AdVenture

Dashboad's Main Page
Dashboard's Login Screen

Inspiration

We wanted an idea to revolutionize the way people advertises, sees and interacts with Billboards!

What it does

The main purpose of the app is to advertise products in stores, based on the people that watch the ads. At first the store owner uploads products to the system using our site. We created machine learning models to assert the values of Gender, Age and Type of a product to make it relevant for a customer later. Now the product is ready to be advertised. Our dual camera billboard detects people in a distance that are interested in our store furthermore it can recognise their faces, their ages and the time they spend watching an ad. When the camera detects someone interested, analyzes a photo of him by sending some values to our ML Model and gets his style. By the style we can understand what kind of products he possibly wants to see and we showing him an ad accordingly. The second camera detects his emotions about our ad and the time he spends watching it. By that values we create a feedback for the AI that learns if the choice of the ad was accurate.

How we built it

We tried to build our app by using state of the art technologies like Azure Machine Learning Studio,Microsoft's Cognitive services and more specifically the Custom Vision API. For our backend we used Microsoft's Web Api 2, MVC which we combined with Entity Framework. Our frontend was made entirely with Vue.js for the owner's endpoint and for the charts we used the chart.js library.

Challenges we ran into

The main challenge was the data mining and scraping we had to do, in order to train our models correctly. We also faced some difficulties when we need to make our multiple projects communicate with each other. But that is the best thing about mistakes you can always learn from them!

Accomplishments that we're proud of

First of all we managed to work as a team untill the end and overcome all problems together. We are different people from different areas of expertise but we managed to cooperate and communicate with each other.

What we learned

Firstly, Machine Learning. Two days ago we were not familiar with the technology as we are today. Secondly, we learned that "Tuxedo" doesn't necessarily makes you a formal man, or at least that was what our ML Model was telling us for about 10 hours!!

What's next for Ad.Venture

We will not stop now! After Junction we plan to continue developing our idea together. One of our goals is to improve our dashboard so it can handle real time data from our models.