dfki-av/PanoSAMic

PanoSAMic is a classic academic research implementation (from DFKI) addressing the geometric limitations of Meta's Segment Anything Model (SAM) when applied to 360-degree imagery. While the 'dual-view fusion' approach (ERP + CP) is a clever engineering solution to distortion, the project lacks any commercial moat. With only 5 stars and 1 fork, it has negligible market traction. The defensibility is low because the core logic is a wrapper/modification around SAM; once frontier labs (like Meta with SAM 2 or Google with specialized Street View models) release native support for spherical/equirectangular data, the need for this specific fusion architecture largely evaporates. Competitors include other SAM-based adaptations for niche geometries and native 360 models like HoHoNet. The displacement horizon is short (under 6 months) as the pace of SAM-derivative research is extremely fast and more integrated solutions for panoramic video are already emerging.