PaliGemma – Making Gemma 2 see by adding a vision encoder

From the creator

Explore how PaliGemma adds a SigLIP vision encoder to Gemma 2. This model is pre-trained on captioning, question answering, object detection, and even segmentation. Varying image resolution and model size allows to scale the compute by a factor of 155. If you have data available for your task, fine-tuning PaliGemma will result in great performance, especially on text-related tasks. Subscribe to Google for Developers → https://goo.gle/developers #Gemma #GemmaDeveloperDay Speaker: Andreas Steiner Products Mentioned: Gemma

Published 2025-04-02 · Watch on YouTube ↗
Availability and playback restrictions are controlled by the original channel.