Makerere University • Acoustic AI

Non-Invasive, Real-Time Poultry Health Monitoring via Acoustic AI

VScreen is an edge-deployed, acoustic intelligence platform that abandons reactive, error-prone visual inspections in favor of real-time vocalization analysis. By utilizing Multi-Strategy Knowledge Distillation, we deliver complex disease detection algorithms directly onto low-cost farm infrastructure.

About the Platform

What is VScreen?

VScreen is an advanced agricultural technology platform that functions as a continuous, non-invasive health monitor for large-scale poultry farms. Instead of stressing birds with physical examinations, VScreen analyzes the ambient audio of the flock. By extracting high-level acoustic features and processing them through a compressed LSTM neural network, the system categorizes flock vocalizations into healthy, unhealthy, or environmental noise in real-time.

Commercial Poultry Farmers

Early warning alerts before catastrophic flock mortality

Veterinary Professionals

Clinical decision-support for precise medical diagnosis

AgTech Integrators & IoT Providers

Lightweight models for farm management systems

Acoustic Feature Extraction

MFCCs, spectral rolloffs, and zero-crossing rates from ambient audio

Multi-Strategy Ensemble Learning

Bagging + boosting algorithms for robust farm environment generalization

Knowledge Distillation

Compressing complex Teacher models into lightweight LSTM Student models

Optimized Edge Deployment

Running offline on microcontrollers and farm smartphones without cloud connection

99%+ Teacher Model Accuracy After 5-fold cross-validation
Sub-Second Diagnostic Alerts Real-time detection at the edge
256 Hz Acoustic Biomarker Low-frequency rumble of respiratory distress
VScreen in your hands

From flock sound to timely care

Explore how the mobile app brings flock checks, detection history, and nearby care providers together.

Record a flock sample

Start an audio check from the home screen and capture poultry vocalizations with your phone.

Get the Android app

Screen 1 of 8 · Home

App screen

The Problem We Solve

Flawed Visual Inspections

Current industry standard relies on manual visual inspection—highly prone to human error and slow to detect severe symptoms.

Cloud-Compute Dilemma

Audio signals are highly dimensional and complex. Processing them with deep learning usually requires cloud compute power that rural farms cannot access reliably.

Environmental Noise Interference

Background environmental noise creates severe data interference, causing generic models to overfit or trigger false alarms.

Why Uganda, Why Now

Our Strategic Market Opportunity

Across Africa, millions of farmers rely on poultry production for their livelihoods and national food security. Yet access to professional veterinary care remains severely limited, leading to critical delays in treatment.

By engineering a highly compressed, edge-deployable model locally, VScreen operationalizes the “Buy Uganda, Build Uganda” mandate. We provide a frugal, accessible, and sovereign technology that democratizes veterinary intelligence for the African farmer.

Locally Engineered Edge-Deployable Offline-Ready Sovereign Technology

Impact at a Glance

99%+teacher model accuracy
Sub-Secondreal-time diagnostic alerts
256 Hzacoustic biomarker of distress
4foundational pillars of architecture

Founders & Strategic Leads

Ishami Kwisanga Andric

Co-Founder & Technical Lead

Specializing in deep learning, software architecture, algorithm design, and system deployment.

Kyomuhendo Sumayah

Co-Founder & AI Engineer

Specializing in data science, acoustic signal processing, cybersecurity, and scalable software systems.

Ggaliwango Marvin (PhD)

Strategic Advisor & AI Mentor

Providing oversight on AI model optimization, research-to-product commercialization, and responsible technology governance at Makerere University.

Emmanuel Mugejjera (PhD)

Co-Investigator

Contributing expertise in agricultural IT, IoT, machine learning, and AI to guide VScreen's farm applications, connected-device integration, and practical deployment in poultry production.

Sarah Kaddu

Co-Investigator

Supporting VScreen's knowledge organisation, metadata, and information-governance work while helping strengthen ethical, inclusive, and responsible use of AI in the platform.

Alice Gitta Kutyamukama

Co-Investigator

Contributing semantic indexing, ontology design, and intelligent knowledge-management expertise to improve VScreen's information structure, contextual insights, and accessibility for research and innovation.

Published Research

Acoustic Feature Extraction And Multi-strategy Knowledge Distillation With Poultry Signal Vocalizations To Detect Diseases

IEEE Access · 2026

Authors: Ishami Kwisanga Andric, Kyomuhendo Sumayah, Ggaliwango Marvin, Emmanuel Mugejjera, Sarah Kaddu, and Alice Gitta Kutyamukama.

Our peer-reviewed publication documents the acoustic feature extraction, machine-learning evaluation, and knowledge-distillation research underlying VScreen.

Read the published paper