Nevus-Detection-via-DIP-Machine-Learning

Kindly read the problem statement first to understand my approaches, you need to download the ph2 folder from the link given below and give directory to the create_dataset code to generate format (class wise image seperation) for model genearation.

Dataset and Model file link: https://drive.google.com/drive/folders/1c8tH-UwQfXQkHwmLfSPPtWpcaVODaq17?usp=drive_link

DIP:

1. Data seperation and Classification
2. Feature extraction
3. Feature scalar value assigning for further graphical representation
4. Box plot graph
5. Analysis: Box plot shows variation among provided images in each of these 3 classes
6. using different models for detection and checking the accuracy: a) Decision Tree Classifier 47.50% b) Logistic Regression 55.00% c) AdaBoost Classifier 55.00% d) KNeighbors Classifier 55.00% e) Random Forest Classifier 60.00% f) Grid Search CV 55.00%
7. test images accuracy table

ML:

1. Data set fornat for model training
2. threshold, feature extraction (hog) and epoche = 10 setting for model training (tensorflow)
3. preprocessing image
4. Dedection using trained model: a) Best Accuracy: 95.3% b) Worst Accuracy: 49.0%