Page 211 - Ai_C10_Flipbook
P. 211

Step 28    Double-click on the Confusion Matrix widget to open it.
                            You will see a table showing the distribution of correct and incorrect predictions for each class, typically

                            as follows:
                               • True Positives (TP): Correctly predicted instances of the positive class.
                               • False Positives (FP): Incorrectly predicted instances of the positive class.

                               • False Negatives (FN): Incorrectly predicted instances of the negative class.
                               • True Negatives (TN): Correctly predicted instances of the negative class.
                            The confusion matrix will first show the results for the Logistic Regression algorithm.

                            You will see the counts of correct and incorrect predictions (TP, TN, FP, FN) for both classes (Bleached

                            and Unbleached).
                            Similarly,  the  confusion  matrix  will  show  the  results  for  the  SVM  (Support  Vector  Machine)

                            algorithm.
                            This  will  allow  you  to  compare  how  SVM  performs  against  the  other  two  models  in  terms  of

                            classification errors.
                            Lastly, the confusion matrix will display the results for the Random Forest algorithm.
                            The matrix will show how well Random Forest performed in correctly and incorrectly predicting the

                            classes.
                 Step 29    Click the Close button to close the Confusion Matrix dialog box.
                            Once you have built and evaluated your model, the then you need to apply the model to new (unseen)

                            test data in order to make predictions.
                 The steps for predicting model are as follows:
                 Step 30    Drag the Import Image widget onto the canvas to bring in the test data (images you want to

                            classify based on the trained model).
                 Step 31    Right-click on the Import Image widget, and rename it to Testing Data. This helps you easily identify

                            the test dataset in your workflow.
                 Step 32    Double-click  on  the  Testing Data  widget,  and  select  the  directory  that  contains  the  test  dataset
                            (images of bleached and unbleached corals).
                 Step 33    Click the Close button to close the Training Data dialog box.

                 Step 34    Drag and drop the Image Viewer widget onto the canvas.

                 Step 35    Connect the Training Data to the Image Viewer.

                 Step 36    Double-click the Image Viewer widget to view the images in the training dataset.
                            The Image Viewer widget will show the test dataset that you selected, allowing you to visually inspect

                            the images that will be used for predictions.
                 Step 37    Drag  and  drop  the  Image Embedding  widget  to  the  canvas,  which  will  convert  the  images  into

                            numerical features for prediction.
                 Step 38    Connect the Test Data widget (test dataset) to the Image Embedding widget, which processes the

                            images into numerical representations that can be used by the model.

                                                                                             Computer Vision    209
   206   207   208   209   210   211   212   213   214   215   216