Insights and approaches using deep learning to classify wildlife
Problem
Conservationists need to automatically identify and classify wildlife species in large camera-trap image datasets, which is slow and error-prone manually.
Why it matters
Camera-trap image volumes far exceed human identification capacity, limiting wildlife monitoring and management.
What the AI did
A convolutional neural network was trained on 111,467 camera-trap images to classify 20 African wildlife species, with Grad-CAM used to highlight the most salient pixels.
Results
87.5% overall accuracy classifying 20 African wildlife species from 111,467 images
Impact
Demonstrates deep learning can automate camera-trap species classification at high accuracy.
Record
| Venue | Scientific Reports |
|---|---|
| Published | 2019 |
| Work type | article |
| Authors | Zhongqi Miao, Kaitlyn M. Gaynor, Jiayun Wang, Ziwei Liu, Oliver Muellerklein, Mohammad Sadegh Norouzzadeh, Alex McInturff, Rauri C. K. Bowie et al. |
| Citations | 136 (OpenAlex, at import time) |
| References | 38 |
| Model or method | Convolutional neural network (CNN) with Grad-CAM |
| Organization | not stated in source |
| Primary source | https://doi.org/10.1038/s41598-019-44565-w |
| Open access | gold |
| Keywords | deep learning, camera traps, wildlife classification, CNN, conservation, Grad-CAM |
Importance rating
real but narrower advance
Automating camera-trap species ID aids wildlife monitoring/management; abstract cites 111,467-image dataset but no broader human/industry impact scale.