ECTS Credits: 3

Course Parameters:

• Language: English
• Level of Course: PhD course
• Semester/Quarter: Q1 2024
• Hours per Week: 3 weeks + homework = 80 hours in total.
• Capacity Limits: 2 participants

Objectives of the Course:

The course aims to provide the PhD students with foundational knowledge and practical skills in using AI and machine learning techniques for image segmentation, with a focus on applications in various fields.

Learning Outcomes and Competences:

Upon completing the course, the student should be able to:

• Understand the fundamentals of AI and machine learning algorithms.
• Apply AI and machine learning techniques to image segmentation tasks.
• Design and implement image segmentation strategies for real-world applications.
• Evaluate and fine-tune machine learning models for image segmentation.
• Present and discuss the results of image segmentation projects.

Compulsory Program:

Entire course

Course Contents:

Week1:

Introduction to AI and Machine Learning

• Overview of AI and machine learning concepts.
• The fundamentals of machine learning
  • Types of learning (supervised, unsupervised, semi-supervised)
  • The main challenges of machine learning: Data quality and quantity, model fit.
  • Training models and gradient descent.
  • Model testing, validation, selection, and evaluation.
  • Model hyperparameters and tuning.
• Introduction to common machine learning models
  • Linear regression, Logistic regression, SVM, CART and Random Forests
• Dimensionality reduction
  • The curse of dimensionality
  • Principal component analysis (PCA)

Practical Application

• Worked through examples with Python/Jupyter Notebooks

Week 2:

Fundamentals of Image Processing

• Introduction to images and image segmentation.
• Image preprocessing techniques: Noise reduction and image enhancement.
• Image filters and feature extraction.
• Neural networks

Supervised Learning for Image Segmentation

• Training data preparation.
• Building and training convolutional neural networks (CNNs) for image segmentation.
• Evaluating model performance.

Practical Application

• Machine learning models for image processing
• Implementing a CNN for image segmentation.

Week 3:

Project Work

• Project work on real-world image segmentation tasks.
• Fine-tuning models and optimizing segmentation performance.
• Data presentation and discussions of project outcomes.

Prerequisites:

 Basic understanding of coding programs such as Python or R

Name of Lecturer:

Mihailo Azhar

Type of Course/Teaching Methods:

The course combines lectures on campus, hands-on exercises (tutorials), and project work. Students will work on practical image segmentation projects.

Literature:

 Course materials and references will be provided at the beginning of the course.

Participants are expected to have acquired “Hands-On Machine Learning with Scikit-Learn, Keras, and TensorFlow 3e: Concepts, Tools, and Techniques to Build Intelligent Systems” – Aurelien Geron, before course start.

Course homepage:

None

Course Assessment:

Course participation will be assessed based on:

- Active attendance

- Completion of hands-on exercises (tutorials)

- Successful implementation of image segmentation in a real-life project and presentation of findings

Provider:

 Aarhus University, Ecoscience AU. Marine Biodiversity and experimental ecology

Special Comments on This Course:

The course will take place at Aarhus University, Ecoscience at Risø, 399 Frederiksborgvej, 4000 Roskilde, Denmark

Course Fee:

PhD students enrolled at a Danish University may not be charged a course fee. for PhD students enrolled at non-Danish university a fee of DKK 3600 will be charged.

Registration:

Deadline for registration is 30 November 2023. Information regarding admission will be sent out no later than 4 December 2023.

For registration: send an e-mail to mihailo.azhar@ecos.au.dk stating your name and affiliation (University)

If you have any questions, please contact Mihailo or Marc, mihailo.azhar@ecos.au.dk, mca@ecos.au.dk