Back to projects

Smart Aquaculture AI / Firmatech

Prototype smart aquaculture platform with fish counting, disease checks, biomass estimates, pond monitoring, alerts, reports, mobile flows, and an admin dashboard.

PrototypeNo public repoSmart aquaculture AI platform / 2024-2025

Problem

Aquaculture operators need to track water quality, fish health, biomass, feeding, and alerts, but manual checks and disconnected records slow response.

Solution

Split the MVP into an Expo app, Next.js dashboard, TypeScript/MongoDB API, FastAPI AI backend, data simulator, and notification service.

Key Features

AI fish counting from video frames using YOLOv8-style Roboflow detection and centroid tracking
Fish weight and biomass estimation from image-based length detection and an empirical formula
Disease classification for healthy fish, Aeromonas, Streptococcus, Tilapia Lake Virus, and uncertain cases
Sensor monitoring for pH, oxygen, temperature, water level, salinity, nitrate, and suspended solids
Gemini/Agno report generation from tank measurements and fish details
Expo mobile app for pond overview, alerts, profiles, notifications, and multi-pond navigation
Next.js admin dashboard for real-time pond monitoring, role management, feeds, and analytics
Cron-based data simulator for hackathon demos with generated measurements and dummy fish images

Pond monitoring pipeline

two backends, one app
POND INPUTSsensorspH · O₂ · tempcamerascount · weightsimulatorcron + dummy imagesstands in for hardwareExpress coreusers · tanks · feedsTypeScriptFastAPIvision + reportsPythonrecordsframesMongoDBMongooseYOLOv8countingdisease classifierRoboflowGemini + Agnofarm report textExpo appfarmer viewAdmin dashboardNext.jsalertspush + emailreads / writesORIGINbachelor end-of-study project · graded 18/20AWARDBest Potential Startup
Two backends because they fail differently: the Express core serves farm records that must always answer, and the FastAPI service runs models that are slow and occasionally wrong. The simulator is the honest part — the ponds had no instrumented hardware yet, so cron jobs and dummy images stood in for sensors so the rest of the system could be built and demonstrated against realistic input.

Firmatech AI Screenshots

10 views
Backend data model for users, water tanks, measurements, reports, alerts, feeding, and fish details.
Mobile farm dashboard with pond counts, fish totals, alerts, status, and navigation.
Push-notification evidence for critical pond events, including a temperature alert.
Feed stock alert flow using Expo/Firebase-style mobile notifications.
Fish counting flow: simulated frames, Roboflow/YOLOv8 inference, and centroid tracking to avoid duplicate counts.
Weight and biomass flow: detect length, convert pixels to centimeters, then apply the empirical formula.
Disease detection flow with Roboflow classification for healthy fish, Aeromonas, Streptococcus, Tilapia Lake Virus, and uncertain cases.
Daily report flow that formats tank measurements and generates operator-readable recommendations through Gemini/Agno.
Fish detection training curves covering loss, precision, recall, mAP50, and mAP50-95.
Disease model training curves for classification and detection quality review.