Inspiration
Didn't want to just use an API. Wanted to use Computer Vision to improve on Delta's current airport wait time tracker.
What it does
Analyzes footage from security cameras to count number of people waiting in queues for service. Estimates waiting time for consumers and has special features for enterprise (proactive kiosk allocation, crowd management). Works with OpenCV on a Haar Cascader. We then analyze the change of crowd density and size over time to provide suggestions.
How we built it
Upon virtualizing the environment for speedy, responsive and seamless Python development. We calibrated OpenCV's Haar Cascade Feature Recognition Algorithm to a highly par standard. Designed a flawless schema for the storage of highly codified data. Gathered test footage from the GT gym to mimic airport baggage check-in. Modeled waiting times with a Uniform distribution (with invariant scaling).
Challenges we ran into
Preventing image detection "popping", trying to remove sudden false positives or persistent missed negatives. Handling concurrent processes that share data without using a redundantly large solution. Lack of UI for the system and hence a good way to represent the data.
Accomplishments that we're proud of
Spoke to Delta engineer and found out what to do with this CV application. Implemented algorithm to minimize image pop-in. It is a multi-threaded application.
What we learned
How OpenCV works and the intricacies within.
What's next for Headhunter
Probably going to be abandoned after it becomes a unicorn startup and we drive it into the ground because of extreme mismanagement and racking up fines. And stealing trade secrets from Google.
Log in or sign up for Devpost to join the conversation.