Our original inspiration came from the #trash-tag event that has been trending recently where you take a picture of yourself at a place with trash littered everywhere and another picture after cleaning it, furthermore challenging others to contribute towards the social cause of saving our environment.
What it does
Trash-Tag brings in an intuitive game where multiple people sign up and register themselves with their organization. A person can register their face through the system, and have every good token of citizenship accurately rewarded to their profile per garbage eradication. Any registered person gets a point to throw trash in respective segregated trash bins. The software also detects the type of object (in our case we decided to limit object detection to cans/bottle/others).
How I built it
We used Google Cloud Vision object detection (cans/bottles), Amazon-Web-Services Rekognition for Facial detections and recognition, MySQL DBMS over Google Cloud Storage API, Open-CV for detecting, registering faces and sensing the presence of multiple objects in front of the camera sensor.
Challenges I ran into
This was the first time that we delve into Cloud APIs and image processing. Certainly, being pretty new we naively thought GCP would cater to all needs of our project. However, we soon realized we need more powerful and a combination of OpenCV and AWS services to make it work. There were many times
Accomplishments that I'm proud of
We're proud to have gained solid experience with not one but TWO of the worlds biggest cloud service providers.
What I learned
Even when your team is more than capable, it's a great idea to start off small and expand upon a minimum viable product. This MVP not only provides a great springboard for future features but also motivates the team with an early win :).
What's next for Trash-Tag
Geolocated trash cans, continued improvement in computer vision, and an improved UI for users and organizations.