Skip to content

CLAUDE.md

Latest commit

 

History

History
92 lines (62 loc) · 3.49 KB

CLAUDE.md

File metadata and controls

92 lines (62 loc) · 3.49 KB

CLAUDE.md

This file provides guidance to Claude Code (claude.ai/code) when working with code in this repository.

Project Overview

BigDataViewer-BIOP-Tools is a Fiji plugin package providing tools for BigDataViewer, including image fusion, GPU-accelerated deconvolution, registration, and data processing. Developed by BIOP (BioImaging and Optics Platform) at EPFL.

Build Commands

# Full build
mvn clean install

# Build without tests (faster)
mvn clean install -DskipTests

# Run all tests
mvn test

# Run a specific test class
mvn test -Dtest=DemoDeconvolution

Architecture

Command-Based Design

The codebase uses SciJava Commands as the primary extension mechanism. All user-facing functionality is implemented as commands annotated with @Plugin(type = Command.class). Commands extend BdvPlaygroundActionCommand and use @Parameter annotations for inputs/outputs.

Key command packages:

  • ch.epfl.biop.scijava.command.bdv/ - BDV visualization commands
  • ch.epfl.biop.scijava.command.source/ - Source manipulation (deconvolution, registration, export)
  • ch.epfl.biop.scijava.command.spimdata/ - SPIM data handling
  • ch.epfl.biop.scijava.command.transform/ - Transform operations

Source Processing Pipeline

The SourcesProcessor interface (ch.epfl.biop.source.processor) defines a functional pattern for chaining image operations:

SourceAndConverter[] → SourcesProcessor → SourceAndConverter[]

Processors are composable, serializable (JSON), and include: affine transforms, channel selection, fusion, identity operations.

GPU Deconvolution (CLIJ2-based)

Deconvolver class implements lazy, tiled Richardson-Lucy deconvolution:

  • Uses CLIJ2-FFT for GPU acceleration (OpenCL)
  • Processes images in configurable overlapping tiles
  • Results cached via ImgLib2's CachedCellImg
  • Supports multi-GPU and GPU pooling

Alpha Blending System

bdv.util.source.alpha/ and bdv.util.source.fused/ provide layer-based compositing:

  • AlphaSource wraps sources with alpha channels
  • LayerAlphaProjectorFactory handles multi-layer blending
  • Blending modes: smooth average, sum, max, min, distance-weighted

Image Fusion

Block-based fusion with bounded cache (triggers at 50% RAM) for large datasets:

  • Addresses BigStitcher memory issues with Soft References
  • Pre-filters tiles overlapping each block before fusion
  • Exports to OME-TIFF (8/16-bit, RGB) and XML-HDF5

Registration Framework

ch.epfl.biop.registration/ provides pair-based registration:

  • Elastix integration for affine/spline deformable registration
  • SIFT-based registration
  • BigWarp transform support
  • 2D registration in selected rectangular regions

Key Dependencies

  • BigDataViewer ecosystem: bigdataviewer-core, bigdataviewer-playground, bigdataviewer-vistools
  • GPU: clij2-fft_ for deconvolution (requires OpenCL devices)
  • Image I/O: bigdataviewer-image-loaders, quick-start-czi-reader for CZI files
  • Registration: bigwarp_fiji, mpicbg_ (SIFT)

Test Data

Test classes in src/test/java/ download datasets from Zenodo. Internet connectivity required for running demos like DemoDeconvolution or DemoLLS7.

Important Notes

  • Lazy evaluation: BDV sources compute on-demand; actual processing happens during visualization or export
  • SciJava annotations: All commands must have proper @Plugin and @Parameter annotations for Fiji integration
  • JSON serialization: Processors use JSON adapters - maintain compatibility when modifying processor classes