Envirolyst

Landing Page
Map Selection
Smart Dashboard
AI Recomendation
Communities
Tech Stack

Inspiration

Indonesia is a country rich in natural resources and green landscapes, but business practices that neglect environmental sustainability have accelerated ecosystem degradation, deforestation, and regional pollution. Many industries continue to expand without considering the long-term environmental impact, resulting in declining air, water, and soil quality. Envirolyst is built to address this issue by providing satellite-based environmental analysis to support responsible and sustainable decision-making.

What it does

Envirolyst analyzes satellite imagery submitted by users to identify land usage patterns within a specific area. Using AI-based image segmentation, the app classifies regions such as vegetation, water bodies, built-up zones, and bare land. To provide deeper environmental context, Envirolyst also integrates real-time air pollution and weather data relevant to the selected location. After collecting and processing this data, a Large Language Model (LLM) is used to generate smart, context-aware sustainability recommendations tailored to the unique characteristics of the area.

Key Features

🛰️ AI-Powered Land Use Classification
Detects and classifies land types (vegetation, water, buildings, etc.) from satellite imagery using deep learning segmentation models.
🌫️ Air Pollution Data Integration
Gathers up-to-date air quality index (AQI) and pollutant levels to assess environmental risks in the selected area.
🌦️ Weather Data Retrieval
Adds current weather conditions to improve the relevance of the recommendations.
💡 LLM-Generated Sustainability Recommendations
Uses a Large Language Model to provide insightful, adaptive suggestions based on combined image and environmental data.
🗺️ Interactive Map Interface
Allows users to freely select areas from a map, trigger analysis, and receive a detailed sustainability report.

How we built it

Frontend: React + Vite, Google Maps API, Tailwind CSS Backend: Flask, Groq API, OpenWeather API, Hugging Face

We used React as the main frontend framework due to its component-based architecture, which allows for modular development and easier maintenance. To speed up the UI styling process, we integrated Tailwind CSS, enabling us to rapidly build responsive and clean interfaces. For geolocation and map interaction, we chose Google Maps API because of its polished visuals, excellent developer documentation, and rich event listener capabilities that allow us to build a highly interactive mapping experience.

On the backend, we chose Flask because it works well with Python-based AI pipelines. Flask is responsible for retrieving the satellite imagery and running it through an open-source image segmentation model from Hugging Face, which identifies land usage types such as vegetation, water, built-up areas, and more. We then query the OpenWeather API to retrieve current weather and air quality data for the selected location.

Finally, all the processed data—segmentation results, weather, and pollution context—is passed to an LLM via Groq API, which generates intelligent, context-aware sustainability recommendations tailored to the environmental conditions of the selected area.bout the area. Then all the data will be given to LLM to give the recomendation based on the situation.

Challenges we ran into

We ran into some a few challenge when using the flask to integrate the models, because the model itself is not well documented and can't be modified, so there will be some accuracy issue about the image segmentation models.

Accomplishments that we're proud of

Integration of AI for web application (this was never done before from our team 😁)
A lot of research has been done to processing satellite imagery (image processing, unsupervised learning, using pre-trained models)
Very efficient and effective teamwork

What we learned

We learned to present data and make it more interesting for overall users
We learned a lot about image processing and image segmentation with various method