---
license: apache-2.0
---

DepictQA (with vision abstractor to reduce vision token numbers) model weights trained on **DQ495K**, **KonIQ**, and **SPAQ** datasets. 

See https://github.com/XPixelGroup/DepictQA for details.