planetree

Inspiration

Our inspiration stems from a profound commitment to environmental conservation and the urgent need to address the escalating crisis of deforestation. Witnessing the widespread destruction of forests globally, we recognized the potential of technology to not only raise awareness but also empower individuals to take direct action.

The idea was sparked by the power of satellite technology, which has transformed our understanding of Earth's landscapes and can provide unequivocal evidence of environmental change. We were particularly moved by stories of communities and small organizations that felt helpless against the massive scale of deforestation. This inspired us to create a tool that could bring together concerned citizens from all over the world, transforming their isolated efforts into a united front.

By merging real-time satellite imagery with interactive mapping and community features, we aim to create a platform that not only educates but actively involves users in monitoring and responding to deforestation. Our goal is to harness collective vigilance and action, making each user a guardian of our planet's forests.

What it does

Our app utilizes cutting-edge satellite technology to provide a dynamic, interactive platform for global users to engage in combating deforestation effectively. Here's a detailed look at its functionalities:

Features

Interactive Earth Map

Explore the World: Users can access a detailed, interactive map on our website, allowing them to zoom, drag, and inspect different regions globally.

Real-Time Satellite Imagery

Scan the Earth: With the simple click of the "scan" button, the app processes satellite snapshots for the selected area, enabling real-time environmental monitoring.

Color-Coded Masking

Visual Analysis: The app applies a color-coded mask over the satellite images, highlighting forested areas in green and deforested areas in red. This feature makes it easy for users to identify and understand the scope of deforestation.

Community Pinpoints

Mark and Share: Users can drop pinpoints on locations where deforestation has been observed. These markers contribute to a shared global map, pinpointing deforestation hotspots and fostering a sense of community and collaborative effort.

Awareness and Action

Empower Change: By collecting and displaying user-generated data and observations, the app not only raises awareness but also promotes collective environmental action. It serves as a platform for advocacy and intervention, providing crucial information on areas that may require immediate or sustained attention.

How We Built It

The construction of our app involved a mix of state-of-the-art web technologies and machine learning frameworks to deliver a seamless, interactive, and impactful platform for monitoring deforestation. Here’s a detailed look at the core components and their integration:

Backend Development with FastAPI

FastAPI Framework: We chose FastAPI for our backend due to its high performance and ease of use for creating APIs. FastAPI is particularly suited for asynchronous tasks and is highly scalable, crucial for handling extensive data and high user traffic.
Database Interactions with PostgreSQL: Our backend manages interactions with a PostgreSQL database, ensuring robust, secure, and scalable data management. We leveraged SQLAlchem as our ORM tool to facilitate database interactions efficiently.

AI Model with PyTorch

PyTorch for Computer Vision: Our custom computer vision model was built using PyTorch, known for its flexibility and efficacy in deep learning tasks. This model processes satellite images to identify deforestation areas, overlaying a color-coded mask to distinguish between forested and deforested zones.
API Integration of Model: The AI model is hosted as a separate microservice and is integrated with the main API. This architecture allows the computationally intensive image processing tasks to be handled without affecting the main application’s performance.

Frontend Development with React

React Framework: We developed the frontend with React, which is renowned for its efficient rendering and state management features. React's component-based structure was ideal for managing the diverse UI elements.
Styled-Components for Styling: We utilized styled-components for CSS-in-JS styling, allowing CSS to be written within React components directly. This method kept the styling modular and easily scalable, especially for dynamic styles based on application state changes.
Interactive Map Integration with HERE Maps: For the mapping functionality, we integrated HERE Maps instead of more conventional alternatives. HERE Maps provided advanced mapping solutions that enhanced our application's ability to let users manipulate maps, zoom, drag, and drop pinpoints with high accuracy and responsiveness.

Challenges We Ran Into

Throughout the rapid development phase of our project, we faced several significant challenges that tested our adaptability and learning speed:

Limited Experience with Computer Vision

Steep Learning Curve: Our team had limited prior experience in the field of computer vision, which required us to quickly come up to speed on the latest practices and technologies in image processing and machine learning. Navigating the complexities of training and integrating a PyTorch model to analyze satellite imagery was particularly daunting but ultimately rewarding as we learned to effectively apply these techniques.

Introduction to MLOps

MLOps Proficiency: Implementing MLOps practices was a new area for us, involving the automation and monitoring of machine learning models in production. The challenge was not just in developing a model but ensuring its scalability, efficiency, and reliability through proper model lifecycle management within the constraints of our project timeline.

Building a Full Application in a Short Time

Time Constraints: Perhaps the most pressing challenge was the time limit we faced. Completing a functioning application with multiple complex components including backend, frontend, and AI integration within just 36 hours required intense focus, efficient task management, and effective teamwork. Balancing feature development with testing and debugging under such tight deadlines tested every aspect of our project management and technical skills.

Accomplishments That We're Proud Of

Despite the various challenges we faced during the hackathon, our team achieved several significant accomplishments that we take great pride in:

Overcoming Technical Challenges

Rapid Learning and Application: We are immensely proud of our ability to quickly learn and apply new technologies in the realms of computer vision and MLOps. This capability not only allowed us to effectively build a sophisticated AI-powered application but also enhanced our technical skill set significantly.

Development of a Comprehensive Business-Product

Holistic Solution: We successfully developed a product that addresses a critical environmental issue with a viable business model. Our application not only serves an environmental purpose but also has the potential for substantial impact, scalability, and engagement across various stakeholders.

Completing the Project On Time

Efficient Project Execution: Completing a fully functional and well-integrated application within the stringent 36-hour deadline is an achievement we hold in high regard. Our ability to manage time effectively, prioritize tasks, and collaborate under pressure was key to delivering the project on schedule.

What We Learned

The hackathon was not only a test of our technical abilities but also a profound learning experience that extended across various dimensions of software development and team collaboration. Here are some of the key learnings:

Technical Skills

Advanced Computer Vision Techniques: We gained hands-on experience with PyTorch and delved deeper into the nuances of building and deploying computer vision models. This has significantly bolstered our understanding of image processing and analysis.
MLOps Implementation: We learned how to integrate MLOps practices to ensure our machine learning models are scalable, maintainable, and reliable in a production environment. This included everything from model training, versioning, to deployment.

Project Management and Collaboration

Agile Development in a Tight Schedule: Managing to develop a full-fledged application in just 36 hours taught us valuable lessons in agile development practices, task prioritization, and time management. We learned to make quick decisions and adapt to changing requirements dynamically.
Effective Team Collaboration: Working in a high-pressure environment emphasized the importance of clear communication and leveraging each team member's strengths. We honed our ability to work as a cohesive unit, ensuring everyone’s input was valued and utilized.

Environmental Awareness

Deeper Understanding of Environmental Issues: By focusing our project on deforestation, we gained a deeper understanding of environmental challenges. This increased our awareness and commitment to using technology for social and environmental impact.

What's Next for Osapiens Case

As we look to the future, our roadmap for the Osapiens app is clear and ambitious. Here are the steps we plan to take to enhance, expand, and ensure the app’s success:

Evaluating Business Models

Sustainable Revenue Streams: We will explore various business models to identify the most sustainable and effective strategy. This includes subscription models, partnerships, and possible government grants, considering our environmental impact and community-driven approach.

Establishing Partnerships

Strategic Alliances: Finding partners who align with our mission will be crucial. This includes environmental NGOs, government bodies, and technology companies that can provide support in terms of resources, expertise, and data.

Finalizing the App

Development and Testing: We will continue to refine the app’s features, focusing on enhancing the user experience and ensuring robust functionality. This includes integrating feedback from initial users to improve the interface and features.

Expanding Data Sources

Obtaining Mosaic GeoTIFF Dataset: To enhance the app's capabilities, we plan to acquire a comprehensive mosaic GeoTIFF dataset. This will allow us to provide users with more detailed and accurate imagery, improving our deforestation detection and mapping features.

Enhancing User Interface

Smoother UI: We aim to redesign the UI to be more intuitive and user-friendly, ensuring that it is accessible to a broader audience and enhances user engagement.

Improving Recognition Capabilities

Enhanced AI Model: We will continue to refine our computer vision model to improve its accuracy and efficiency in detecting deforested areas. This will involve training the model with more diverse datasets and optimizing its algorithms.

Bug Fixes and Optimization

Continuous Improvement: Regular updates and bug fixes will be part of our ongoing development cycle to ensure the app remains reliable and effective.

Identifying and Engaging Target Audience

Marketing and Outreach: Identifying the right target audience and crafting tailored marketing strategies to engage them will be critical. We will focus on users who are activists, educators, researchers, and policy-makers who can leverage our tool for the greatest impact.