Page 229 - AI for Good Innovate for Impact

AI for Good Innovate for Impact

Use Case 11: AI biodiversity monitoring

Organization: SBER

Country: Russia Change 4.2-Climate

Contact Persons:
Primary Contact: Iuliia Zemtsova – YMZemtsova@ sberbank .ru
Secondary Contact: Oleg Artyugin – oyartyugin@ sberbank .ru
Other Contacts:
Konstantin Gongalskii – Gongalsky@ sev -in .ru
Jose A Hernandez-Blanco – j.a.hernandez.blanco@ yandex .ru
Denis Malikov – zoolog.22@yandex.ru

1 Use Case Summary Table

Item Details
Category Climate Change/Natural Disaster

Problem Addressed AI biodiversity monitoring
Key Aspects of Solu- Use of AI for optimized camera trap monitoring to classify 25 native
tion species in the Russian Far East and Altai

Technology machine learning, computer vision, conservation, ecology, camera traps,
Keywords wildlife, MegaDetector

Data Availability Private (Plans to open-source repository in 2025)
Metadata Photo

Model Training 73,687 images (60/20/20 train/val/test split) using EVA2 model (F1-macro:
0.93)
Pilot Deployment https:// biodiversity .ai4good .ru/

Code Repositories Not publicly available

2 Use Case Description

2�1 Description

Camera traps, deployed globally for ecological monitoring, capture vast amounts of imagery
triggered by motion or on a set schedule. These images are traditionally reviewed manually,
which is labor-intensive and time-consuming—often cluttered with irrelevant (empty or human/
vehicle) images.

This project integrates AI to automate species identification from camera trap images. Our model
suite includes a detector (MegaDetector) followed by a classifier capable of distinguishing 25
native species, thereby reducing manual effort and accelerating biodiversity assessments.

193