🗑️ ESP32-CAM Waste Classification System

An Edge AI-powered waste classification system using ESP32-CAM with on-device machine learning inference. Features a retro-styled Arcade dashboard for real-time monitoring with live video streaming and prediction visualization.

📸 Dashboard Preview

🎮 Retro Arcade-style Dashboard with live camera feed and real-time predictions

[Add your dashboard screenshot here]

✨ Features

🎯 Edge AI Classification

9 Waste Categories: Battery, Biological, Cardboard, Clothes, Glass, Metal, Paper, Plastic, Shoe
On-Device Inference: ML model runs directly on ESP32-CAM (no cloud required)
60% Confidence Threshold: Only high-confidence predictions are reported
3-Second Inference Interval: Balanced between responsiveness and performance

📹 Live Video Streaming

MJPEG Stream: Real-time video from ESP32-CAM at 320x240
Snapshot Endpoint: On-demand image capture
Low Latency: Direct camera-to-dashboard streaming

🎮 Arcade Dashboard

Retro Gaming Aesthetic: CRT scanline effects, neon colors, pixel-style fonts
Live Camera Feed: Embedded video stream with overlay
Real-time Predictions: Animated category display with confidence bars
Statistics Panel: Total counts, category distribution, session stats
Prediction History: Color-coded scrolling list of recent detections
Connection Status: Visual indicators for ESP32 and stream connectivity

🌐 System Features

WiFi Auto-Reconnection: Robust connectivity handling
HTTP API: RESTful endpoints for predictions and status
CSV Logging: Automatic prediction logging for analysis
Serial Commands: Runtime control via serial monitor
Status LED: Visual feedback on ESP32-CAM

🏗️ System Architecture

┌─────────────────────────────────────────────────────────────────┐
│                        ESP32-CAM                                │
│  ┌─────────────┐  ┌─────────────┐  ┌─────────────────────────┐  │
│  │   Camera    │→ │  Edge AI    │→ │   Async Web Server      │  │
│  │  OV2640     │  │  Inference  │  │  - /stream (MJPEG)      │  │
│  └─────────────┘  └─────────────┘  │  - /snapshot (JPEG)     │  │
│                                    │  - /status (JSON)       │  │
│                                    └───────────┬─────────────┘  │
└────────────────────────────────────────────────┼────────────────┘
                                                 │
                        WiFi Network             │
                                                 │
┌────────────────────────────────────────────────┼────────────────┐
│                   Arcade Dashboard (Python)    │                │
│  ┌─────────────────┐  ┌───────────────────────▼──────────────┐  │
│  │  HTTP Server    │  │      Video Thread                    │  │
│  │  (Port 5000)    │  │  - MJPEG Stream Consumer             │  │
│  │  - Predictions  │  │  - Frame Queue Management            │  │
│  └────────┬────────┘  └──────────────────────────────────────┘  │
│           │                                                     │
│  ┌────────▼────────────────────────────────────────────────┐    │
│  │              Arcade Game Window                          │    │
│  │  - Live Video Panel    - Stats Panel                     │    │
│  │  - Prediction Display  - History List                    │    │
│  │  - CRT Effects         - Connection Status               │    │
│  └──────────────────────────────────────────────────────────┘    │
└──────────────────────────────────────────────────────────────────┘

🚀 Quick Start

Prerequisites

Component	Requirement
Hardware	ESP32-CAM (AI-Thinker), USB-to-Serial adapter
Software	Python 3.8+, PlatformIO (VS Code extension)
Network	2.4GHz WiFi (ESP32 doesn't support 5GHz)

1️⃣ Clone Repository

git clone https://github.com/SarangPratap/Waste_Classification.git
cd Waste_Classification

2️⃣ Configure ESP32-CAM

Edit WiFi and backend settings in src/main.cpp:

#define WIFI_SSID "Your_WiFi_Name"
#define WIFI_PASSWORD "Your_WiFi_Password"
#define BACKEND_HOST "192.168.1.100"  // Your PC's IP address
#define BACKEND_PORT 5000

3️⃣ Flash ESP32-CAM

# Using PlatformIO CLI
pio run --target upload

# Monitor serial output (115200 baud)
pio device monitor -b 115200

Note: After uploading, press the reset button on your ESP32-CAM.

4️⃣ Setup Dashboard

cd dashboard
pip install -r requirements_arcade.txt

5️⃣ Configure Dashboard

Edit dashboard/config.json:

{
  "esp32_ip": "YOUR_ESP32_IP",
  "stream_port": 80,
  "http_port": 5000,
  "confidence_threshold": 0.6,
  "window_width": 1400,
  "window_height": 900,
  "enable_scanlines": true,
  "max_history": 20
}

Tip: Find ESP32's IP from the Serial Monitor output after WiFi connects.

6️⃣ Run Dashboard

cd dashboard
python arcade_dashboard.py

🎉 Done! The dashboard will open with live video and predictions.

🎨 Waste Categories

Category	Icon	Color	Examples
Battery	🔋	Gold	AA/AAA batteries, power cells, lithium batteries
Biological	🍎	Lime Green	Food waste, fruit peels, organic matter
Cardboard	📦	Brown	Boxes, packaging, corrugated cardboard
Clothes	👕	Purple	Textiles, fabrics, garments
Glass	🫙	Turquoise	Bottles, jars, glass containers
Metal	🔩	Gray	Cans, foil, metal objects, screws
Paper	📄	Sky Blue	Documents, newspapers, notebooks
Plastic	🍾	Blue	Bottles, containers, plastic bags
Shoe	👟	Orange	Footwear, sneakers, sandals

📁 Project Structure

Waste_Classification/
├── 📄 platformio.ini           # PlatformIO build configuration
├── 📄 README.md                # This file
│
├── 📂 src/
│   └── main.cpp                # ESP32-CAM firmware (WiFi, Camera, ML, Web Server)
│
├── 📂 include/
│   ├── config.h                # Hardware configuration
│   └── camera_pins.h           # ESP32-CAM pin definitions
│
├── 📂 lib/
│   └── Waste_classification_inferencing/  # Edge Impulse ML model
│       ├── src/
│       │   ├── edge-impulse-sdk/          # EI SDK
│       │   ├── model-parameters/          # Model config
│       │   └── tflite-model/              # TensorFlow Lite model
│       └── examples/                      # Platform examples
│
├── 📂 dashboard/
│   ├── arcade_dashboard.py     # Python Arcade dashboard application
│   ├── config.json             # Dashboard configuration
│   ├── requirements_arcade.txt # Python dependencies
│   ├── predictions_log.csv     # Logged predictions (auto-generated)
│   └── dashboard.log           # Application logs (auto-generated)
│
├── 📂 docs/
│   ├── SETUP.md                # Detailed setup guide
│   ├── API.md                  # API documentation
│   ├── EDGE_IMPULSE_INTEGRATION.md  # ML model integration guide
│   └── WIFI_SETUP.md           # WiFi troubleshooting
│
└── 📂 test/                    # Test files

🔌 API Reference

ESP32-CAM Endpoints

Endpoint	Method	Description
`/stream`	GET	MJPEG video stream
`/snapshot`	GET	Single JPEG image
`/status`	GET	System status (JSON)

Dashboard HTTP Server (Port 5000)

Endpoint	Method	Description
`/api/prediction`	POST	Receive prediction from ESP32

Example: Status Response

{
  "status": "running",
  "wifi": true,
  "ip": "192.168.1.50",
  "category": "plastic",
  "confidence": 0.87
}

⚙️ Configuration Reference

ESP32 Settings (`src/main.cpp`)

Define	Default	Description
`WIFI_SSID`	-	Your WiFi network name
`WIFI_PASSWORD`	-	Your WiFi password
`BACKEND_HOST`	-	Dashboard PC's IP address
`BACKEND_PORT`	5000	Dashboard HTTP port
`CONFIDENCE_THRESHOLD`	0.6	Minimum confidence (0.0-1.0)
`INFERENCE_INTERVAL`	3000	Time between inferences (ms)
`CAMERA_QUALITY`	12	JPEG quality (0-63, lower=better)

Dashboard Settings (`dashboard/config.json`)

Key	Default	Description
`esp32_ip`	-	ESP32-CAM's IP address
`stream_port`	80	ESP32 web server port
`http_port`	5000	Dashboard prediction server port
`confidence_threshold`	0.6	Display threshold
`window_width`	1400	Dashboard window width
`window_height`	900	Dashboard window height
`enable_scanlines`	true	CRT scanline effect
`max_history`	20	Prediction history length

🔧 Serial Commands

Control ESP32-CAM via Serial Monitor (115200 baud):

Command	Description
`pause`	Pause ML inference
`resume`	Resume ML inference
`status`	Show system status
`reset`	Restart ESP32-CAM

🐛 Troubleshooting

ESP32-CAM Issues

Problem	Solution
WiFi won't connect	Verify credentials, ensure 2.4GHz network, check power supply (500mA min)
Camera init failed	Check PSRAM, verify pin connections, try lower quality setting
Brownout detected	Use better power supply, add capacitor, reduce WiFi transmit power
Upload fails	Connect GPIO0 to GND during upload, use shorter USB cable

Dashboard Issues

Problem	Solution
Can't connect to stream	Verify ESP32 IP in config.json, ensure same WiFi network
Stream disconnects	Check WiFi signal strength, reduce inference interval
Window won't open	Install/update graphics drivers, check arcade installation
Unicode errors in logs	Windows console issue, logs still work in file

Network Issues

Problem	Solution
Predictions not received	Check `BACKEND_HOST` matches your PC's IP
High latency	Reduce `CAMERA_QUALITY`, increase `INFERENCE_INTERVAL`
Connection refused	Ensure dashboard is running before ESP32 sends predictions

🛠️ Hardware Requirements

Required

ESP32-CAM (AI-Thinker module with OV2640 camera)
USB-to-Serial Adapter (FTDI FT232RL or CP2102)

🔮 Future Enhancements

🤝 Contributing

Contributions are welcome! Please:

Fork the repository
Create a feature branch (git checkout -b feature/AmazingFeature)
Commit your changes (git commit -m 'Add AmazingFeature')
Push to branch (git push origin feature/AmazingFeature)
Open a Pull Request

📝 License

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

👨‍💻 Author

Sarang Pratap - GitHub

🙏 Acknowledgments

Edge Impulse - ML model training and deployment
Espressif - ESP32 platform
PlatformIO - Development environment
Arcade Library - Python game framework for dashboard

⭐ Support

If you find this project useful, please give it a ⭐ on GitHub!

Found a bug? Open an issue

Made with ❤️ for a cleaner planet 🌍

Reduce • Reuse • Recycle • Classify

Name		Name	Last commit message	Last commit date
Latest commit History 29 Commits
assests		assests
dashboard		dashboard
docs		docs
include		include
lib		lib
src		src
test		test
.gitignore		.gitignore
README.md		README.md
_codeql_detected_source_root		_codeql_detected_source_root
platformio.ini		platformio.ini

SarangPratap/Waste_Classification

Folders and files

Latest commit

History

Repository files navigation