Gemma explained: PaliGemma architecture- Google Developers Blog

This post details the PaliGemma architecture, a vision-language model (VLM) that combines the SigLIP vision model with the Gemma language model. It explains the components: the vision tower (SiglipVisionModel) with its patch embedding and positional embedding, the multi-modal projector, and the language model (GemmaForCausalLM). It also provides an example of object segmentation output and explains how coordinates and segmentation masks are encoded.