Inspiration
When deciding on an idea for our hackathon submission, GrowthFactor's problem caught our attention.
Our application utilizes funny magic words to identify accessible parking near requested locations. With the ability to scan available slots, retailers are able to save time and resources using our application to gauge foot traffic.
What it does
After developing a segmentation transformer to determine parking lot locations from satellite images, we used zoomed in parking lot locations, a pretrained YOLOv8m model, and Hough line detection to determine how many parking spots are lot has.
How we built it
We used a Python frontend and a React Backend. We trained two models. One model's goal was to find the parking lots, and the other model's goal was to count the parking lots. We trained the models based off of data found on Roboflow.
Challenges we ran into
A big challenge when approaching this problem was finding efficient data for our models to learn. The sparsity of quality data led to our team having to work around blockers and work with what was publicly available. Once acquired, the ability to work with datasets of image data became an issue due to connection and hardware limitations
Accomplishments that we're proud of holder
Despite the challenges, we were able to create a model that accurately identifies parking spots. Achieving this in such a short time span has felt rewarding, and we are confident that our software can assist users to find real-world solutions.
What we learned
Throughout our project timeline, we learned to stick to our goal better. While we had ideas to expand on our MVP, it would have been more beneficial to keep it simple.
What's next for Benji Lots
Our next step is to locate diverse satellite data to strengthen our model's capabilities. From there, we aim to explore how we can utilize our tool to effectively support business owners in their entrepreneurship journey
Log in or sign up for Devpost to join the conversation.