Introduction to DeepCell

This material is intended to help users become acclimated with the DeepCell ecosystem. DeepCell addresses three key needs for deep learning and biological images:

1. How can I use deep learning easily on my data?
2. How can I interact with these predictions?
3. How can I improve these model predictions?

This tutorial will provide a gentle introduction to all three areas. Additionally, we have included a "Getting Started" section for users that may be unfamiliar with the basic tools covered in this material.

Table of Contents

Getting started

• Required software installations
• Intro to Unix, Docker, and Git
• Python best practices
• Basic Python, Numpy, and Scipy exercises
• Intro to Python image processing for live-cell imaging
• Intro to deep learning with tensorflow

Analyzing my images with pre-trained models

• Summary of available models
• Running pre-trained models in the cloud
• Running pre-trained models locally

Labeling my data with DeepCell Label

• Load files
• Use DeepCell Label

Building new and improved models with deepcell-tf

• Introduction to deepcell-tf
• Training a model in Google Colab

Publications

To learn more about the various systems and software that comprise DeepCell, please refer to the publications below. Relevant links are highlighted below each publication.

Greenwald, Miller et al. Whole-cell segmentation of tissue images with human-level performance using large-scale data annotation and deep learning

• The TissueNet dataset is available at deepcell.datasets
• The Mesmer pipeline can be run from deepcell.org or via ImageJ/QuPath plugins by reading our documentation on running pretrained models in the cloud
• The Mesmer pipeline can be run locally via a jupyter notebook as shown in this example notebook, or from the command line using our application Docker
• The ark-analysis multiplexed image analysis pipeline is at this github link
• All code for model training and figure generation for the paper can be found in our publication figures repo

Bannon et al. DeepCell Kiosk: scaling deep learning–enabled cellular image analysis with Kubernetes

• The DeepCell Kiosk can be downloaded from the github repo and additional documentation is hosted on Read the Docs
• A persistent deployment of the Kiosk is hosted at https://deepcell.org/
• The code used to generate figures from the paper is available in our publication figure repo

Copyright

Copyright © 2016-2021 The Van Valen Lab at the California Institute of Technology (Caltech), with support from the Shurl and Kay Curci Foundation, Google Research Cloud, the Paul Allen Family Foundation, & National Institutes of Health (NIH) under Grant U24CA224309-01. All rights reserved.

License

This software is licensed under a modified APACHE2. See LICENSE for full details.

Trademarks

All other trademarks referenced herein are the property of their respective owners.

Credits