Skip to content

DevPool1/ecobin

BranchesTags

Repository files navigation

ECOBIN Logo

♻️ ECOBIN — Smart Waste Sorting Bin

AI-powered waste classification and automatic sorting using computer vision

Status Platform AI License

About • Architecture • Features • Tech Stack • Getting Started • Project Structure • License

🇵🇹 Sobre

Projeto académico para a unidade curricular Laboratório IoT — Universidade do Algarve (UAlg)

O ECOBIN é um contentor de lixo inteligente que classifica resíduos automaticamente usando visão computacional e IA (Google Gemini 2.0 Flash). O sistema captura uma imagem do resíduo depositado, classifica-o em 4 categorias, e aciona um carrossel mecânico rotativo para depositar o lixo no compartimento correto — tudo de forma autónoma.

🌍 About

ECOBIN is an AI-powered smart waste bin that automatically classifies and sorts waste using computer vision. When a user drops waste into the bin, an ESP32-CAM captures an image, a Raspberry Pi sends it to the Google Gemini 2.0 Flash API for classification, and a motorized carousel rotates to deposit the waste into the correct compartment — all autonomously.

Key Highlights

  • 🤖 Real-time AI classification using Google Gemini 2.0 Flash
  • 🔄 4-compartment motorized carousel with stepper motor
  • 📡 Distributed IoT architecture with 4 hardware nodes via MQTT
  • 📊 Live web dashboard with classification history and fill levels
  • 🔌 Offline resilience — basic operation continues without internet
  • 🖨️ 3D-printed enclosure — custom cylindrical design

🏗️ Architecture

The ECOBIN uses a distributed architecture with 4 hardware nodes communicating via Wi-Fi + MQTT:

┌─────────────────────────────────────────────────────────────┐
│                    Wi-Fi Network (MQTT)                      │
├─────────────┬──────────────┬──────────────┬─────────────────┤
│             │              │              │                 │
│  📷 Vision  │ ⚙️ Container │ 🖥️ Interface │ 🧠 Gateway      │
│  ESP32-CAM  │ ESP32-WROOM  │ Arduino R4   │ Raspberry Pi    │
│             │              │   WiFi       │                 │
│ • Camera    │ • Stepper    │ • OLED       │ • MQTT Broker   │
│   OV2640    │   motor      │   display    │ • Gemini API    │
│ • HTTP      │ • Servo      │ • NeoPixel   │ • SQLite DB     │
│   endpoint  │ • Reed SW    │ • PIR sensor │ • Web Dashboard │
│             │ • Ultrasonic │ • Buzzer     │ • REST API      │
└─────────────┴──────────────┴──────────────┴─────────────────┘

Operation Flow

sequenceDiagram
    participant U as 👤 User
    participant PIR as PIR Sensor
    participant CAM as ESP32-CAM
    participant RPI as Raspberry Pi
    participant AI as Gemini API
    participant MOT as ESP32 (Motors)
    participant DB as SQLite

    U->>PIR: Approaches bin
    PIR->>RPI: Motion detected (MQTT)
    RPI->>CAM: Wake up & prepare
    U->>CAM: Drops waste on tray
    CAM->>RPI: JPEG frame (HTTP)
    RPI->>AI: Classify image
    AI->>RPI: JSON {category, confidence}
    RPI->>MOT: Rotate carousel (MQTT)
    MOT->>MOT: Stepper → correct compartment
    MOT->>MOT: Servo opens trapdoor
    MOT->>RPI: Reed switch confirms close
    RPI->>DB: Log event + image
Loading

✨ Features

Feature Description
🤖 AI Classification Real-time waste classification via Gemini 2.0 Flash
🔄 Auto-Sorting 4-compartment carousel with stepper motor (90° increments)
📷 Vision System ESP32-CAM with OV2640 for waste imaging
💡 Visual Feedback NeoPixel LED ring — color-coded by waste category
🔊 Audio Feedback Buzzer for accept/reject confirmation
📊 Web Dashboard Real-time fill levels, classification history with images
🌙 Night Mode LDR-based ambient light detection
📡 MQTT Communication Decoupled, event-driven architecture
💾 Event Logging SQLite database with full history and timestamps
🔌 Offline Mode Core functionality works without internet
🖨️ 3D Printed Custom cylindrical enclosure in PLA

🛠️ Tech Stack

Layer Technology
AI / Classification Google Gemini 2.0 Flash API
Communication MQTT (Mosquitto broker)
Database SQLite
Backend Python 3 (Flask/FastAPI)
Dashboard HTML / CSS / JS (WebSocket)
Firmware Arduino IDE / PlatformIO (C/C++)
3D Printing PLA filament, custom design
Gateway OS Raspberry Pi OS

🚀 Getting Started

Prerequisites

  • Raspberry Pi 4 (2GB+) with Raspberry Pi OS
  • ESP32-CAM (OV2640)
  • ESP32-WROOM-32UE
  • Arduino UNO R4 WiFi
  • Python 3.9+
  • Arduino IDE or PlatformIO
  • Google Gemini API key

Installation

# Clone the repository
git clone https://github.com/<YOUR_USERNAME>/ecobin.git
cd ecobin

# Set up the gateway (Raspberry Pi)
cd gateway
python -m venv venv
source venv/bin/activate
pip install -r requirements.txt

# Configure environment variables
cp .env.example .env
# Edit .env with your Gemini API key and MQTT settings

# Start the system
python main.py

Firmware Upload

# ESP32-CAM (Vision Node)
cd firmware/esp32_cam
# Open in Arduino IDE → Select board → Upload

# ESP32-WROOM (Container Node)  
cd firmware/esp32_contentor
# Open in Arduino IDE → Select board → Upload

# Arduino R4 WiFi (Interface Node)
cd firmware/arduino_interface
# Open in Arduino IDE → Select board → Upload

📁 Project Structure

ecobin/
├── 📄 README.md
├── 📄 CLAUDE.md                  # AI assistant context
├── 📄 LICENSE
├── 📄 .gitignore
│
├── 📂 firmware/                  # Microcontroller code
│   ├── 📂 esp32_cam/             # Vision Node (ESP32-CAM)
│   │   └── esp32_cam.ino
│   ├── 📂 esp32_contentor/       # Container Node (ESP32-WROOM)
│   │   └── esp32_contentor.ino
│   └── 📂 arduino_interface/     # Interface Node (Arduino R4)
│       └── arduino_interface.ino
│
├── 📂 gateway/                   # Raspberry Pi software
│   ├── 📄 main.py                # Entry point
│   ├── 📄 requirements.txt
│   ├── 📄 .env.example
│   ├── 📂 classifier/            # Gemini AI pipeline
│   ├── 📂 mqtt/                  # MQTT client & handlers
│   ├── 📂 database/              # SQLite schema & queries
│   └── 📂 web/                   # Dashboard (REST + WebSocket)
│
├── 📂 3d_models/                 # 3D printing files
│   ├── 📂 stl/                   # Print-ready STL files
│   └── 📂 source/                # Editable design files
│
├── 📂 docs/                      # Documentation
│   ├── 📂 assets/                # Images, diagrams
│   ├── 📂 wiring/                # Wiring diagrams
│   └── 📄 api.md                 # API documentation
│
├── 📂 .github/                   # GitHub automation
│   ├── 📂 workflows/             # CI/CD pipelines
│   ├── 📂 ISSUE_TEMPLATE/        # Issue templates
│   └── 📄 PULL_REQUEST_TEMPLATE.md
│
└── 📂 tests/                     # Test files

📋 Roadmap

  • Project proposal approved
  • Bill of materials defined
  • Hardware assembly & wiring
  • ESP32-CAM firmware (Vision Node)
  • ESP32-WROOM firmware (Container Node)
  • Arduino R4 firmware (Interface Node)
  • Raspberry Pi gateway software
  • Gemini AI classification pipeline
  • MQTT communication layer
  • SQLite database & event logging
  • Web dashboard (real-time)
  • 3D printed enclosure
  • System integration & testing
  • Documentation & demo video

🤝 Contributing

This is an academic project, but suggestions and feedback are welcome! Feel free to open an issue or submit a pull request.

📄 License

This project is licensed under the MIT License — see the LICENSE file for details.

👤 Author

Alexandru Tutunaru
🎓 LESTI — Universidade do Algarve (UAlg)
📧 [Your Email]
🔗 [Your LinkedIn]
🐙 GitHub

Built with ❤️ for Laboratório IoT @ UAlg

About

ECOBIN - AI-powered smart waste sorting bin with computer vision (Gemini 2.0), MQTT, ESP32, Raspberry Pi. IoT Lab project @ UAlg

Topics

mqtt raspberry-pi iot arduino computer-vision esp32 3d-printing smart-bin gemini-api waste-classification

Resources

Readme

License

MIT license
Activity

Stars

0 stars

Watchers

0 watching

Forks

0 forks
Report repository

Releases

No releases published

Packages

 
 
 

Contributors 1

Languages