University of Wyoming Raccoon Project

About the University of Wyoming Raccoon Project

As the human population continues to grow, urbanization of the land is expected to grow as well. Although many species around the globe are in currently in decline due to urbanization, we see that some species not only survive, but actually thrive in urban habitats. Animals like raccoons have demonstrated an incredible ability to adapt to new environments, including cities, and are expanding their range into new areas where they have not existed previously. Raccoons use the cover of darkness to explore human environments and are experts at exploiting human refuse, earning them the affectionate title of "trash pandas". They are well-known for many of their iconic physical traits, such as their dexterous forepaws, ringed tail, and bandit-like mask. Raccoons also have a reputation for being intelligent, which likely aids them in their successful conquest of human environments. Interestingly however, raccoons have been largely overlooked in the field of animal cognition, likely because of the difficulty involved in studying these nocturnal, cunning animals.

The University of Wyoming Raccoon Project was established in 2015 to better understand raccoons and their cognition, especially their ability to learn independently or from others, their ability to innovate solutions to new problems, and their ability to behave flexibly. We design various cognitive challenges, or "puzzles", for wild raccoons and study how they approach and solve these challenges. We use a variety of high-tech equipment and new experimental methods to conduct our trials at night without the involvement of human experimenters. We also conduct comparative research with other successful urban carnivores, such as striped skunks and coyotes.

About Our Experiment

These videos were collected during an experiment on the learning, problem-solving, and social behavior of free-ranging raccoons and skunks in the town of Laramie, Wyoming. We set up the experiment at night, when raccoons and skunks are most active. Below is a list of the equipment we used that you may be curious about.

Puzzle Box

A puzzle box is an apparatus an animal must figure out how to open in order to get a food reward. The puzzle box you see in the videos has 12 doors on each side, and behind each door is a small amount of sardines and dog kibble (a tasty treat for animals like raccoons and skunks!). We are able to measure the learning and problem-solving skills of animals in the videos by measuring (1) the amount of time it takes the animal to open each door and (2) the diversity of behaviors they use to open the doors. We also record other items of interest, such as social behaviors.

Night Vision Cameras

We set up four infrared (or "night vision") cameras around the puzzle box to record trials. Like us, the animals in the videos cannot see this light.

PIT tag reader

Every year the UWRP humanely traps, identifies, and releases raccoons and skunks as a part of our research. Each animal we trap is marked with a passive integrated transponder tag ("PIT tag," like a microchip used with pets). We place a PIT tag reader at our study site, and the reader scans and records the PIT tag number of the animals that come to the puzzle box. This way, we can match up the PIT tag records to our camera footage to find out who is who!

Ear Tags

Many of the raccoons and skunks also receive ear tags for visual identification.

Radio Collars

A subset of raccoons in our study population have been fitted with very high frequency (VHF) radio collars. We use VHF telemetry to detect the signal emitted from the collars and find the location of the raccoons after we have released them. We use this information to better understand the movement and home range size of raccoons in the town of Laramie. The majority of raccoon tracking is performed by University of Wyoming undergraduate students that are building skills for a future career in wildlife biology and management.

About the aims of our Citizen Science Project

The overall goal of this project is to gather data that will allow us to train a machine learning algorithm to distinguish between individual raccoons. We want to be able to say whether the raccoon in a video is Alice the raccoon or Bob the raccoon based on the video data alone. To get there, we need to be able to determine a few things about each video

1. how many animals are in a video,
2. where they appear in the frame,
3. what species each animal in the video is,
4. the individual identity of each of the animals in the video.

We could ask each of these questions in separate citizen science projects. However, by gathering the bounding box data for each frame in a video, we are gathering the data required to address the first three points. We will aim to use the data you provide to train a machine learning algorithm to automatically draw bounding boxes and identify the species in each box. It will take a lot of training data for the algorithm to learn to do this and it will be an iterative process. As you annotate more and more images, we will use these to improve the algorithm, allowing you to focus attention on the data that we need the most help with. For example, as the algorithm improves we should be able to remove more of the images that do not contain any animals allowing you to focus on those that do.

Once we have a reliable algorithm that can draw a box around each animal in an image and identify its species, we can link the PIT tags to the raccoons in the bounding boxes (addressing point 4. above). While this should be easy for videos with one or two raccoons, it will be a lot trickier when there are many raccoons. We might need your help again in the future to deal with these situations in order to track raccoons between frames or even identify who each of the raccoons is.

We hope to use what we learn from this work to help us identify individuals in other camera trap projects. This will allow ecologists to more accurately estimate populations of animals and enhance the study of wild animal behavior and personality.