Recycling Reimagined ♻️

---

This project focuses on using image recogonition techinques to identify different types of waste products, helping imporve recycling processes. We utilized the YOLOv9 model for object detection and created a front-end interface using Streamlit to allow live recognition using a web camera.

About

You can find here the link to the recorded presentation

https://projects.lewagon.com/projects/recycling-reimagined

Architecture



Facility representation setup



Manual labeling demo

Features:

• Waste product image recognition using YOLOv9
• Training with both a sourced dataset and custom images.
• Live object detection thorugh web camera on a Streamlit-based website. +Deployed using Docker on the cloud.

Dataset:

• We used a dataset sourced from Roboflow, which contained approximately 7000 images of various waste products, notably cardboard, glass, metal, paper and plastic. Source: https://universe.roboflow.com/image-processing-home-assignment/trash-detection-kfzaq

• Custom Additions: WE manually annotated 100 custom images using Roboflow to further improve the model's recognition capabilites.

• Annotation Tool : Annotation was done using Roboflow.

Model:

We used the latest YOLOv9 model for object detection due to it's high efficiency and accuracy. Our process involved:

1. Training on sourced images from Roboflow. Source: https://universe.roboflow.com/image-processing-home-assignment/trash-detection-kfzaq
2. Updating the model with our custom images to further refine and improve detection accuracy.

Web Interface:

• The front end was built using Streamlit to enable an interactive and user-friendly interface.
• Live Recognition: A webcam setup was integrated into the website to allow users to perform live recognition of waste products.

Deployement:

• The entire application was containerized using Docker, ensuring ease of deployement.
• The project was then hosted on the cloud, making it accessible for live demonstrations and testing.

How to use:

Installation:

1. Clone the repository:

git clone https://github.com/MiyuNori18/Recycling_Reimagined.git

2. Install the dependencies:

pip install -r requirements.txt

Running the Application:

1. Run the YOLOv9 model for waste product recognition:

python main.py

2. Start the streamlit interface

streamlit run app.py

3. Access the live recognition feature through the webcam interface and activate your webcam for real-time detection.

Contributors:

• Anton Wenemoser
• Miyuki Niyungeko