to fuse missile-warning sensors across orbits