smart-fish-ponds-algeria/firmatech-backend-ai

Smart aquaculture AI backend built with FastAPI — enabling fish counting, disease detection, biomass estimation, and more. Designed for real-world farm automation with modular service architecture and AI integration. # 🐟 FIRMATECH-BACKEND-AI Smart aquaculture AI backend built with **FastAPI** — enabling fish counting, disease detection, biomass estimation, and more. Designed for real-world farm automation with modular service architecture and AI integration. --- ## 📁 Table of Contents * Features * Architecture * AI Models & Evaluation * Endpoints Overview * Project Structure * Getting Started * Environment Setup * Future Work * License --- ## ✅ Features | Module | Description | | -------------------- | -------------------------------------------------------------------- | | 🎥 Fish Counting | Real-time video-based fish counting using YOLOv8 + Centroid Tracking | | ⚖️ Weight Estimation | Biomass calculation via length-based prediction formula | | 🩺 Disease Detection | Classify fish sickness from image (e.g., Aeromonas, Tilapia Virus) | | 🧪 Farm Monitoring | Generate natural language reports from sensor data using Gemini API | | 🍽️ Food Prediction | *(Planned)* Predict fish feeding needs based on AI heuristics | --- ## 🧱 Architecture ``` FastAPI 🔼 Routers (API Endpoints) 🔼 Controllers (Logic Handlers) 🔼 Modules (Pydantic Schemas) 🔼 Services (Roboflow / Agent) 🔼 Utils (Tracking, Math utils) ``` > Uses **Roboflow** for AI inference and **Gemini Pro** for report generation. --- ## 📊 AI Models & Evaluation ### 🤖 1. Fish Detection & Counting (YOLOv8) * **Model**: Custom YOLOv8m (Roboflow) * **Input**: Video frames * **Output**: Bounding boxes of detected fish * **Post-processing**: `CentroidTracker` avoids double counting ### 🧬 2. Disease Classification * **Model**: Roboflow classification model * **Classes**: * `Healthy-Fish` * `Tilapia Lake Virus` * `Bacterial Aeromonas Disease` * `Streptococcus` * `null` (uncertain) ```json { "isSick": true, "sicknessType": "Streptococcus" } ``` ### ⚖️ 3. Weight Prediction * **Formula**: ``` Weight (g) = 0.0196 × Length(cm)^ …