segment-anything-model

Zero-shot image segmentation allows segmenting any object without task-specific training. The Segment Anything Model uses point, box, or previous-mask prompts to guide quick mask generation across medical, satellite, or data-labeling workflows.

For image segmentation, you can use point, box, or previous-mask prompts to guide the mask generation. The model also supports automatic mask generation to segment objects across images without manually providing prompts.

Yes, you can export the segmentation model to ONNX for deployment. The Skill supports ONNX export alongside multiple model variants like ViT-B, ViT-L, and ViT-H to fit various deployment environments and batched workflows.

Zero-shot segmentation works for medical and satellite imagery. It applies to annotation tools and data-labeling workflows, allowing quick object segmentation in specialized imagery without requiring any model re-training.

To handle batched or multi-image segmentation workflows, the Skill supports processing multiple images efficiently. You can load a pretrained model variant and apply automatic mask generation or prompt-based segmentation across your batch.