What AI says about FG-CLIP

FG-CLIP is the official implementation of a bilingual fine-grained vision-language alignment model (FG-CLIP 2) by 360CVGroup, enabling cross-modal understanding between images and Chinese-English text. It delivers fine-grained alignment across 29 datasets and 8 diverse tasks, outperforming baselines like SigLIP 2 and MetaCLIP 2 in both languages. The repository provides code and model weights, API access, an MCP server, and demos to support testing, retrieval, and visualization of dense features.