An autonomous dual-gate waste sorting system powered by YOLOv11, Raspberry Pi, and 8 precision servos.
Designed to identify and sort waste into “Recyclable” or “Landfill” bins in real time.

The Automated YOLOv11 Waste Sorter combines AI-based object detection, embedded hardware control, and mechanical automation.
A custom YOLOv11 model detects items through the Pi Camera, and a Finite State Machine (FSM) controls servo-driven gates and bins to sort waste automatically.

This system ensures continuous, one-by-one item processing using two alternating gates for uninterrupted operation.

• Detects and classifies waste instantly using a custom-trained YOLOv11 model.
• Powered by Ultralytics YOLOv11 and optimized for Raspberry Pi.

• Two alternating gates (Gate 1, Gate 2) ensure non-stop item flow.
• While one gate is closed and detecting, the other prepares for the next item.

• Uses RPi.GPIO for precise hardware control.
• Uses picamera2 for efficient image streaming.
• Lightweight and fully compatible with Raspberry Pi 4/5.

git clone https://github.com/Rie109/Hackathon.git
cd Hackathon

pip install ultralytics opencv-python-headless RPi.GPIO picamera2

If picamera2 fails to install:

sudo apt update
sudo apt install -y python3-picamera2

Place your trained model file best.pt at:

/home/harry/Hackathon/best.pt

Or modify this line in the script:

MODEL_PATH = '/your/path/to/best.pt'

Execute the main script:

python waste_sorter.py

You’ll see this startup menu:

RASPBERRY PI WASTE DETECTION SYSTEM - BALANCED BIN VERSION
===========================================================

1. Run detection (with display)
2. Run detection (headless - no display)
3. Run detection + servo control (with display)
4. Run detection + servo control (headless)
0. Exit

💡 Tip:

• Use 1 or 2 to test camera and detection accuracy.
• Use 3 or 4 for full sorting system with all 8 servos.

Tweak angles and timings at the top of the script:

SERVO_CLOSED_ANGLE = 90
SERVO_OPEN_ANGLE   = 45
GATE_CLOSED_ANGLE  = 90
GATE_OPEN_ANGLE    = 45

bin_open_duration = 2.5  # seconds
clear_wait_time   = 1.0  # seconds

• Current version uses time.sleep() in servo control, causing short camera freeze periods.
• Next iteration: implement non-blocking timers for smoother detection.
• Future goal: add more waste categories and adaptive gate control for improved accuracy.

(Insert a GIF or image of the system in action here)

Sample Output:

[INFO] Gate 1 - Object Detected: ♻️ Recyclable
[INFO] Opening Recyclable Bin...
[INFO] Item Sorted Successfully!
[INFO] Switching to Gate 2...

Bochkovskiy, A., Wang, C. Y., & Liao, H. Y. M. (2024). Ultralytics YOLO11: Redefine what’s possible in AI [Software]. Ultralytics. Retrieved from https://docs.ultralytics.com/models/yolo11/ :contentReference[oaicite:0]{index=0}

Dwyer, B., Nelson, J., Hansen, T., et al. (2025). Roboflow (Version 1.0) [Software]. Retrieved from https://roboflow.com/ :contentReference[oaicite:1]{index=1}

Bradski, G. (2000). “The OpenCV Library.” Dr. Dobb’s Journal of Software Tools. :contentReference[oaicite:2]{index=2}

Croston, B. (n.d.). RPi.GPIO: A Python module to control the GPIO on a Raspberry Pi [Software]. Retrieved from https://sourceforge.net/p/raspberry-gpio-python/wiki/Examples/ :contentReference[oaicite:3]{index=3}

Raspberry Pi Ltd. (n.d.). Picamera2: The libcamera-based Python interface for Raspberry Pi cameras [Software]. Retrieved from https://github.com/raspberrypi/picamera2 :contentReference[oaicite:4]{index=4}