Neptune Core

Neptune is a field-dependent SMLM localization pipeline with continuous zmap conditioning, standard training, full-FOV inference, reconstruction, and evaluation workflows.

This repository is the public implementation package. It intentionally excludes large raw data, full checkpoints, historical outputs, and internal experiment logs.

Quick Start

pip install -e .
neptune train --config configs/standard_v0_1/train_single_left.yaml
neptune infer --config configs/infer_recon_v0_2/single_left_full8000.yaml
neptune eval --config configs/eval_v0_1/density1_evalset.yaml
neptune fit-zmap --config configs/zmap_v0_1/microtube_left.yaml
neptune render --config configs/infer_recon_v0_2/render_subpixel.yaml

Layout

• src/neptune/: Python package.
• configs/: versioned public configs.
• scripts/: shell and SLURM wrappers.
• tests/: public test suite.
• docs/: standards, architecture, and API notes.
• outputs/: runtime output root, ignored by git except documentation.

Public Standards

• Training standard: docs/standards/standard_v0_1.md
• Dual-channel standard: docs/standards/dual_standard_v0_1.md
• Inference/reconstruction standard: docs/standards/infer_recon_v0_2.md

Public CLI

• neptune train: standard training entry.
• neptune infer: full-ROI infer entry.
• neptune eval: metric computation and aggregation.
• neptune fit-zmap: NAT zmap fitting and promotion.
• neptune render: standard reconstruction renderer.