Group Assignment

Group Assignment: Deep Learning for Ecology

Objective

The objective of this assignment is to develop a small research project investigating how deep learning can be used to address challenges in ecology. Working in groups, you will choose a research question, explore an ecological dataset, implement and evaluate a deep learning approach, and discuss the implications of your findings.

Your project can focus on either:

• An ecological question (e.g. how many species are present at a specific location?).

• A methodological question (e.g. investigating how a particular methodological choice affects deep learning performance).

You can choose to work on biologging data, wildlife images or passive acoustic monitoring. The project should demonstrate not only that you can implement a model, but also that you understand your data, evaluate your approach correctly and interpret your results in an ecological context.

Expected outcomes

By the end of the project, you should be able to:

• Formulate a clear ecological research question.

• Explore and prepare an ecological dataset.

• Implement a baseline and a deep learning approach.

• Evaluate model performance using appropriate metrics.

• Interpret results, analyse errors and discuss limitations.

• Present your work as a small scientific study.

Deliverables

1. Project proposal: One paragraph describing the ecological motivation, research question, dataset, planned experiments and expected challenges. (Wednesday Week 2)

2. Progress demonstration: A short demonstration of the data pipeline and initial baseline results.(Optional) 

3. Final presentation: A 10-minute presentation followed by questions. (to be adapted based on the number of groups)

4. Code and documentation: A runnable notebook or equivalent code, with instructions and a README explaining the workflow on github.

Assessment

The assignment will be assessed based on the research question, data preparation, model implementation, experimental design, evaluation, interpretation, presentation and reproducibility.

A high-performing model is not required to obtain a good grade. The quality of the scientific reasoning, the validity of the experiments and the interpretation of the results are equally important.

Grading :  60 points