How thenukadharmaseelan/eyebridge works
EyeBridge is a batch and Python-API tool for registering retinal FAF images onto IR images. Its main workflow loads image pairs and vessel masks, estimates a transform from a field-of-view scale seed and LoFTR/RANSAC matching, optionally refines it with dense flow, and returns metrics and diagnostic outputs. Landmark-based fallback and landmark evaluation are additional capabilities. The graph follows the README and indexed references; unsampled implementation details are not inferred beyond those references.
Main entry points: pipeline.py Registration API · cli.py Batch CLI · registration.py FOV seed and matching · matching.py LoFTR transform fit
Components 16
Inputs and interface
- Pair CSV
- Batch CLI
src/cli.py - Image and mask loading
src/io.py
Registration
- Registration API
src/pipeline.py - Vessel preprocessing
src/preprocessing.py - FOV seed and matching
src/registration.py - LoFTR transform fit
src/matching.py - LoFTR model loader
src/models/loftr.py - Dense-flow refinement
src/registration.py - Landmark fallback
src/landmarks.py
Evaluation
- Image and vessel metrics
src/metrics.py - Landmark evaluation
src/landmarks.py
Results and diagnostics
- Batch result tables
src/cli.py - Diagnostic visualizations
src/visualization.py
Other components
- Research user
- Output directory
Connections 18
- Research user → Batch CLI: runs batch
README.md - Research user → Registration API: calls API
README.md - Pair CSV → Batch CLI: supplies pairs
README.md - Batch CLI → Image and mask loading: loads inputs
src/cli.py - Batch CLI → Registration API: registers pairs
src/cli.py - Registration API → Image and mask loading: loads images
src/pipeline.py - Registration API → FOV seed and matching: estimates transform
src/pipeline.py - FOV seed and matching → LoFTR transform fit: fits matches
src/registration.py - LoFTR transform fit → LoFTR model loader: runs matcher
src/matching.py - FOV seed and matching → Vessel preprocessing: uses vessel helpers
src/registration.py - Registration API → Dense-flow refinement: refines transform
README.md - Registration API → Landmark fallback: uses fallback
README.md - Registration API → Image and vessel metrics: assembles metrics
src/pipeline.py - Batch CLI → Landmark evaluation: evaluates landmarks
README.md - Registration API → Diagnostic visualizations: creates diagnostics
src/pipeline.py - Batch CLI → Batch result tables: writes summaries
README.md - Batch result tables → Output directory: writes files
README.md - Diagnostic visualizations → Output directory: saves images
README.md