fudan-zvg/Semantic-Segment-Anything

Health v2 · maintenance only

Flags: no_releases

Adoption not part of the score

2301 stars · 142 forks observed · 2026-08-28

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

Semantic Segment Anything (SSA) is a Python framework that adds semantic category prediction to the Segment Anything Model (SAM) by combining SAM's precise masks with existing semantic segmenters like Segformer, Oneformer, and CLIPSeg. It also includes SSA-engine, an automated dense open-vocabulary annotation engine used to generate the initial semantic labels for the SA-1B segmentation dataset.

Use cases

• add semantic category labels to SAM segmentation masks
• automatically annotate large image datasets with dense semantic categories
• generate open-vocabulary segmentation annotations without retraining SAM
• build a semantic segmentation dataset from SA-1B masks
• combine multiple segmenters with SAM for better mask boundaries
• reduce manual annotation costs for segmentation datasets

When to choose

• you need semantic categories on top of SAM's class-agnostic masks
• you want to auto-label a large image dataset for segmentation training
• you want to plug existing segmenters into SAM without fine-tuning its weights

When to avoid

• you need a simple off-the-shelf semantic segmenter with no SAM dependency
• you require real-time segmentation on edge devices
• you need actively maintained software with frequent updates

Facets

library · maturity maintenance

computer-vision image-processing machine-learning data-generation computer-vision image-processing machine-learning artificial-intelligence python segment-anything semantic-segmentation sam annotation-engine open-vocabulary dataset-labeling linux gpu

Member repositories

For agents

markdown · JSON · MCP: product_card(name="fudan-zvg/Semantic-Segment-Anything")