Skip to content

This project focuses on using multi deep learning model to get best possible accuracy on disease image classification

Notifications You must be signed in to change notification settings

aatansen/Multi-Model-Comparison-Deep-Learning-Project

Folders and files

NameName
Last commit message
Last commit date

Latest commit

 

History

12 Commits
 
 
 
 
 
 
 
 
 
 

Repository files navigation

Multi Model Comparison Deep Learning Project

  • Pretrained model is used
  • freezing all layer except x number of layer (experimental)
  • ImageDataGenerator is used for preprocessing images
  • In output Softmax activation is used as it is multiclass
  • Total num epochs : 50
  • Adam optimizer is used with 0.0001 learning rate
  • Dropout & early_stop function used to avoid overfitting.
  • Changes
    • Rotation Range changed from 20 to 10
    • Width , Height shift & zoom range changed from 0.2 to 0.1
    • Freezing all layer
    • Model saved based on max validation accuracy
    • Roboflow dataset is used (Previous ModelV01 was on Kaggle Dataset)
    • Total seven model is trained and measured
  • Removed
    • Shear range Removed
    • Dropout, New layer (experimented previously)
    • Early stop function removed
  • Added
    • Preprocessing function added
    • Shuffle added
  • Advantage
    • Accuracy increased
    • Complexity reduced
    • Less number of params
    • Accurate ROC,AUC curve
    • Accurate performance matrices
  • Changes
    • Kaggle and Roboflow both dataset is Measured separately
    • Roboflow dataset decreases after filtering
  • Removed
    • Few number of images removed after filtering process
  • Added
    • Image filtering added
    • labels added for image preview and other task
    • Total training time measured
    • Epoch number measured by highest training and validation accuracy
  • Advantage
    • Black images detected
    • Two dataset models are compared
    • Accurate result analysis
  • Changes
    • Kaggle and Roboflow dataset are combined
  • Removed
    • Nothing removed
  • Added
    • More comparison added
  • Advantage
    • Accuracy improved

About

This project focuses on using multi deep learning model to get best possible accuracy on disease image classification

Resources

Stars

Watchers

Forks

Releases

No releases published

Packages

No packages published