Researchers at the University of Wyoming used a database of 3.2 million photos taken by hidden camera traps in the Serengeti National Park in Tanzania to train a deep-learning system to distinguish between 48 animal species, including elephants, giraffes, and gazelles.
During testing, the system correctly identified the species present in an image 92% of the time.
The researchers say the system is better at identifying the most common animals in the dataset, and it has greater difficulty identifying rarer species.
The team notes the system could be used to classify most of the photos in the database, and it could be further trained on hand-labeled images to improve its performance at recognizing rarer species.
The researchers also plan to test whether the system can identify animal behavior in images.
From New Scientist
View Full Article
Abstracts Copyright © 2017 Information Inc., Bethesda, Maryland, USA
No entries found