SegMate: A Segmentation Toolkit

This repository is a toolkit for using the Segment Anything Model (SAM) developed by Meta AI. It is capable of accurately “cutting out” any object from an image with just a single click.

SAM is a promptable segmentation system that exhibits zero-shot generalization to unfamiliar objects and images, eliminating the need for additional training. By providing prompts that specify what needs to be segmented in an image, SAM can perform a wide range of segmentation tasks without requiring additional training.

Features

• Easy inference with SAM, supporting bounding boxes, points, masks, and text prompts

• Automatic masking without the need for prompts

• Finetune SAM on custom datasets

• GroundingDINO integration for text prompt segmentation

• Training a custom decoder to auto segment a specific type of object

---

Contents:

• Home
• SAM Architecture
  • Image Encoder
  • Prompt Encoder
  • Mask Decoder
• Getting Started
  • Installation
  • Example Usage
• API Documentation
  • Segmentation Model
  • Object Detector
  • Dataset
• Credits and License
  • Main Contributors
  • License