Onda Auto-See Detection

Inspiration

Parking lots at GMU are often congested, leading to frustration for students. Seeing this as an opportunity, we were inspired to create a streamlined parking solution that combines License Plate Recognition (LPR), Facial Recognition, and Object Detection technologies. The goal was to provide an efficient and secure parking experience that benefits both users and property managers, making parking faster and safer.

What it does

Onda Auto-See Detection enhances parking lot management by using:

• License Plate Recognition to monitor vehicle entries, exits, and durations.
• Facial Recognition to verify registered personnel, enhancing access control and security.
• Object Detection to track vacant parking spots and detect potential security risks, ensuring a safer environment for users. Through a mobile app, users can receive real-time updates on spot availability, reducing the time spent searching for parking and enhancing their overall experience.

How we built it

We built Onda Auto-See Detection using computer vision and machine learning:

• LPR, facial recognition, and object detection models were integrated to provide accurate and real-time data.
• A website interface allows users to see real time data.
• We configured a Raspberry Pi 4 and Arducam to capture real time footage.
• Frames are then processed and pushed through separate models to determine the contents of each frame and to record the respective data.

Challenges we ran into

Developing a solution that integrates LPR, facial recognition, and object detection required substantial optimization for real-time processing and the time constraints to build out models for each parameter. To provide proof of concept we realized it would be best to constrain our current abilities to LPR in order to provide a working prototype. Ensuring system accuracy is complex, and as a team new to hackathons, frameworks, and the sheer size of this project we found ourselves tackling challenges we have never faced and learning along the way. This is both of our first projects at this scale. That being said we ran into challenges being that our ambitions were high and we had to constrain our working product to insure we produced a prototype that did its job efficiently.

Accomplishments that we're proud of

We are proud to have created a reliable, real-time parking management solution product plan and that we built out a capable LPR system. The website allows users to see real time recording of the LPR program. Generating a SaaS model that provides actionable insights has been a significant achievement as we can see the product being viable in real world use cases.

What we learned

This project highlighted the importance of seamless system integration and optimization for real-time performance. We learned to address the nuances of high-accuracy AI models in a practical, real-world environment and to enjoying the challenges of realizing your ambitions. We learned how to work with various model training platforms, React frameworks, creating a User Interface, and simply going from idea to feasible business plan.

What's next for Onda Auto-See Detection

Moving forward, we plan to implement the facial recognition + object detection and increase accuracy further to expand our solution applications. We aim to refine our analytics capabilities, giving clients more insights into parking trends and user behaviors, ultimately helping them maximize space utilization. We see the potential for the project and we would love to see it scaled and tested in a real world application.

Built With

• google-cloud
• https://platerecognizer.com
• node.js