Inspiration
As climate change, urbanization, and extreme weather events increasingly threaten forests, we saw a growing need for tools that help people understand tree health at a local, actionable level. While large amounts of environmental and forestry data exist, it is often inaccessible or difficult to interpret without technical expertise. We wanted to bridge that gap by turning complex environmental data into clear insights that anyone could explore interactively. This motivated us to create GrowWiseAI, a platform that helps users assess tree survivability at specific locations using machine learning.
What it does
GrowWiseAI allows users to select a location on an interactive map, which prompts our platform to analyze the environmental conditions within that area. GrowWise then uses a machine learning model trained on 10,000+ datapoints to predict how optimal the environmental conditions are for tree growth. Results are displayed through an intuitive interface that includes a health score, contributing environmental factors, and a natural-language explanation of the prediction. The user is also able to customize the parameters and simulate different conditions. This enables users to quickly understand potential risks to tree health and make more informed environmental or planning decisions.
How we built it
We built GrowWiseAI as a full-stack application with a clear separation between the frontend and backend. The frontend was developed using HTML/CSS, React, Vite, and Leaflet, providing an interactive map-based interface where users can select locations, adjust parameters, and view results in real time.
The backend was built in Python, where we trained an XGBoost machine learning model using environmental and forest health data. We used Pandas, NumPy, and scikit-learn for data preprocessing, feature scaling, model evaluation, and validation. The trained model, along with its preprocessing components, was serialized using joblib to make it ready for deployment and integration with the frontend.
We used GitHub for version control and collaboration.
Challenges we ran into
Data availability: We spent a lot of time looking for reliable, large-scale datasets directly measuring optimal conditions for tree growth as well as various other environmental factors.
Dataset bias: In some of the datasets, we found that a select number of features had far stronger correlations to the labels than other features, leading to biased predictions and limited generalization.
Accomplishments that we're proud of
Having a working product completed in a short time span. Creating a machine learning model that accurately predicts results.
What we learned
We learned how to implement machine learning algorithms that are designed for categorization tasks such as Random Forest, XG Boost, and Logistic Regression. We learned how to build a data pipeline to bridge data between our APIs, our predictive model, and our frontend framework. We also learned how to create a smooth and interactive UI.
What's next for GrowWise
Greater location coverage. Accounting for more environmental variables. Implementing a future sight option that takes in data from climate/environmental predictions to create forecasts.
Log in or sign up for Devpost to join the conversation.