We credit our ideas to Juliana Shihadeh's research described in "Deep Learning Based Image Classification for Remote Medical Diagnosis."
What it does
When given a data set of moles each labeled as either cancerous or noncancerous, a neural network classifies uncategorized images into their proper classifications.
How we built it
We used MatLab's pertained network AlexNet by utilizing transfer learning we were able to modify the network to fit our needs.
Challenges we ran into
Another application to this project is creating a deep learning model to identify whether or not a lung x-ray is cancerous. This was our original idea. However, with the x-ray lung data sets used, the neural network was unable to properly classify images as cancerous vs. noncancerous. Due to this, we created a proof of concept by using the neural network to identify moles as either cancerous or noncancerous. We are looking forward to working on improving the neural network in order to properly identify x-ray lung images in the future.
Accomplishments that we're proud of
What we learned
We were all fairly unfamiliar with neural networks and deep learning, and were all able to learn quite a bit about that process.
What's next for HackForHumanity2019