Pri-matrix Factorization

Data scientists from more than 90 countries around the world drew on 300,000 video clips in a competition to build the best machine learning models for identifying wildlife from camera trap footage. The results are powerful and – equally important … #climate

€20,000 in prizes
dec 2017
320 joined

About the project

Camera traps have become powerful tools for non-invasive wildlife research and conservation. Triggered by motion or heat, these automatic devices allow ecologists, anthropologists, and other researchers to study valuable footage of various species without requiring long hours of latent human time between sightings.

As part of the Pan African Programme: The Cultured Chimpanzee, over 8,000 hours of camera trap footage has been collected across various chimpanzee habitats from camera traps in 15 African countries. The programme is also collecting a wide variety of organic samples from these sites, such as feces, hair, plant matter, and information on the ecology and environment of each habitat. By scanning the videos from these traps and identifying the types of species and activity that can be seen, they hope to be able to understand the lives of these apes—their behaviors, relationships, and environments—and to extrapolate new ideas about human origins.

In addition to helping us better understand cultural evolution, the project is also documenting wildlife populations and biodiversity in these areas. Already they have both documented new chimpanzee behaviors and made some startling finds of animals in locations where they were no longer thought to live! They hope that drawing attention to a great many of these sites will incite conservation organizations to take an interest in these areas and move to protect them.

Knowledge of the species present in the videos can help project organizers better study, explore, and learn from the ecology of these habitats. However, labeling the species in camera trap footage is no small task. It takes a lot of time to determine whether or not there are any animals present in the data, and if so, which ones. This is where machine learning can help.

To date, thousands of citizen scientists have manually labeled Chimp&See data. In partnership with experts at The Max Planck Institute for Evolutionary Anthropology, we have created a well-labeled dataset of nearly 2000 hours of camera trap footage from Chimp&See's database.

Utilizing both crowdsourced labels as well as crowdsourced algorithms, this ambitious computer vision competition is in a league of its own. The winning techniques developed here will provide a starting point for production-level automated species tagging for use in camera trap systems around the world. By decreasing the time that experts spend watching empty footage, we can improve their ability to focus on the outcomes that matter most.

On the role of crowdsourced analysis and machine learning in evolutionary anthropology, project manager Mimi Arandjelovic, PhD says

"We partnered with the Zooniverse to develop where people from around the world have been recording what they see in each video. These citizen scientists also carry out more complex tasks such as identifying unique chimpanzee individuals and tagging primates and other animals to the species level. With this competition we hope to use machine learning to accomplish the first task more time efficiently thereby allowing citizen scientists to focus on the more complex tasks of the project. "


Setting a camera trap. Image courtesy of Flikr user teresehart.

This competition is generously funded by the ARCUS foundation.

Project Team

Christophe Boesch - Co-director of the Pan African Programme: The Cultured Chimpanzee Christophe is Professor and Director of the Department of Primatology of the Max Planck Institute for Evolutionary Anthropology in Germany. He has been studying apes for over 40 years across Africa, having established both the Tai Chimpanzee Project in Cote d’Ivoire and the Loango Ape Project in Gabon during this time. His research takes a holistic approach to studying chimpanzees and uses this to improve our understanding of the evolution of humans and their cognitive and cultural abilities. He is the author of several popular books and as the founding president of the Wild Chimpanzee Foundation he fights for a better future for the remaining wild ape populations at a grassroots level.

Hjalmar Kuehl - Co-director of the Pan African Programme: The Cultured Chimpanzee Hjalmar is a Robert Bosch Junior Professor and research group leader at the German Centre for Integrative Biodiversity Research and the Max Planck Institute for Evolutionary Anthropology. His research has focused on various issues in ape conservation, the development of wildlife survey and monitoring techniques, as well as questions in ape population ecology. More recently he has been focusing on new approaches to reduce the negative effects of natural resource exploitation in great ape habitats and combining evidence-based environmental protection with complexity science.

Mimi Arandjelovic - Programme manager of the Pan African Programme: The Cultured Chimpanzee Mimi is a junior scientist at the Max Planck Institute for Evolutionary Anthropology. Her research focuses on primate genetics, molecular ecology, conservation biology and citizen science and finding efficient means of studying wild animal populations non-invasively. In addition to her current research, she currently manages data collection from over 39 temporary research sites across Africa from all four chimpanzee subspecies.

Colleen Stephens - Statistician for the Pan African Programme: The Cultured Chimpanzee Colleen is lead statistician for the Chimp&See citizen science project at the Max Planck Institute for Evolutionary Anthropology. Her current research focuses on assessing the effectiveness of citizen science for annotating camera trap videos of wildlife and methods for improving video processing.

You can learn more here: