Project Artage

Inspiration

Just a mild interest in paintings...maybe a bit more than mild...

What it does

Predicts year artwork was made

How we built it

Convolutional Neural Networks

Challenges I ran into

Alot of the issues came from pre-processing the data. We had trouble cleaning the data to the specifications we wanted that would suit the neural network.

Accomplishments that I'm proud of

We were able to make the neural network train a model successfully as well as run it which was already a huge accomplishment for us.

What I learned

We learned too much about numpy arrays today. We now know why people use numpy instead of lists but we still have to get on grips with how to fully use it as I'm not used to working with ML models.

What's next for [Table 53] Project Artage

We only ran it for one epoch and we had to run the resizer script to resize it to only a 50x50 image due to time constraints (it was a 12-hour hack), and so we got an accuracy of 21% which is surprisingly good. With just more time, a better GUI and more epochs, we could definitely develop this project and boost the accuracy rating. We could also make it look for other parameters such as genre or artist instead of just the year.

Built With

• cnn
• machine-learning
• python