Alibaba Cloud Developer

Nov 19, 2019 · Artificial Intelligence

Can AI Imagine Visually? Seq‑SG2SL for Scene‑to‑Semantic Layout

This article introduces the Seq‑SG2SL framework, which tackles the challenge of granting AI visual imagination by converting scene graphs into semantic layouts, discusses the limitations of existing text‑to‑image methods, proposes the SLEU metric for automatic evaluation, and presents experimental results demonstrating its effectiveness.