Trash 'n' Dash

Logo
Architecture

What Inspired Us

Technology will lead to a sustainable future. The EPA estimated that the United States generated 292.4 million tons of municipal waste in 2018. 25% of the recyclable waste does not actually undergo recycling due to contamination with other types of waste. Team members Manvitha Kalicheti, Aishwarya Sheelvant, and Nemath Ahmed have seen this problem firsthand in India. India generates 26,000 tons of plastic. According to India Times, waste is only 10-15% recycled. Waste Segregation is where the solution lies.

Parent Idea

Trash ‘n’ Dash started in Spencer Tate’s Dorm Room 2 years ago. Spencer realized a small percentage of his classmates knew how to recycle. He then bought a 64-gallon recycling can and a 64-gallon trash can. Spencer would verbally tell his customers during pick-up what they had recycled incorrectly during the Spring of 2020. This strategy caught the attention of other students who liked the convenience of someone picking up their trash and recycling from their door. With over 250 students using Spencer's trash valet service now, he had to change his education strategy to meet demand.

During the Fall of 2021, the team iterated on their previous product. Trash ‘n’ Dash added clear laminated labels to all customer recycling bins. Instead of verbally telling customers what they had done wrong, the team would write the recycling errors in each student bin during trash collection. During the first three weeks, we provided no writing on cans. Almost everyone recycled incorrectly! We wrote on the containers for the remainder of the semester, and recycling behavior improved dramatically. We tracked everything by collecting data on six metrics to track behavior. After empirically showing Trash ‘n’ Dash's effect, Spencer turned to the university to expand his service. To grow, additional funding was needed to operationalize to a larger scale. After speaking with His university, they were not interested in integrating the service, which led to the end of Trash ‘n’ Dash. Spencer learned that this problem is prevalent for people of all ages, and habits will change when positive reinforcement occurs during the educational process.

Waste has never been at higher levels worldwide. With AI advancing and integrating into our everyday lives, there is no better time to create a transformational waste system that reinforces the user on what they can and cannot recycle at the point of use.

How We Built The Project

Trash ‘n’ Dash integrates a small camera into a waste bin system and classifies the object and what type of waste bin the trash goes in.

We used a YOLOv5 model to classify trash into six different categories. The YOLOmodel was deployed on AWS Sagemaker after training the model specifically for waste. From a live video stream, upon detecting an arm holding trash using a media pipe, we snapshot five frames and use our model to get confidence scores for each class for each image. We sum up these scores to find the highest-scoring class - this is our final prediction of the trash class. We implemented a visual guideline to direct the trash to the correct bin using Streamlit. We used MongoDB Atlas to save the data from integrating it with PowerBI to perform other analytics efficiently.

Challenges

The main challenge was limited computational power. The deep learning model requires a large amount of memory and processing power. Unfortunately, our computer could not handle the data to train the model. This forced us to devise alternative solutions, such as using a cloud-based service to perform the training.

We also needed help accessing the hardware components for our project. We had planned to use sensors to detect and measure the distance between objects and the camera. However, access to the lab room where these sensors were stored was restricted, leaving us with limited options. As a result, we had to pivot and use computer vision techniques to identify and capture photos of objects at a specific distance from the camera. Finally, our project proposal used an intelligent bulb to indicate the trash category. However, we soon discovered that the smart bulb did not have support for us to interact with it through an API, making it impossible to control the bulb from our code. This forced us to change our approach and develop a web application to display the results instead. This significantly deviated from our original plan and required us to learn new technologies and programming languages. Overall, these challenges made the hackathon a valuable learning experience, forcing us to be creative and resourceful in finding solutions to the problems we encountered.

Accomplishments

We are proud of our significant achievements as part of this task. First and foremost, we take pride in successfully incorporating various tools and technologies into our work process, enabling us to keep up with trends and develop cutting-edge solutions.

We are also proud of our capacity to overcome obstacles encountered during the hackathon, such as the RAM restrictions on our computers. We came up with innovative ideas that allowed us to continue progressing with our project by integrating computer vision and creating a web app. We are pleased with the potential influence that our solution may have on reducing waste and increasing recycling rates globally. We seek to motivate others to take action and contribute to building a more sustainable future by showcasing our approach's financial and environmental advantages. Overall, we are pleased with our development on this project and want to keep growing and improving our answer in the future.

In addition to the benefits of integrating tools and technologies into our work process, the hackathon also taught us several new skills and techniques. We learned how to effectively use computer vision to identify and capture photos of objects, which was a new challenge for us. We also gained experience developing web applications outside our normal scope of work. This allowed us to expand our skill set and increase our technical knowledge.

Furthermore, the hackathon allowed us to work under time pressure and solve real-world problems. This allowed us to develop our problem-solving skills and learn how to approach complex challenges in a structured and efficient manner. By working on a project with limited resources and facing unforeseen difficulties, we developed our ability to be flexible and adaptable. Overall, the hackathon was a valuable learning experience and allowed us to understand our project's potential as a business model. By combining technical knowledge with business acumen, we have equipped ourselves with the skills necessary to bring our project to market and create a successful and profitable business.

Future Work

The goal is to increase the amount of recyclable waste in each country that is actually recycled. Recycling has been shown to create more jobs and revenue compared to incineration or landfilling. For example, recycling 10,000 tons of waste creates 36 jobs, while incinerating or landfilling the exact amount only makes 1 or 6 positions. We plan on integrating sensors instead of models for object detection. This would drastically reduce the cost of build and computation. It would be easier to install without much initial investment.

To make the project more scalable, we also plan to develop a cloud-based platform that will allow us to store and analyze the data collected by the sensors. This platform will also enable us to remotely monitor the system's performance and quickly respond to any issues that may arise.

To make this project more attractive to investors, we plan to demonstrate its potential for reducing waste and increasing recycling rates and its positive economic impact. We aim to add a feedback loop that can add authentic images captured back into the database so we can improve the model with time. These images would be added only when we detect an object with a high confidence score. We also plan to conduct a comprehensive market analysis to identify key industry trends and target markets for our solution. In conclusion, by improving our technology's accuracy and reliability and developing a scalable and cost-effective solution, we aim to make this project a successful and investable business model that will help reduce waste and increase recycling rates worldwide.