A First Step Towards Automated Species Recognition from Camera Trap Images of Mammals Using AI in a European Temperate Forest
Problem
Manually labeling camera trap images of mammals for wildlife monitoring is slow and hard to scale; the paper automates species identification in Białowieża Forest, Poland.
Why it matters
Camera traps are used worldwide to monitor wildlife, and automated labeling could make large-scale conservation monitoring practical.
What the AI did
A YOLOv5-based image recognition pipeline automatically labeled camera trap images of 12 common medium-size and large mammal species, with data managed via TRAPPER software.
Results
Average accuracy of 85% F1-score identifying the 12 most common mammal species, using 2659 images with animals.
Impact
Preliminary demonstration that lightweight deep learning can automate camera trap labeling with limited training data.
Record
| Venue | Lecture notes in computer science |
|---|---|
| Published | 2021 |
| Work type | conference-paper |
| Authors | Mateusz Choiński, Mateusz Rogowski, Piotr Tynecki, Dries P. J. Kuijper, Marcin Churski, Jakub W. Bubnicki |
| Citations | 33 (OpenAlex, at import time) |
| References | 19 |
| Model or method | YOLOv5 |
| Organization | not stated in source |
| Primary source | https://doi.org/10.1007/978-3-030-84340-3_24 |
| Open access | hybrid |
| Keywords | camera traps, YOLOv5, species recognition, wildlife monitoring, deep learning |
Importance rating
real but narrower advance
Abstract states camera traps are used worldwide for wildlife monitoring and that automated classification would significantly speed data processing, but demonstrated only on 2659 images of 12 species in one forest.