Automating the taxonomic classification of malaria-vector mosquitoes to address global shortages in trained entomologists.

Manual identification of Anopheles mosquito species requires years of expert training, a resource that is critically scarce in malaria-endemic regions. This project builds a computer vision framework that automates species-level classification from specimen images, making vector surveillance accessible to non-experts.

Architectures Benchmarked:

• ResNet-50
• Vision Transformer (ViT)
• EfficientNet
• ConvNeXt
• MobileNet-V3 (selected for edge deployment)

Dataset: 2,400+ labeled Anopheles specimens

Key Engineering Decisions:

• MobileNet-V3 selected over heavier Transformer-based models for edge deployment suitability without sacrificing accuracy
• Local GPU caching strategy engineered to resolve cloud storage latency — reduced training epoch duration by 30×
• Translator Module built to convert numerical model outputs into human-readable entomological reports for non-expert users

digital-entomologist/
│
├── data/                    # Dataset loading and preprocessing
├── models/                  # Architecture implementations
│   ├── resnet50.py
│   ├── vit.py
│   ├── efficientnet.py
│   ├── convnext.py
│   └── mobilenet_v3.py      # Final selected model
├── training/                # Training loops and config
├── evaluation/              # Benchmarking and metrics
├── translator/              # Translator Module (output → readable reports)
├── notebooks/               # Exploratory analysis
├── requirements.txt
└── README.md

git clone https://github.com/JananiV3010/digital-entomologist.git
cd digital-entomologist
pip install -r requirements.txt

Python PyTorch OpenCV scikit-learn MobileNet-V3 Google Colab Kaggle

Malaria kills over 600,000 people annually. Accurate vector surveillance (knowing which mosquito species are present) is critical for targeted interventions. This tool democratizes that capability for field workers without specialized entomological training.

Janani Vaiyapuriappan — MSE Biomedical Engineering, Johns Hopkins University
LinkedIn · GitHub

Course project — AI-ML in Global Health, Johns Hopkins University (Fall 2025)