Skip to content

Module 3: AI Literacy: Key Modalities and Modeling Paradigms

Module Summary

Introduction to Machine Learning and Computer Vision

This module introduces students to the core concepts and tools underlying artificial intelligence for ecological research. Students begin with foundational approaches in machine learning and computer vision, building an understanding of how AI can be used to extract information from images and other complex ecological data. The module then expands beyond imagery to explore multimodal ecosystem sensing, and how data from these measurements can be integrated to understand ecological systems. Finally, students are introduced to edge computing and AI-paired coding, exploring how AI models and computational tools can be brought closer to the point of data collection and how emerging AI-assisted programming approaches can support ecological research.

Week 1 - Introduction to Machine Learning and Computer Vision

Class Date: October 6, 2026

Instructors: Harry Chao, Sara Beery

Recording from 10/06 synchronous session (Slides)

Lesson Summary

This lesson introduces the field of machine learning, its basic concepts, essential components, different subcategories, and corresponding algorithms. This lesson also introduces the field of computer vision, covering key topics including Image classification, object detection, semantic segmentation, and representation learning, etc.

Learning Objectives

At the end of this lesson, students will be able to:

  • Know that machine learning is 'learning from data'
  • Understand the difference between supervised and unsupervised learning
  • Understand the limitations of ML algorithms
  • Understand what representation learning and feature extraction mean in computer vision
  • Understand that training neural network models can be both time-consuming and data-intensive
  • Describe a use case in which machine learning and computer vision algorithms can be useful for a real-world application
  • Identify and describe relevant computer vision and machine learning algorithms that could be useful for a given task
  • Select different candidate algorithms that are useful but not over-engineered for a given task

Pre-Lecture Activities and Reading

due by 2pm ET on Tuesday, October 6, 2026

Watch

  1. The Basics of Machine Learning View Slides
  2. Intro to Computer Vision
  3. Deep Learning for Computer Vision (optional) View Slides

Read

  1. Read Chapter 1 of Understanding Deep Learning
  2. Data Bias and Shifts in Foundations of Computer Vision
  3. Training for Robustness and Generality in Foundations of Computer Vision
  4. Transfer Learning and Adaptation in Foundations of Computer Vision

Post-class Assignments

  1. Journal Entry
    • Prompt: Start from an ecological problem you would like to solve. Is there a role for Machine Learning and/or Computer Vision in solving this problem? Why or why not? If there is, how would you map the ecological problem into a CV/ML task, and how would you evaluate success? Do you anticipate any risks or difficulties?

Class Session Office Hours

Who: Jacob Beattie

Date: Thursday, October 3, 2026

Time: 2pm Eastern Time

Zoom: Office Hour Zoom Link

Week 2 - Multimodal Ecosystem Sensing

Class Date: October 13, 2026

Instructors: Justin Kitzes, Devis Tuia, Steve Bullock

Recording from 10/13 synchronous session

Lesson Summary

This lesson explores multimodal ecosystem sensing and how data collected through different sensing technologies - including acoustic and remote sensing - can be integrated to better understand ecological systems and biodiversity at scale, and how to operate remote systems safely and legally. Students will consider how automated sensing and machine learning can support ecological research, from detecting and monitoring organisms to connecting observations across different modalities and scales.

Learning Objectives

At the end of this lesson, students will be able to:

  • Identify diverse technologies for data collection and processing,
  • Consider how automated approaches can support ecological research.
  • Recognize the importance of operating remote sensing technologies in a safe, ethical, and legal manner, within appropriate frameworks.

Pre-Lecture Activities and Reading

due by 2pm ET on Tuesday, October 13, 2026

Watch

  1. Remote Sensing Crash Course Devis Tuia
  2. Understanding and Protecting Biodiversity at Scale Using Automated Acoustic Surveys Justin Kitzes (Internal) (Click on 'View raw' to download and watch video)

Read

  1. Perspectives in machine learning for wildlife conservation
  2. Integrating AI models into ecological research workflows: The case of terrestrial bioacoustics
  3. Computational bioacoustics with deep learning: a review and roadmap (OPTIONAL)
  4. MaskSDM with Shapley values to improve flexibility, robustness and explainability in species distribution modelling (An example of multimodal data processing for ecology)
  5. A brief introduction to safety, ethics, and legal frameworks

Complete

  1. Submit 3 general questions about bioacoustics research based on the pre-read materials to the Module 3: Week 2 - Bioacoustics Questions Discussion.
  2. Field Guide Assignment: Look at any field guide (physical or online) or use other online resources to find three species of either (1) thrushes, or (2) sparrows that occur in your area. Then go to the Peterson Field Guide to Bird Sounds and find and listen to the songs made by those species. Answer the following three questions:
    • How hard would it be to identify this species in the field if you heard its song? Do you think you could do it if I placed you outside and played one of these songs through a speaker?
    • How would you explain to an undergraduate student, in words, how to tell the songs of these three species apart?
    • How would you "explain" or teach a computer how to tell the songs of three species apart? How well do you think it would work? (If you have little experience with machine learning, feel free to wildly speculate.)
  3. Make some progress on familiarisation with regulation, ethics, and safety:

Post-class Assignments

  1. Journal Entry

    • Prompt: Considering your skillset, interests, and research direction, what sensing and processing technologies might you employ in the field? What technological and wider frameworks might you employ?

Class Session Office Hours

Who: Steve Bullock

Date: Friday, October 16, 2026

Time: 2:00 PM ET

Zoom: Office Hour Zoom Link

Week 3 - Introduction to Edge Computing & AI-paired Coding

Class Date: October 20, 2026

Instructors: Hari Subramoni, Anirudh Potlapally, Zhongqi Miao, Pete Beckman, Andrew Quitmeyer, James Crall, Kalindi Fonda, Karina Torres

Recording from 10/20 synchronous session

Lesson Summary

This lesson will provide students with a practical framework for thinking about how AI can be deployed in ecological field settings. The session will emphasize key concepts and real-world examples, centered around a fundamental question: When ecological data are collected in the field, what computation should happen where, and why? The lecture will emphasize that deploying AI in the field is about more than model accuracy. It requires considering how sensing, computation, communication, data storage, power, connectivity, and scientific objectives interact to shape the design of a field-based AI system. Students will also be introduced to a range of edge computing systems and tools that they will encounter and use during the field course. Additionally, students will have the opportunity to engage in a live vibe-coding workshop for extra credit.

Learning Objectives

At the end of this lesson, students will be able to:

  • Describe how edge computing places computation closer to where data are generated.
  • Articulate how edge, gateway, cloud, and HPC form a continuum.
  • Discuss how different sensors and applications create different computational and communication requirements.
  • Extrapolate how the same hardware may require different processing strategies for different scientific questions.
  • Understand how computing decisions can influence the ecological data ultimately available for analysis.
  • Differentiate between different field course tools.

Pre-Lecture Activities and Reading

due by 2pm ET on Tuesday, October 20, 2026

Watch

  1. Computer for Ecology pre-lecture
  2. MothBox: inexpensive, open-source, automated insect monitor
  3. Sentinel: Upleveling Biodiversity Protection in the Peruvian Amazon
  4. Project Sparrow

Read

  1. Scout: Open-World Species Recognition on the Edge
  2. Towards edge processing of images from insect camera traps
  3. Reliable and efficient integration of AI into camera traps for smart wildlife monitoring based on continual learning
  4. Mothbox website and documentation (SKIM)
  5. Mothbox and Mothbot: automated light trap and data processing system for scalable insect monitoring
  6. Developing the Mothbox automated light trap (OPTIONAL)
  7. MothBox Open Hardware for Automated Insect Monitoring (OPTIONAL)
  8. Sentinel Documentation
  9. Project SPARROW and the future of conservation technology
  10. Sparrow Overview
  11. Sage (EXPLORE)

Post-class Assignments

  1. Journal Entry

    • Prompt: Considering the project themes presented earlier in the course, how do you think we could apply one of the technologies explored in this lesson to answer a related ecological question?

  2. Read:

Class Session Office Hours

Who: Anirudh Potlapally

Date: October 23, 2026

Time: 1:30pm ET

Zoom: Office Hour Zoom Link