Project 17 · Conservation ML
BIO
WildEye
Wildlife Species Classifier
01 Abstract
Wildlife Species Classifier
A conservation toolkit that classifies camera-trap imagery with an efficient edge model and computes ecological biodiversity metrics for ecosystem monitoring.
02 Highlights
Methods & contributions
- Biodiversity indices — Shannon H′, Gini–Simpson (1−D), Pielou evenness, species richness, and relative abundance index, all formula-verified.
- Ecology analytics — naive occupancy, observed-vs-expected co-occurrence, and diel-activity classification (diurnal / nocturnal / crepuscular).
- Edge model — MobileNetV3 with IR/night-vision handling, exported to ONNX.
- Serverless — AWS Lambda classification backed by DynamoDB.
03 Architecture
Data flow
camera trap → MobileNetV3 (ONNX) → species
→ biodiversity engine (Shannon · Simpson · occupancy)
04 Stack