My Style

Have you ever taken a ton of photos but not known which ones are the best? Well let My Style tell you!

Comment

Inspiration

Have you ever gone on a trip and taken way too many photos? Or had your photo library get a little out of hand? Well a couple people on our team have and decided the best way to tackle this problem would be to create an application that learns your preference from images and then classifies new ones.

What it does

MyStyle allows you to tag photos by how you feel about them, extracts various features about them, and then uses Microsoft Azure to train a machine learning model on these features. From here it can be used to classify new images to the data set.

How we built it

We used Azure computer vision to extract some of the features that the image contained. From here we moved to using Azure Machine Learning to classify the data.

Challenges we ran into

We did not choose our project until very late. At this point, we were unable to connect to all of Azure services, so we were unable to completely implement the project in one place.

Accomplishments that we're proud of

Finally finishing! As a team, we changed out project about 10 times throughout the day so finding the project that we all enjoyed and were able to implement was very nice.

What we learned

The ease of Microsoft Azure, it has a very simple drag and drop interface for machine learning, and we really liked the modular naure of everything. You can switch out one machine learning algorithm for another in literally 5 seconds.

What's next for My Style

As mentioned before, we want to actually have a one-stop app built. A few things to make the model better include adding a few more features, such as saturation and hue value and maybe even a pixel representation of the image. Finally, and this one is probably the easiest to do because of Azure structure, we want to look in to running a regression on the data instead to be able to show relative preferences versus categorized.

Built With

Share this project:

Updates