Build Your Own Image Segmentation Pipeline in Python

Workshop Objectives
Workshop Poster

About:

Participation will be limited. The format will be a combination of tutorials and hands-on exercises.

The Image Analysis Collaboratory is running a workshop on building your own deep learning image segmentation pipeline in Python. This workshop will introduce participants to the fundamentals of image segmentation and how to train a U-Net for this,including creating ground truth labels and understanding key deep learning concepts. We will then dive into building and training a deep learning model (U-Net) for image segmentation and explore evaluation metrics to assess model performance.

This hands-on workshop will combine tutorials with interactive exercises to give participants practical experience in developing and evaluating segmentation models using deep learning methods in Python.

The methods covered will be largely agnostic to the imaging modality, meaning that images can be from bright-field, fluorescence, confocal, light-sheet, or electron microscopy or any other modality. Though we focus on 2D images, extension to 3D images is relatively straightforward in many cases.

Time:

Monday December 9th 2024, from 9 AM to 5 PM

Place:

Harvard Medical School.

Format:

In-person.

Skill level:

Intermediate.

Requirements:

• Gmail account (to access Google Colab)
• Coding experience in Python
• A laptop to perform the hands-on exercises
• An interest in learning some bioimage analysis

Scope:

• Introduction to image segmentation
• Tools to create ground truth labels
• Introduction to deep learning
• Building your own deep learning model (UNet)
• Evaluation metrics

Cost

$200.

Eligibility and billing

Eligible individuals include internal HMS researchers, HMS affiliates, and external researchers.  Attendees/users will be required to be setup in the PPMS System, which coincides with the billing/invoicing process to collect payment for attending the workshop.  Information required for system setup and billing is collected only from those accepted to the workshop.
Billing occurs monthly, and you will be invoiced within the month following the workshop.  Individuals internal to HMS will be billed for payment via 33-digit code, and all others with be billed for payment (PO or CC) via electronic invoice automated through the PPMS system.  Payment is due net 30 days for all payments.  A no show fee of the full registration cost will be charged for 48 hour or less (not including weekends or holidays) notice of cancellation/no-show.

Applying:

You apply here: https://tinyurl.com/IACsegmentation
If you want to be notified for future events and workshops, join our mailing list, if you didn’t already.

Deadline for application:

11:59 pm on Friday, November 15, 2024

Selection:

Mostly based on your short motivation and skill level. In case of over-subscription, the available spaces will be distributed evenly between labs and departments. You will be informed by email about your application.
Selection announcement: November 22, 2024

NB

Our workshops are often oversubscribed, so please be considerate to your fellow applicants and let us know well in advance if you are accepted and cannot make it.

Looking forward to see you there
Ranit Karmakar, Antoine Ruzette, Federico Gasparoli, Maria Theiss, and Simon F. Nørrelykke.