SkyPark

Skypark Website
SkyPark Data
SkyPark Framework Diagram
Confusion Matrix
Results Curve
Train Batch 1
Train Batch 2
Train Batch 3

The Machine Learning Goats (MLGs) 🌴🚀 created SkyPark, a revolutionary innovation that utilizes a custom machine learning model and drone computer vision to transform the urban parking system in Wynwood. The purpose is to make the search of parking spaces easier and fast, arriving to your location without endlessly driving in circles.

Inspiration 💡

The MLG's inspiration in creating this project is rooted in the very moment when we were trying to find parking at Wynwood to participate in Miami Hack Week. We recognized that this is a typical problem that Miami residents face. Whether it's arriving to your next appointment, meeting or party! The MLGs have the solution to getting to your location and valuing your time.

What it does ⚙️

SkyPark utilizes drone footage with a custom trained YOLOv8 model that detects available parking spots around Wynwood from an aerial view. Users will be able to go to our website to find open parking spot when they are detected, the drone will notify the user availability and provide the street address of the parking spot. SkyPark also generates a 3D map that lets the user visualize simultaneously localization, mapping, and pinpointing.

How we built it ⚒️

The MLGs built SkyPark by integrating machine learning algorithms, such as PyTorch and YOLOv8 with a custom manually annotated dataset using computer vision aid tool with over 200 images of parking spaces in Wynwood captured with our DJI Mini SE 2 drone. We then used custom python scripts to capture images every 3 seconds from our footage, annotate them, and train our model with an OpenParkingSpace class; essentially training the model with over 250 epochs for accuracy. Then we used Slam to create a 3D mapping terrain of the footage from the drone to pinpoint specifically where there is open parking and alert users. Finally, we embedded ML drone footage, the Slam model, and GPS zip code tracking to a website using Flask.