Skip to content

API Overview

Introduction

Welcome to the SABER Python API, your programmatic gateway to high-resolution segmentation in electron tomography (cryo-ET) and electron microscopy (EM). Whether you're integrating segmentation into large-scale pipelines or building custom tools, the SABER API gives you full control over the segmentation workflow.

With SABER's modular and extensible API, you can:

  • Load and process tomographic datasets and micrographs
  • Interface with SAM2 and custom domain expert classifiers
  • Apply segmentations in 2D or propagate across 3D volumes
  • Save, visualize, and evaluate segmentations with ease
  • Build custom workflows for your specific research needs

The API is designed for flexibility and reproducibility—ideal for researchers, developers, and power users who need programmatic control over their segmentation pipelines.

🏗️ Core Architecture

SABER's API is built around three main components:

1. Segmenter Classes (saber.segmenters)

The core segmentation engines that handle SAM2 integration and 3D propagation:

  • cryoMicroSegmenter - 2D micrograph segmentation with SAM2
  • cryoTomoSegmenter - 3D tomogram segmentation with video propagation
  • saber2Dsegmenter - Base class for 2D segmentation workflows
  • saber3Dsegmenter - Base class for 3D segmentation workflows

2. Entry Points (saber.entry_points)

High-level functions that provide complete segmentation workflows:

  • segment_micrograph_core - Complete 2D micrograph processing
  • segment_tomogram_core - Complete 3D tomogram processing
  • Parallel processing utilities for batch operations

3. Utility Modules

Supporting functionality for data handling and visualization:

  • saber.utils.io - Data loading and saving utilities
  • saber.utils.preprocessing - Image processing and helper functions
  • saber.visualization - Result visualization and galleries

🚀 Quick Start Paths

Choose your entry point based on your needs:

2D Micrograph Segmentation

Perfect for single-particle EM, SEM images, or 2D slices:

  • 2D Quickstart - Learn how to segment 2D EM images with SAM2 and SABER

3D Tomogram Segmentation

Ideal for cryo-ET, FIB-SEM, or any 3D volume data:

🔧 Key Features

Flexible Input Formats

# Support for various microscopy data formats
from copick_utils.io import readers
from saber.utils import io

# Load micrographs
image, pixel_size = io.read_micrograph("path/to/image.mrc")

# Load tomograms from Copick projects
vol = readers.tomogram(run, voxel_size=10, algorithm="denoised")

SAM2 Integration

# Direct access to SAM2 models with custom parameters
from saber.segmenters.micro import cryoMicroSegmenter

segmenter = cryoMicroSegmenter(
    sam2_cfg="large",  # Model size: tiny, base, large
    min_mask_area=50,  # Filter small artifacts
    window_size=256,   # Sliding window for large images
)

Domain Expert Classifiers

# Integrate custom classifiers for improved accuracy
from saber.classifier.models import common

classifier = common.get_predictor(
    model_weights="path/to/model.pth",
    model_config="path/to/config.yaml"
)

3D Propagation

# Automatic 3D segmentation from 2D annotations
from saber.segmenters.tomo import cryoTomoSegmenter

segmenter = cryoTomoSegmenter(sam2_cfg="large")
vol_masks = segmenter.segment(vol, slab_thickness=32)

📊 Output Formats

SABER provides flexible output options:

  • Zarr volumes - Efficient storage for large datasets
  • NumPy arrays - Direct access for custom analysis
  • Visualization galleries - PNG galleries for result review
  • Copick integration - Native support for collaborative annotation

📚 Next Steps

Ready to dive deeper? Follow these paths:

The SABER API is designed to be both powerful and accessible. Start with the quickstarts to see immediate results, then explore the detailed workflows for advanced usage.