Urban Space

Home
Features
Map for map city
Vacant space detection ai
Vacant space detection: Results
Visualizations

Inspiration

It all started with addressing a crucial issue in real life. When we look for places to jog, go on morning walks, or find relaxing areas, we find that most people in urban areas do these activities on the road! They run in a lane, making sure no vehicle will hit them. Brainstorming why this is happening, we discovered that there are very few green spaces in cities, and most of them are located offsite from the city centre. In other words, these green spaces are not easily accessible for daily purposes as they are hard to reach. Our goal is to transform vacant urban spaces into vibrant green havens, bringing nature closer to where people live and promoting a healthier, more sustainable urban environment.

What it does

Based on the user's location, Urban space gives map visualizations. Urban Space performs vacant space detection on a city map using advanced vision techniques to list the top suitable vacant spaces. It determines the best vacant space based on accessibility, potential impact, and suitable geographic conditions. Urban Space gives suggestions for setting up green spaces and the required infrastructure details to build for creating green spaces. Additionally, users can upload images of specific land areas, and our platform generates visual simulations of how the green space would look after implementation.

How we built it

Urban Space is a website built using Flask. During the development process, we utilized Streamlit for testing purposes.

Explore city: We integrated the Geopy library to obtain latitude and longitude information based on the user's input and Maptiler API to display map visualizations for it.
Vacant Space Detection: We utilized generative AI and vision techniques to identify vacant spaces within urban areas. We leveraged the vision capabilities of a Gemini LLM, to process images along with structured prompts. This allowed us to detect vacant or underutilized spaces in the city. Based on the analysis, we provided recommendations on the most suitable options for transforming these vacant areas into green spaces.
Planner for generating Green Space: • We performed prompt tuning on the Gemini LLM to specialize it for the specific use case of understanding urban conditions, analyzing vacant spaces, and generating plans for creating green spaces. • This prompt tuning process enabled the LLM to become highly proficient in this domain. • The tuned LLM identified suitable vacant areas for developing green spaces, provided comprehensive accessibility information, such as available travel options to reach the proposed green spaces from various sub-regions within the city. Additionally, the LLM analyzed and reported on the potential environmental impact of transforming these vacant areas into green spaces.
Visualizations: We fine-tuned the Controlnet Stable Diffusion model to get green space visuals on vacant areas in the city.

Challenges we ran into

• Generating realistic visualizations of proposed green spaces that aligned with the existing surroundings and incorporated various design elements was a complex task. • Deploying the user interface (UI) with visualization features required at least a T4 GPU, which we currently do not have access to.

Accomplishments that we're proud of

We are proud to have developed a platform that addresses a significant urban issue by promoting the creation of green spaces within cities. Our integration of various technologies, including machine learning, APIs, and visualization techniques, along with a minimal UI, has enabled this solution.

What we learned

This project allowed us to gain invaluable experience in working with geospatial data, machine learning models, and web development frameworks. We learned how to effectively integrate different technologies to develop a comprehensive solution. Gaining an understanding of the problem statement and domain knowledge has educated us on the importance and necessity of green spaces and renewable energy initiatives, encouraging us to take a step towards environmental sustainability for better humanity!

What's next for Urban Space

Looking ahead, we plan to enhance our platform's accuracy in identifying vacant spaces and generating more detailed and context-aware plans for green space development. We aim to incorporate additional data sources, such as zoning regulations and environmental factors, to provide a more feasible solution. • Feature - To build a sustainable green space platform, here users will allot a space for gardening & caretaking that will spotlight them in the community, to attend eco-friendly seminars or workshops hosted at Green Space, adding gamification for high engagement. Furthermore, we intend to connect with more subject experts to refine this solution, find opportunities to scale this application and provide automation to government to implement this