lxtGH/OMG-Seg

Health v2 · maintenance only

Flags: no_releases no_license

Adoption not part of the score

1354 stars · 55 forks observed · 2026-08-28

What it is AI-extracted, prompt v1, taxonomy v1, 2026-08-30, confidence not recorded

Official research codebase for OMG-Seg (CVPR 2024) and OMG-LLaVA (NeurIPS 2024), unified models for image-level, object-level, and pixel-level visual perception and reasoning. OMG-LLaVA combines a universal segmentation encoder with an LLM to support text and visual prompts for flexible interaction.

Use cases

• run promptable image segmentation with a single unified model
• combine an LLM with pixel-level segmentation for visual reasoning
• reproduce OMG-Seg and OMG-LLaVA paper results
• build a multimodal chatbot that outputs segmentation masks
• train a universal segmentation model on multiple benchmarks
• interact with images using point, box, or text prompts

When to choose

• you need one model handling semantic, instance, and panoptic segmentation plus reasoning
• you want a research baseline for LLM-driven segmentation
• you need pixel-level outputs from a multimodal LLM

When to avoid

• you need a lightweight production segmentation service without LLM overhead
• you lack GPU resources for large multimodal training or inference
• you need a permissively licensed library for commercial embedding

Facets

library · maturity active

computer-vision image-processing machine-learning deep-learning llm-inference nlp computer-vision image-processing large-language-models artificial-intelligence deep-learning python segmentation multimodal visual-reasoning research-code cvpr neurips promptable-segmentation gpu linux

Member repositories

For agents

markdown · JSON · MCP: product_card(name="lxtGH/OMG-Seg")