Greener Future Through Vision AI and Ultralytics YOLO

‍TrashBestie is a new app that helps us sort and manage waste in a different and better way using computer vision. TrashBestie uses deep learning and advanced technology to help people take action to make the planet cleaner and more sustainable.

The team behind TrashBestie envisions a future where waste is no longer a nuisance but an opportunity for positive change. Sorting waste is important for protecting the environment, saving resources, and reducing pollution. With this in mind, TrashBestie became the digital solution that empowers individuals to make informed waste management decisions effortlessly. The goal is clear: inspire a collective movement toward responsible waste management and foster a cleaner planet for generations to come.

Meet the Team Behind TrashBestie

Before we dive into the innovative technology behind TrashBestie, let's meet its creators:

• Helge Rölleke: Experienced in healthcare sales, Helge transitioned to data science and conducted groundbreaking research on company performance and executive compensation. He's also a mushroom enthusiast and open to new data science opportunities.
• My: A Data Scientist and Frontend Developer who combines skills to tackle complex challenges and create user-friendly web applications.
• Simantini Shinde: A Junior Data Scientist with expertise in data analysis, machine learning, and more. Simantini is a strong advocate for open-source development who constantly explores new technologies and pursues a balanced, sustainable lifestyle.
  

The Journey to Machine Learning and Vision AI

Helge began studying machine learning during his master's thesis, examining how a manager's pay relates to a company's success. This involved using regression models and machine learning techniques. Helge was able to dive deeper into the world of vision AI at Spiced Academy's Bootcamp. Here, he experimented with deep learning and determined the usefulness of Ultralytics YOLO models.

My had a friend who shared his data science projects, which sparked her interest in machine learning. The way data could uncover insights and optimize processes fascinated her. That's why she joined the Bootcamp, where she met Simantini and Helge.

Simantini started exploring machine learning during her master's thesis. She discovered its potential in her field of work, which involves assessing building damages caused by earthquakes. Following her graduation, Simanti had different jobs involving data. These jobs eventually led her to a data science bootcamp and piqued her interest in ML and vision AI.

Choosing Ultralytics YOLO for TrashBestie

TrashBestie's use of Ultralytics YOLOv8 as the primary tool is strategic.

• User-Friendly: Because YOLOv8 is open-sourced and easy to use, it was highly accessible to the team.
• Accuracy: YOLOv8 provided better accuracy, especially in precision scores.
• Flexibility: The team could integrate YOLOv8 seamlessly with Roboflow, enhancing their workflow.
  

How Does TrashBestie Work?

TrashBestie operates as a personal waste-sorting assistant, using artificial intelligence to simplify the process into four straightforward steps:

1. Detect with Your Camera. Use your device's camera to capture an image of the waste item you're unsure how to dispose of.
2. Instant Recognition. Giving credit to YOLOv8's image recognition technology, the app can quickly analyze images and identify different types of waste.
3. Educational Insights. TrashBestie doesn't stop at recommendations. This tool provides educational insights to users to understand suggested waste disposal methods. In turn, this promotes long-term learning and conscious waste disposal habits.
4. Easy to Use and Accessible. The app is user-friendly and accessible to all, making environmentally responsible waste sorting achievable for anyone using an Android device.

Try it out

Building TrashBestie

The development journey of TrashBestie involves a series of crucial steps:

1. Labeling and Annotation. Images are carefully labeled and annotated using tools like Roboflow to create a robust dataset for training.
2. Exporting the Dataset. After exporting the labeled dataset, the object detection dataset is ready for training.
3. Training with YOLOv8. The YOLOv8 model is trained on the exported dataset, focusing on fine-tuning its parameters for improved object detection accuracy.
4. Streamlit Deployment. The YOLOv8 model is integrated into the Streamlit application, ensuring efficient and accurate object detection. This app is hosted on GitHub using YOLOv8 and Streamlit for object detection and tracking.
   

The Future of TrashBestie

TrashBestie is continuing to improve by adding localization, making it more accessible on iOS and Android, and refining image processing techniques. The team is committed to continually improving the app's performance and precision.

Check out their project on Devpost, which includes an image gallery and a YouTube video showcasing the details of their work.

TrashBestie is on a mission to revolutionize waste management and make our planet cleaner and more sustainable. This is a first step into the future, that could even revolutionize the conception of waste management careers. Join them on this exciting journey toward a greener future!

‍

Get in touch with the TrashBestie team:

Helge: LinkedIn, GitHub

Simantini: LinkedIn, GitHub, Medium

My: LinkedIn