Recyclable?

Check recyclability through yours lens and save the world! ♻

Comment

Project Story

about the project

I decided to clean my room the other day, and uncovered bags of random objects and small items that were lost within the piles. Along the way, I picked up draw-string bags from spontaneous events, Amazon package bags, and cups and bottles from small businesses advertising their companies. Amidst the junk, they all seemed to be made of some kind of plastic and I questioned if they were recyclable. Yet, having no resource to directly find the object, I casually threw the rest of it in the trash with a slight hint of guilt that washed away in seconds.

According to the World Bank, an average of 2.01 billion tons of waste is produced every year, with over 33% of it managed in an environmentally unsafe way. The average person can create up to 0.74 kilograms of waste, though it ranges as high as 4.5 kilograms. This is especially prevalent in high-income countries, where just 16 percent of the world's population generates 34% of the earth's waste, equivalent to 683 million tons. If this projection continues, daily per capita waste generation is predicted to increase by 19% by 2050. Though there are existing infrastructures and systems to encourage people to recycle what they can, most do not have clear information on specific instructions, which leads to a failure in utilizing those systems. Thus, our project aimed to create an application that will help people identify what they can recycle within their immediate surroundings, along with simple instructions on where to do so. With a specialized app that will be accessible on the web, people will be able to find and scan objects around them and effectively recycle, reducing global waste by a significant amount.

What it does

The Object Recyclable is a project that aims to help people recycle their waste by detecting objects in a web camera image and giving information about how to recycle it. Users will hold the object up for a scan, where the website will then load various articles of recycling tips for the object. Articles include specific statistics on the object's waste impact on the environment, sectioned articles that explain ways to possibly reuse the object, and a map to show recycling sites. Users will be able to use the sorted data to find the best way for them to recycle.

How we built it

The Recycle? website is developed using Python with Flask framework, OpenCV for web camera function, and APIs discovered at RapidAPI.

Challenges we ran into

Initially, during our app development process, we couldn't configure IP Location API and Google Maps Autocomplete Plus API to retrieve the user's location so that we could suggest nearby recycling centers. We concluded that the case might be due to user's IP address being restricted from access for privacy concerns. Hence, we hard-coded the feature without utilizing the APIs for the app prototype. As included in api_queries.py, we also learned the idea of building that feature using APIs.

Accomplishments that we're proud of

The accessibility with which people of all ages can view and interact with the website interface. We are also happy that our web application had a functioning prototype that allowed users to perform manual article searches for information on recycling of the objects.

What we learned

We learned about utilizing OpenCV as a main way to detect objects in images, and experimented with applying masks and contours to categorize shapes. We also learned about several concepts in Machine Learning such as neural networks, the different types of learning, and combining its outputs with our design. We used APIs like Google Image Search API to request the data of articles, images, and website links to lay out on the application.

What's next for Object Recyclable

Following the development of the application, the first improvement would be to develop a custom Machine Learning model to expand the number of items to detect, as well as refine the accuracy of the detection weights. We can add a larger series of reference pictures into a dataset, allowing the application to more accurately label the scanned image. Furthermore, aside from the common objects that the app can detect now, we can also detect text on the objects such as brands, the materials, and manufacturer to give more detailed instructions on its recyclability to minimize pollution and resource use.

According to the World Bank, an average of 2.01 billion tons of waste is produced every year, with over 33% of it managed in an environmentally unsafe way. The average person can create up to 0.74 kilograms of waste, though it ranges as high as 4.5 kilograms. This is especially prevalent in high-income countries, where just 16 percent of the world's population generates 34% of the earth's waste, equivalent to 683 million tons. If this projection continues, daily per capita waste generation is predicted to increase by 19% by 2050. Though there are existing infrastructures and systems to encourage people to recycle what they can, most do not have clear information on specific instructions, which leads to a failure in utilizing those systems. Thus, our project aimed to create an application that will help people identify what they can recycle within their immediate surroundings, along with simple instructions on where to do so. With a specialized app that will be accessible in all high-income countries, people will be able to find and scan objects around them and effectively recycle, reducing global waste by a significant amount.

What it does

The Object Recyclable is a project that aims to help people recycle their waste by detecting objects in a web camera image and giving information about how to recycle it. Users will hold the object up for a scan, where the website will then load various tabs of instructions on the side. This includes sectioned articles that explain ways to possibly reuse the object and a map to show recycling sites. Users will be able to use the sorted data to find the best way for them to recycle.

How we built it

The website was developed using Python, OpenCV, and the Earth911 API. We set up a detection model using TensorFlow and the cvlib library, and returned a string of the detected object. This is then passed to the API and google search and returns information about the object. The website was designed using the Flask framework with multiple pages, HTML, CSS to set up the layout.

Challenges we ran into

We had trouble incorporating TensorFlow and the package into the project. There were issues where the library would not install locally, so a virtual environment was needed to start the detection function, which lead to delays in the actual application. Another challenge was displaying the cv2 feed onto the website, so we figured out how to loop and display each frame to mimic the video. There were also challenges in using the Flask framework and learning all the syntax, along with setting up the jinja and css layouts.

Accomplishments that we're proud of

We're proud of successfully completing a project that has several unique features that work together to provide a service. The process of learning Flask and honing our web skills is definitely something that will help us. We're also glad that we were able to introduce ourselves to ML in the project, which is something that can be versatile in future projects.

What we learned

We learned about utilizing OpenCV as a main way to detect objects in images, and experimented with applying masks and contours to categorize shapes. We also learned about several concepts in Machine Learning such as neural networks, the different types of learning, and combining its outputs with our design. We used APIs like Earth911 to request information from datasets, along with web scraping articles and website links to lay out on the application.

What's next for Object Recyclable

Following the development of the application, the first improvement would be to develop a custom Machine Learning model to expand the number of items to detect, as well as refine the accuracy of the detection weights. We can add a larger series of reference pictures into a dataset, allowing the application to more accurately label the scanned image. Furthermore, aside from the common objects that the app can detect now, we can also detect text on the objects such as brands, materials, and manufacturers to give more detailed instructions on its recyclability to minimize pollution and resource use.

Built With

Share this project:

Updates