Page 338 - AI Ver 3.0 class 10_Flipbook
P. 338

Stage 2   Data Acquisition
              Relevant datasets are identified, collected, and prepared for analysis. For the coral bleaching project, the dataset
              used is from the manuscript titled "Bag of Features (BoF) Based Deep Learning Framework for Bleached
              Corals Detection." This dataset is intended for research and experimentation to develop AI models for detecting
              bleached corals.
              You can download the Coral Bleaching dataset from the following link:
              https://orangewebsupport.co.in/AI/Coral_Bleaching.zip

              OR
              https://bit.ly/orange_computer_vision
              Note, Image Analytics can be added by selecting Image option from Add-ins from Options tab.
              The steps for acquiring training dataset are as follows:
               Step 1     Open Orange Data Mining Tool and locate the Import Image widget.
               Step 2     Drag and drop this widget onto your workflow canvas.

               Step 3     Right-click on the Import Image widget on the canvas.
               Step 4     Select the Rename option.
               Step 5     Enter the new name: Training Data to make it easier to

                          identify the widget's purpose.
               Step 6     Double-click the renamed Training Data icon to open

                          its configuration window.
               Step 7     Click on the Browse button.
                          The Select Top Level Directory dialog box appears.
               Step 8     Navigate the location containing your training dataset.
               Step 9     Click on the Select Folder button.





















                          Ensure that your dataset is organised into subfolders, where each subfolder corresponds to a category

                          or class (e.g., "Bleached" and "Unbleached").
                          Orange will automatically recognise the 2 categories based

                          on  the  subfolder  structure  in  the  training  dataset.  These
                          categories represent the 2 classes of the problem:
                             • Bleached: Images of corals that are bleached.
                             • Unbleached: Images of healthy corals.



                    336     Touchpad Artificial Intelligence (Ver. 3.0)-X
   333   334   335   336   337   338   339   340   341   342   343