Seol mar théacs é seo: Learning texton models for real-time scene context