Inspiration
The Health and Bioech Track, as well as a desire to learn MLM's gave me the idea to detect brain tumors from MRI's.
What it does
Takes input in of an MRI brain scan image, and detects 3 different types of tumor as well as a tumor free brain with 92.7% acuracy.
How we built it
Uses a pre sorted and formatted image set from Kaggle, a pre-trained neural network using Xception and TensorFlow, and a visiual app created using TKinter.
Challenges we ran into
The detection for glioma tumors is around 10% less than the other categories. I think this is due to not being trained on enough images, or not enough time to train the model.
Accomplishments that we're proud of
I had no TensorFlow or MLM experience before Friday night, so I am proud of my 92.7% accuracy, and the app.
What we learned
How to use Kaggle to find data sets, how to train neural network to detect tumors using Xception and Tensorflow, reading images with openCV, and developing Python GUI's using TKinter.
What's next for Detecting Brain Tumors using Xception and Tensorflow
Implenting a larger data set, training the model for longer, adding support for other types of tumors, improve app.
Log in or sign up for Devpost to join the conversation.