localization3d

Object Detection and 3DLocalization

https://github.com/arminabdeh/localization3d

Science Score: 57.0%

This score indicates how likely this project is to be science-related based on various indicators:

  • CITATION.cff file
    Found CITATION.cff file
  • codemeta.json file
    Found codemeta.json file
  • .zenodo.json file
    Found .zenodo.json file
  • DOI references
    Found 2 DOI reference(s) in README
  • Academic publication links
  • Academic email domains
  • Institutional organization owner
  • JOSS paper metadata
  • Scientific vocabulary similarity
    Low similarity (11.3%) to scientific vocabulary
Last synced: 10 months ago · JSON representation ·

Repository

Object Detection and 3DLocalization

Basic Info
  • Host: GitHub
  • Owner: arminabdeh
  • Language: Jupyter Notebook
  • Default Branch: smlm
  • Homepage:
  • Size: 196 KB
Statistics
  • Stars: 0
  • Watchers: 1
  • Forks: 0
  • Open Issues: 0
  • Releases: 3
Created over 2 years ago · Last pushed 10 months ago
Metadata Files
Readme Citation

README.md

LUENN: PyTorch Package for 3D Single Molecule Localization Microscopy

LUENN (Localization3D) is an open-source PyTorch package for 3D Single Molecule Localization Microscopy (SMLM).
It delivers end-to-end pipelines for synthetic data generation, deep learning model training, sub-pixel emitter localization, and 3D super-resolution image rendering.
Validated in peer-reviewed research, LUENN enables fast, accurate, and robust live-cell imaging, bridging the gap between computational microscopy and biomedical applications.

📄 Results published in: Applied Optics, 2024


✨ Key Features

  • Data Generation → create synthetic microscopy datasets for training and benchmarking.
  • Sampling Utilities → extract training/validation subsets efficiently from large datasets.
  • Custom Model Training → PyTorch-based CNNs for SMLM with flexible architectures and loss functions.
  • Post-Processing Functions → refine localization results, suppress artifacts, and enhance reconstruction quality.
  • 3D Localization → achieve sub-pixel emitter positioning in 3D, surpassing classical resolution limits.
  • Rendering Tools → generate and visualize high-resolution 3D reconstructions, including live-cell time series.

🎥 Example: 3D reconstruction and rendering of a live-cell dataset
3D Reconstruction


🚀 Performance

  • Robust across 2D, 3D, and live-cell microscopy modalities.
  • Achieves sub-10 nm localization accuracy under challenging noise/light conditions.
  • Processes live-cell SMLM data in ~3 seconds, enabling dynamic biological imaging.

📦 Installation

  1. Clone or download this repository.
  2. Create the environment (Linux example):
    ```bash conda env create -f environment_linux.yml conda activate LUENN
  3. Run training, inference, and reconstruction pipelines.

Performance

Luenn has demonstrated exceptional accuracy across a broad spectrum of imaging conditions. Its ability to handle live-cell SMLM data with reduced light exposure in just 3 seconds makes it a valuable asset for dynamic imaging scenarios.


System Requirements

  • Linux (Ubuntu 20.04+) or Windows 10
  • Python 3.8+
  • NVIDIA GPU with CUDA support (tested on Titan Xp, 12 GB VRAM)
  • 32 GB RAM recommended

Contributers:

Armin Abdehkakha, Email: arminabd@buffalo.edu
Craig Snoeyink, Email: craigsno@buffalo.edu

Owner

  • Name: Armin Abdehkakha
  • Login: arminabdeh
  • Kind: user
  • Company: University at Buffalo

Ph.D. Candidate, Mechanical Engineering, University at Buffalo

Citation (CITATION.cff)

# This CITATION.cff file was generated with cffinit.
# Visit https://bit.ly/cffinit to generate yours today!

cff-version: 1.2.0
title: LUENN
message: >-
  If you use this software, please cite it using the
  metadata from this file.
type: software
authors:
  - given-names: Armin
    family-names: Abdehkakha
    email: arminabd@buffalo.edu
    affiliation: University at Buffalo
  - given-names: Craig
    family-names: Snoeyink
    email: craigsno@buffalo.edu
    affiliation: University at Buffalo
    orcid: 'https://orcid.org/0000-0001-7215-2554'
repository-code: 'https://github.com/arminabdeh/LUENN_tf_version'
abstract: >-
  The development of Single-Molecule Localization Microscopy
  (SMLM) has enabled the visualization of sub-cellular
  structures, but its temporal resolution is limited. To
  address this issue, a deep-convolutional neural network
  called LUENN has been introduced, which uses a unique
  architecture that rejects the isolated emitter assumption.
  LUENN is a Python package based on a deep CNN that
  utilizes the Tensorflow tool for SMLM. It is capable of
  achieving high accuracy for a wide range of imaging
  modalities and frame densities. 
keywords:
  - Convolutional Neural Network
  - Single Molecule Localization Microscopy
  - Super-Resolution Microscopy
license: MIT

GitHub Events

Total
  • Push event: 1
Last Year
  • Push event: 1

Dependencies

setup.py pypi
.github/workflows/python-package-conda.yml actions
  • actions/checkout v4 composite
  • actions/setup-python v3 composite