Back to Search Start Over

Leveraging social media and deep learning to detect rare megafauna in video surveys

Authors :
Mannocci, Laura
Villon, Sébastien
Chaumont, Marc
Guellati, Nacim
Mouquet, Nicolas
Iovan, Corina
Vigliola, Laurent
Mouillot, David
Mannocci, Laura
Villon, Sébastien
Chaumont, Marc
Guellati, Nacim
Mouquet, Nicolas
Iovan, Corina
Vigliola, Laurent
Mouillot, David
Source :
Conservation Biology (0888-8892) (Wiley), 2022-02 , Vol. 36 , N. 1 , P. e13798 (11p.)
Publication Year :
2022

Abstract

Deep learning has become a key tool for the automated monitoring of animal populations with video-surveys. However, obtaining large amounts of images to train such models is a major challenge for rare and elusive species since field video-surveys provide few sightings. We propose a methodological framework that takes advantage of videos accumulated on social media for training deep learning models to detect rare charismatic species in the field. We apply our framework to aerial video-surveys of dugongs (Dugong dugon) in New Caledonia (South-Western Pacific). Convolutional neural networks trained with 1,303 social media images yielded only 25% of false positives and 38% of false negatives when tested on independent field video surveys. Incorporating a small number of images from New Caledonia (equivalent to 12% of social media images) into the training dataset resulted in a drop by nearly half of false negatives. We highlight how and the extent to which images collected on social media can offer a solid basis to train deep learning models for rare species detection and that the incorporation of few images from the study site further boosts detection accuracy. Our method provides a new generation of deep learning models able to rapidly and accurately process field video-surveys for the monitoring of charismatic species.

Details

Database :
OAIster
Journal :
Conservation Biology (0888-8892) (Wiley), 2022-02 , Vol. 36 , N. 1 , P. e13798 (11p.)
Notes :
application/pdf, English
Publication Type :
Electronic Resource
Accession number :
edsoai.on1286208035
Document Type :
Electronic Resource
Full Text :
https://doi.org/10.1111.cobi.13798