OpenViSTAI · Agents · Blockchain · Crypto
Submit a project
Menu
AI Models

Sa2Va

Sa2VA is an open-source Python multimodal model that combines SAM-2 with multimodal language models for pixel-level grounded understanding of images and videos.

Explore the practical guide ↓

Independent OpenViST project profile, based on linked sources.

Sa2Va — OpenViST project profile in AI Models.
OpenViST project illustration. See the linked sources for product details.
Understand the project

Practical overview

When it fits

Use Sa2VA when a visual-language task needs grounded outputs, such as referring segmentation, visual prompting, or image and video chat, rather than only high-level image descriptions.

Who it is for

  • Computer-vision researchers working on dense visual grounding and segmentation.
  • Developers exploring image or video conversations linked to specific visual regions or content.

Documented uses

  • Build image and video question-answering with pixel-level grounded responses.
  • Connect language descriptions to regions in images or content in videos.
  • Experiment with dense visual understanding that combines multimodal language models and segmentation.
Before you choose

Requirements & limits

The code repository is public and licensed under Apache-2.0.

  • The model combines SAM-2 with an MLLM, linking language-driven understanding to dense visual regions.
  • The repository documents InternVL2.5/3 and Qwen2.5-VL/Qwen3-VL backbones.
  • Dependencies are managed with uv, with latest and legacy environment options.
  • Training and evaluation procedures are project-specific and are referenced in the related project instructions.

Check current requirements in the official documentation. Source-based guidance does not mean OpenViST has run or independently tested this project.

Source-based practical guide

Set up the shared project environment

Create the repository's shared Python environment for Sa2VA and its related projects.

Before you start

  • The public Sa2VA repository
  • uv installed

Steps

  1. Clone or otherwise obtain the repository.

  2. From the repository root, run the command below to set up the latest environment, or use the command below for InternVL2.5 or earlier.

    Documented command
    bash setup_env.sh
    Documented command
    bash setup_env.sh sa2va legacy

  3. Alternatively, enter projects/sa2va and run the command below or the command below.

    Documented command
    uv sync --extra=latest
    Documented command
    uv sync --extra=legacy

  4. Activate the environment with source projects/sa2va/.venv/bin/activate before following a project's training or evaluation instructions.

Expected outcome

A configured shared project environment, ready for project-specific training or evaluation steps.

Guide limitations

  • This setup guide does not cover running a particular training or evaluation task.
  • The repository documents Hugging Face or OpenRouter tokens and API keys as configuration items for relevant workflows.
Documentation 1 ↗
Evidence and transparency

Sources & review

Official repository ↗
Source record details
Source availability
Active
Sources checked
14 October 2026
Original website title
bytedance/Sa2VA · Official homepage title · retained in its source language
How we review sourcesSuggest a correction