ImageClassification / Methods.py

How to use the program:

• After running the program, give the path of the dataset.
• The dataset must have 2 folder named Train and Test.

$ tree --Images

.
├── Train                       
│   ├── Airplane                
│   ├── Cat                    
│   ├── Dog                     
│   └── Others     
└── Test                   
    ├── Airplane                
    ├── Cat                     
    ├── Dog                     
    └── Others                  

• After giving the path, choose the methods for image classification.
• For exit , enter q.

Pseudocode:

• User enters the path when the app runs.
• Checks the train and test labels are same or not.
• If they are the same, create variables to hold images from loaded images and their labels and triggers create_data function. If not turn back to “1”.
  • Loads the images from the path that is given by user.
  • Resizes the loaded images as specific size.
  • Merges resized images and their labels in a single array.
  • Shuffles the array.
  • Seperates the images and labels from the array. So the main data is created.
• Asks user to choose what method to be used.
• If the input is between 1-4 (kNN-SVM-RandomForest-NaiveBayes)
  • Triggers HOG function to calculate hog features.
  • HOG function takes the images and extract the features of images.
  • After HOG is done, classifier will be created.
    • For 1 (kNN) Classifier method is -> neighbors.KNeighborsClassifier()
    • For 2 (SVM) Classifier method is -> svm.SVC()
    • For 3 (RandomForest) Classifier method is -> RandomForestClassifier()
    • For 4 (NaiveBayes) Classifier method is -> GaussianNB()
  • After classifier is identified, Method function is triggered to train the model and calculate the accuracy of it.
    • Takes the training and testing data.
    • Trains the data with using identified classifier.
    • Calculates the accuracy.
• If the input is 5 (MLP)
  • Triggers PreprocessDataForMLP for reshaping the images to be used in MLP classifer.
  • After PreprocessDataForMLP is done, classifier will be created.
    • For 5 (MLP) Classifier method is -> MLPClassifier()
  • After classifier is identified, Method function is triggered to train the model and calculate the accuracy of it.
• If the input is 6 (Tensorflow-Keras)
  • Triggers PreprocessDataForTensorflow for reshaping the images and converting the labels to one-hot-encoding in order to be used in TF function.
  • After PreprocessDataForTensorflow is done, TF function is triggered.
    • Creates the model.
    • Sets up the layers.
    • Compiles the model.
    • Trains the data.
    • Evaluates the accuracy.
• Return to “4”.

Example Images From:

• Open Images Dataset V4-V5
• Dogs vs Cats dataset from Microsoft

How to download Open Images V4

Open Images V4 Image Downloading Tool

How to download Dogs vs Cats Dataset

Dogs vs Cats Images