Skip to content

Latest commit

 

History

History
71 lines (42 loc) · 2.17 KB

README.md

File metadata and controls

71 lines (42 loc) · 2.17 KB

BRAIN TUMOUR DETECTION

BUSINESS CASE: BASED ON BRAIN MRI IMAGES DATASET WE NEED PREDICT BRAIN TUMOUR

TASK: BINARY CLASSIFICATION

image

DATA SUMMARY

Total 274 Jpg Images are present In two classes

  • No ----> 119
  • Yes ----> 155

DEVICE THE PROJECT INTO MULTIPLE STEPS

  • Prepare training, validation and testing set
  • Get all classes labels
  • Visualise training images
  • Used CNN & VGG19 model
  • Model Compilation
  • Model Training
  • Model Evaluation
  • Model Saving
  • Prediction on test data

Architecture of prepare CNN model

image

Compile & Train Model

  • Use binary crossentropy loss with adam optimizer and metrics is accuracy
  • Balance the class weights
  • Train model of 80 epoch and use model checkpoint to save the best model
  • Plotting training, validation loss as well as accuracy

image

image

Achieved 88.24% testing accuracy

Used VGG19 transfer learning technique to get better accuracy

  • Import vgg19 library and set input image size & used imagnet dataset weight as well as not include fully connected layer at top
  • Freeze the existing weights
  • Add more layers with sigmoid activation function

Architecture of VGG19 model

image

Compile & Train model

  • Use binary crossentropy loss with adam optimizer and metrics is accuracy
  • Train model of 60 epoch and set the class weight
  • Plotting training, validation loss as well as accuracy

image

image

Save model

Achieved 92.16% testing accuracy