About the Series

Introducing the WILDLABS Tech Tutors, our new series that focuses on answering the "how do I do that?" questions of conservation tech. Launched with the support of Microsoft AI for Earth, this series will give you the bite-sized, easy-to-understand building blocks you'll need to try new conservation technology, enhance your research, or DIY a project for the first time.

Taking place every Thursday, each Tech Tutor will present a 20 minute tutorial guiding you through an aspect of conservation tech, followed by a 10 minute live Q&A session with the audience.

For participants, the outcome will be an increased sense of confidence in their technological skills, the ability to actively build off of the skills discussed in these tutorials, and an opportunity to learn and collaborate with other members of the WILDLABS community. Our goal is to customize these tutorials to fit the needs of the community and address your needs, so let us know what you want to see in this season and beyond.

Can't make it? You can find every tutorial after it airs on our Youtube channel.

Meet Your Tutor: Sara Beery, Caltech, Microsoft AI for Earth

My research focuses on machine learning and computer vision for biodiversity monitoring, particularly for detection and recognition of animal species in challenging camera trap data at a global scale. I work closely with Microsoft AI for Earth and Google Research/Wildlife Insights where I help turn my research into usable tools for the ecology/biodiversity community.

Engage with Sara in the WILDLABS Community, Twitter and her website.

We asked Sara...

What will I learn in this episode?

You will learn about existing resources for training computer vision models, and how to curate, annotate, and structure data from your study area to build project-specific detection or classification models for camera trap data.

How do you recommend I learn more about this subject?

Try the Microsoft AI for Earth MegaDetector demo, and look through Dan Morris's excellent Camera Trap ML Survey. 

I recommend the following papers as a starting point, there are many more!

• Three critical factors affecting automated image species recognition performance for camera traps

• Efficient Pipeline for Camera Trap Image Review

• Improving the accessibility and transferability of machine learning algorithms for identification of animals in camera trap images: MLWIC2

• Recognition in Terra Incognita

• Deep Learning Object Detection Methods for Ecological Camera Trap Data

• Context R-CNN: Long Term Temporal Context for Per-Camera Object Detection

You can also look through the iWildCam Competitions. 

[NB: Sara's talk from our Camera Trapping Virtual Meetup also gives a great introduction to AI for Camera Traps. Watch it here)

If I want to take the next step using this technology, where should I start?

This is a great place to find resources!

What advice do you have for a complete beginner in this subject?

Think about what you want your machine learning model to do, then quantify the data and labels that you already have. Be realistic in your computer vision model expectations based on the variability and quantity of your training data.

Learn more about our upcoming Tech Tutorials

Visit the series page on WILDLABS to find the full list of WILDLABS Tech Tutors.