-
-
Overview
-
x-ray image
-
result
Video
You can see the loom video here
Inspiration
In my biology lessons, I learned about the causes of cancer and I also learned that the highest cause of death in women is breast cancer. One of the reasons are late diagnosis and misdiagnosis. So I thought of creating a web app that detects breast cancer early with the help of object detection.
What it does
Detects breast cancer from x-ray images with instance segmantation
How we built it
For my project i used vscode as editor and trained my YOLOv8 model with a x-ray images dataset, the notebook from Roboflow make it possible. The libraries I used and for what are:
dash --> web framework; create web app
plotly --> data visualisation library; create figures and show images
random --> library for generate random number; create random colors for each class
io --> module for dealing various types of input/output; I/O operations on byte data
base64 --> module for encoding and decoding binary data; decode binary data
opencv --> computer vision library for image processing; resize, convert color(BGR to RGB), draw boxes and filled polygons and overlay segmentation masks
numpy --> numerical computing library; convert mask as array and reshape arrays
ultralytics --> deep learning object detection framework; load YOLOv8 model for instance segmentation
Challenges I ran into
Not enough computing power to train my YOLO model with my customer dataset Solution: reduces epochs Consequence: reduces accuracy
Accomplishments that we're proud of
That my project even work
What I learned
I learned how instance segmentation works and the dangers of ignorance and late diagnosis of breast cancer
What's next for Breast Cancer Detection
Train my YOLO model more so it becomes better and train it with other cancer types datasets
How it works
You just upload a x-ray image and in few seconds the program finds out whether it is breast cancer or not
Log in or sign up for Devpost to join the conversation.