Shypd.aiThe largest open index of AI toolsGroma is an AI Agents & Automation tool that offers a grounded multimodal large language model with localized visual tokenization. It excels in region understanding and visual grounding capabilities, taking user-defined region inputs to generate long-form responses.
Trending
About
Groma is a cutting-edge grounded multimodal large language model (MLLM) developed by FoundationVision, designed with exceptional region understanding and visual grounding capabilities. It introduces a novel paradigm of grounded MLLMs by utilizing localized visual tokenization, distinguishing it from models that rely on LLMs for localization or external modules. Groma can process user-defined region inputs, such as bounding boxes, and generate detailed, long-form responses that are deeply grounded in the visual context. This makes it a powerful tool for developing multimodal assistants that require precise visual comprehension and contextualized output. The project is open-source, providing model weights and instructions for installation, data preparation, training, and inference, making it accessible for researchers and developers to customize and build upon.
Capabilities
Pricing & Plans
Open Source
Free
FAQs
Trending
