Believe it or not, CarbonVision began as the result of a debate among students in our living room. Will and Kunga had conflicting opinions on what countries produced more carbon emissions. Will’s idea was countries that had high exports of consumer goods produced high carbon emissions whilst Kunga argued that the production of capital goods created much more carbon emissions due to its high demand. Soon Asher walked in and acted as the arbiter of the argument. As we talked about world challenges, a really interesting concept began to form.
Will noted that CO2 emissions, which signal major environmental changes, could be thought of as the Earth's pulse. Expanding on this idea, Kunga proposed that there could be a direct correlation between industrial activity and economic activity that could be utilized, given that industrial operations release a significant amount of CO2. Asher suggested developing a platform that would be very helpful to economists and policymakers in predicting economic patterns based on CO2 emissions.
This marked the inception of CarbonVision.
Our objective for CarbonVision was to create a revolutionary platform that uses CO2 emissions from major production sites like ports, factories, and metropolitan centers to forecast economic patterns in a given nation. Utilizing cutting-edge AI algorithms and real-time satellite data, our goal was to convert CO2 levels into useful economic insights. Through the prism of environmental data, users would be able to track indications of GDP growth, trade volumes, and industrial activity.
We started our trip by compiling CO2 emissions data from datasets like CMS_CTL_NA_TCCON and Sentinel-5P, as well as from satellites. Asher took the lead on applying Python packages such as sklearn, geopandas, and joblib to handle this geospatial data. He put in several hours making sure the data was clear, standardized, and correctly georeferenced to correspond with ports, cities, and industrial zones.
Will, meanwhile, concentrated on fusing economic and environmental data. We gathered trade volumes, historical GDP numbers, and industrial production indexes from reputable sources such as the World Bank. Will carefully matched these economic indicators with the time-related components of CO2 emissions, an extremely difficult undertaking.
Kunga was in charge of the models for machine learning. He tried algorithms like Random Forests and Gradient Boosting Machines using scikit-learn to see which one would work best for our forecasting requirements. In order to improve the model's accuracy, feature engineering was essential. We took into account temporal alignments, spatial distributions, and outside variables like weather patterns.
Because Flask is so simple and flexible, we decided to use it for the backend when constructing the platform itself. While Will and Kunga worked together on the frontend using HTML, CSS, and JavaScript, Asher created APIs to deliver processed data and predictions.
There were further difficulties with the deployment of CarbonVision. Because of PostMan’s scalability and dependability, we chose it. Using GitHub Actions to set up pipelines for Continuous Integration and Deployment guaranteed that our code would stay clean and that updates could be distributed without hiccups.
Constructing CarbonVision wasn't without difficulties. Processing the enormous volumes of satellite data needed optimization and, occasionally, pure willpower. We frequently found ourselves working late into the night, straining our laptops under the demand. It was difficult to match real-time CO2 emissions data with economic indicators because our computeres frequently lagged and required resourceful techniques. It took several iterations for our machine learning models to find the right mix between avoiding overfitting and retaining predictive power. Our debugging abilities were put to the test by making sure the frontend and backend communicated flawlessly. Managing this large-scale endeavor in addition to academics required careful planning and steadfast dedication.
Despite the difficulties, we accomplished a number of goals that made us proud. With the help of CO2 emissions, we were able to create a functional prototype of CarbonVision that offers precise economic projections. Through our interdisciplinary teamwork, we were able to combine our various areas of expertise into a well-thought-out project. When we presented CarbonVision to our friends and clubmates, we received positive feedback, which confirmed our diligence and commitment. We all developed tremendously, going above and beyond our comfort zones in the classroom and picking up new abilities in the process.
We learned priceless lessons from the journey. Our understanding of how environmental data can be used as a stand-in for economic activity has grown. Good task delegation and communication were crucial; we also learnt how to capitalize on one another's advantages and help one another overcome obstacles. Overcoming challenges strengthened our resilience and problem-solving skills. The initiative served as evidence of the effectiveness of interdisciplinary methods in solving practical issues.
We are only getting started on this journey with CarbonVision. We are eager to add streaming data capabilities so that users can receive real-time updates and increase the platform's dynamic nature. We plan to enhance the insights we provide by adding other environmental indicators, such as NO₂ levels and socioeconomic data. We are creating dashboards that users may customize to the way they want the platform to work better for them. Additionally, a mobile application to enable CarbonVision's mobility is being developed. In order to implement CarbonVision more broadly and support global sustainability initiatives, we are excited to collaborate with federal and international organizations. Another thing we are passionate about is educating others about the
Looking back on our journey, we are incredibly proud of the work that we have produced as a team. From humble beginnings as a concept between three friends, CarbonVision has developed into a technology that has the potential to significantly alter how economies perceive and react to environmental data. We can't wait to keep improving this platform and taking on the new challenges and opportunities that lie ahead.
Log in or sign up for Devpost to join the conversation.