Generating datastores from Bangalore image tiles, train and evaluate classifiers

[elboyran]

• [x] Make a script to generate datastore for combination of desired ROIs, training and testing classifiers: simpleImageCategoryClassifier_Bangalore

• [x] Fix tiles visualization

• [x] Create figure illustrating the generated tiles and include in paper.

• [x] Modify simpleImageCategoryClassifier_Bangalore to be able to save results from runs.

• [x] Train and test classifiers

• [x] Function to plot Acc, Sens, Spec and Prec for all parameters and classes.

• [x] Create script to generate performance figures.

• [x] Make the figures and include them in the paper

• [x] Function to plot FScore all parameters and classes

• [x] Modify script to generate also FScore figures.

• [x] Make the figures and include them in the paper

• [x] Modify visualization scripts to show up to 40m. Recreate figures and replace old ones in the paper.

[elboyran]

Creating datastore for dataset 1:px67m10...

imds = 

  ImageDatastore with properties:

       Files: {
              ' ...\px67m10\ROI1\BuiltUp\Bangalore_ROI1_tile_sr100er166sc100ec166.tif';
              ' ...\px67m10\ROI1\BuiltUp\Bangalore_ROI1_tile_sr100er166sc1057ec1123.tif';
              ' ...\px67m10\ROI1\BuiltUp\Bangalore_ROI1_tile_sr100er166sc1090ec1156.tif'
               ... and 165146 more
              }
      Labels: [BuiltUp; BuiltUp; BuiltUp ... and 165146 more categorical]
    ReadSize: 1
     ReadFcn: @readDatastoreImage

      Label         Count   
    __________    __________

    BuiltUp       1.0653e+05
    NonBuiltUp         55592
    Slum                3030

ans =

  3×2 table

      Label       Count
    __________    _____

    BuiltUp       3030 
    NonBuiltUp    3030 
    Slum          3030 

-----------------------------------------------------------------
Splititng the datastore into Train and Test datastores with fractionTrain: 80%
Training datastore: 
      Label       Count
    __________    _____

    BuiltUp       2424 
    NonBuiltUp    2424 
    Slum          2424 

Testing datastore: 
      Label       Count
    __________    _____

    BuiltUp       606  
    NonBuiltUp    606  
    Slum          606  

-----------------------------------------------------------------
Creating features (BoVW) using vocabulary size 10 and SUFR point locations Detector for the Train datastore keeping the 80% strongest features

Creating Bag-Of-Features.
-------------------------
* Image category 1: BuiltUp
* Image category 2: NonBuiltUp
* Image category 3: Slum
* Selecting feature point locations using the Detector method.
* Extracting SURF features from the selected feature point locations.
** detectSURFFeatures is used to detect key points for feature extraction.

* Extracting features from 7272 images...done. Extracted 13470 features.

* Keeping 80 percent of the strongest features from each category.

* Balancing the number of features across all image categories to improve clustering.
** Image category 2 has the least number of strongest features: 1460.
** Using the strongest 1460 features from each of the other image categories.

* Using K-Means clustering to create a 10 word visual vocabulary.
* Number of features          : 4380
* Number of clusters (K)      : 10

* Initializing cluster centers...100.00%.
* Clustering...completed 33/100 iterations (~0.02 seconds/iteration)...converged in 33 iterations.

* Finished creating Bag-Of-Features

-----------------------------------------------------------------
Training an image categiry classifier

Training an image category classifier for 3 categories.
--------------------------------------------------------
* Category 1: BuiltUp
* Category 2: NonBuiltUp
* Category 3: Slum

* Encoding features for 7272 images...done.

* Finished training the category classifier. Use evaluate to test the classifier on a test set.

-----------------------------------------------------------------
Evaluating perfomance on the Training set

Evaluating image category classifier for 3 categories.
-------------------------------------------------------

* Category 1: BuiltUp
* Category 2: NonBuiltUp
* Category 3: Slum

* Evaluating 7272 images...done.

* Finished evaluating all the test sets.

* The confusion matrix for this test set is:

                          PREDICTED
KNOWN         | BuiltUp   NonBuiltUp   Slum   
----------------------------------------------
BuiltUp       | 0.15      0.50         0.35   
NonBuiltUp    | 0.03      0.78         0.19   
Slum          | 0.09      0.25         0.67   

* Average Accuracy is 0.53.

                  accuracy    sensitivity    specificity    precision    recall    Fscore 
                  ________    ___________    ___________    _________    ______    _______

    BuiltUp       67.533      14.645         93.977         54.869       14.645     0.2312
    NonBuiltUp    67.437      77.517         62.397         50.756       77.517    0.61345
    Slum          70.861      66.584         72.999         55.217       66.584     0.6037

Evaluating perfomance on the Training set
-----------------------------------------------------------------
Evaluating perfomance on the Test set

Evaluating image category classifier for 3 categories.
-------------------------------------------------------

* Category 1: BuiltUp
* Category 2: NonBuiltUp
* Category 3: Slum

* Evaluating 1818 images...done.

* Finished evaluating all the test sets.

* The confusion matrix for this test set is:

                          PREDICTED
KNOWN         | BuiltUp   NonBuiltUp   Slum   
----------------------------------------------
BuiltUp       | 0.18      0.47         0.35   
NonBuiltUp    | 0.04      0.76         0.19   
Slum          | 0.10      0.24         0.66   

* Average Accuracy is 0.53.

                  accuracy    sensitivity    specificity    precision    recall    Fscore 
                  ________    ___________    ___________    _________    ______    _______

    BuiltUp       67.767      17.822         92.739         55.102       17.822    0.26933
    NonBuiltUp    68.427      76.238         64.521         51.794       76.238    0.61682
    Slum          70.407      65.842          72.69         54.658       65.842    0.59731

Saving perfomance on the Test set
-----------------------------------------------------------------
Creating features (BoVW) using vocabulary size 20 and SUFR point locations Detector for the Train datastore keeping the 80% strongest features

Creating Bag-Of-Features.
-------------------------
* Image category 1: BuiltUp
* Image category 2: NonBuiltUp
* Image category 3: Slum
* Selecting feature point locations using the Detector method.
* Extracting SURF features from the selected feature point locations.
** detectSURFFeatures is used to detect key points for feature extraction.

* Extracting features from 7272 images...done. Extracted 13470 features.

* Keeping 80 percent of the strongest features from each category.

* Balancing the number of features across all image categories to improve clustering.
** Image category 2 has the least number of strongest features: 1460.
** Using the strongest 1460 features from each of the other image categories.

* Using K-Means clustering to create a 20 word visual vocabulary.
* Number of features          : 4380
* Number of clusters (K)      : 20

* Initializing cluster centers...100.00%.
* Clustering...completed 30/100 iterations (~0.03 seconds/iteration)...converged in 30 iterations.

* Finished creating Bag-Of-Features

-----------------------------------------------------------------
Training an image categiry classifier

Training an image category classifier for 3 categories.
--------------------------------------------------------
* Category 1: BuiltUp
* Category 2: NonBuiltUp
* Category 3: Slum

* Encoding features for 7272 images...done.

* Finished training the category classifier. Use evaluate to test the classifier on a test set.

-----------------------------------------------------------------
Evaluating perfomance on the Training set

Evaluating image category classifier for 3 categories.
-------------------------------------------------------

* Category 1: BuiltUp
* Category 2: NonBuiltUp
* Category 3: Slum

* Evaluating 7272 images...done.

* Finished evaluating all the test sets.

* The confusion matrix for this test set is:

                          PREDICTED
KNOWN         | BuiltUp   NonBuiltUp   Slum   
----------------------------------------------
BuiltUp       | 0.20      0.47         0.34   
NonBuiltUp    | 0.06      0.75         0.19   
Slum          | 0.11      0.21         0.68   

* Average Accuracy is 0.54.

                  accuracy    sensitivity    specificity    precision    recall    Fscore 
                  ________    ___________    ___________    _________    ______    _______

    BuiltUp       67.561      19.513         91.584         53.689       19.513    0.28623
    NonBuiltUp    69.156      75.371         66.048         52.606       75.371    0.61964
    Slum           71.59      67.574         73.597         56.134       67.574    0.61325

Evaluating perfomance on the Training set
-----------------------------------------------------------------
Evaluating perfomance on the Test set

Evaluating image category classifier for 3 categories.
-------------------------------------------------------

* Category 1: BuiltUp
* Category 2: NonBuiltUp
* Category 3: Slum

* Evaluating 1818 images...done.

* Finished evaluating all the test sets.

* The confusion matrix for this test set is:

                          PREDICTED
KNOWN         | BuiltUp   NonBuiltUp   Slum   
----------------------------------------------
BuiltUp       | 0.21      0.44         0.34   
NonBuiltUp    | 0.08      0.72         0.20   
Slum          | 0.12      0.21         0.67   

* Average Accuracy is 0.54.

                  accuracy    sensitivity    specificity    precision    recall    Fscore 
                  ________    ___________    ___________    _________    ______    _______

    BuiltUp       67.217      21.287         90.182         52.016       21.287    0.30211
    NonBuiltUp    69.142      72.277         67.574         52.708       72.277     0.6096
    Slum          70.902      67.327          72.69          55.21       67.327    0.60669

Saving perfomance on the Test set
-----------------------------------------------------------------
Creating features (BoVW) using vocabulary size 50 and SUFR point locations Detector for the Train datastore keeping the 80% strongest features

Creating Bag-Of-Features.
-------------------------
* Image category 1: BuiltUp
* Image category 2: NonBuiltUp
* Image category 3: Slum
* Selecting feature point locations using the Detector method.
* Extracting SURF features from the selected feature point locations.
** detectSURFFeatures is used to detect key points for feature extraction.

* Extracting features from 7272 images...done. Extracted 13470 features.

* Keeping 80 percent of the strongest features from each category.

* Balancing the number of features across all image categories to improve clustering.
** Image category 2 has the least number of strongest features: 1460.
** Using the strongest 1460 features from each of the other image categories.

* Using K-Means clustering to create a 50 word visual vocabulary.
* Number of features          : 4380
* Number of clusters (K)      : 50

* Initializing cluster centers...100.00%.
* Clustering...completed 25/100 iterations (~0.03 seconds/iteration)...converged in 25 iterations.

* Finished creating Bag-Of-Features

-----------------------------------------------------------------
Training an image categiry classifier

Training an image category classifier for 3 categories.
--------------------------------------------------------
* Category 1: BuiltUp
* Category 2: NonBuiltUp
* Category 3: Slum

* Encoding features for 7272 images...done.

* Finished training the category classifier. Use evaluate to test the classifier on a test set.

-----------------------------------------------------------------
Evaluating perfomance on the Training set

Evaluating image category classifier for 3 categories.
-------------------------------------------------------

* Category 1: BuiltUp
* Category 2: NonBuiltUp
* Category 3: Slum

* Evaluating 7272 images...done.

* Finished evaluating all the test sets.

* The confusion matrix for this test set is:

                          PREDICTED
KNOWN         | BuiltUp   NonBuiltUp   Slum   
----------------------------------------------
BuiltUp       | 0.17      0.49         0.34   
NonBuiltUp    | 0.04      0.78         0.18   
Slum          | 0.08      0.24         0.69   

* Average Accuracy is 0.54.

                  accuracy    sensitivity    specificity    precision    recall    Fscore 
                  ________    ___________    ___________    _________    ______    _______

    BuiltUp       68.509      17.079         94.224         59.654       17.079    0.26555
    NonBuiltUp    68.551      77.764         63.944         51.885       77.764    0.62242
    Slum          71.906      68.606         73.556         56.469       68.606    0.61948

Evaluating perfomance on the Training set
-----------------------------------------------------------------
Evaluating perfomance on the Test set

Evaluating image category classifier for 3 categories.
-------------------------------------------------------

* Category 1: BuiltUp
* Category 2: NonBuiltUp
* Category 3: Slum

* Evaluating 1818 images...done.

* Finished evaluating all the test sets.

* The confusion matrix for this test set is:

                          PREDICTED
KNOWN         | BuiltUp   NonBuiltUp   Slum   
----------------------------------------------
BuiltUp       | 0.18      0.46         0.35   
NonBuiltUp    | 0.04      0.75         0.20   
Slum          | 0.11      0.23         0.66   

* Average Accuracy is 0.53.

                  accuracy    sensitivity    specificity    precision    recall    Fscore 
                  ________    ___________    ___________    _________    ______    _______

    BuiltUp       67.767      18.482         92.409         54.902       18.482    0.27654
    NonBuiltUp    68.482      75.248         65.099         51.877       75.248    0.61414
    Slum          70.132      65.842         72.277         54.286       65.842    0.59508

Saving perfomance on the Test set
-----------------------------------------------------------------
**********************************************************************************

[elboyran]

Creating datastore for dataset 2:px134m20...

imds = 

  ImageDatastore with properties:

       Files: {
              ' ...\ROI1\BuiltUp\Bangalore_ROI1_tile_sr1006er1139sc2145ec2278.tif';
              ' ...\ROI1\BuiltUp\Bangalore_ROI1_tile_sr1006er1139sc2212ec2345.tif';
              ' ...\ROI1\BuiltUp\Bangalore_ROI1_tile_sr1006er1139sc2279ec2412.tif'
               ... and 32974 more
              }
      Labels: [BuiltUp; BuiltUp; BuiltUp ... and 32974 more categorical]
    ReadSize: 1
     ReadFcn: @readDatastoreImage

      Label       Count
    __________    _____

    BuiltUp       22200
    NonBuiltUp    10252
    Slum            525

ans =

  3×2 table

      Label       Count
    __________    _____

    BuiltUp       525  
    NonBuiltUp    525  
    Slum          525  

-----------------------------------------------------------------
Splititng the datastore into Train and Test datastores with fractionTrain: 80%
Training datastore: 
      Label       Count
    __________    _____

    BuiltUp       420  
    NonBuiltUp    420  
    Slum          420  

Testing datastore: 
      Label       Count
    __________    _____

    BuiltUp       105  
    NonBuiltUp    105  
    Slum          105  

-----------------------------------------------------------------
Creating features (BoVW) using vocabulary size 10 and SUFR point locations Detector for the Train datastore keeping the 80% strongest features

Creating Bag-Of-Features.
-------------------------
* Image category 1: BuiltUp
* Image category 2: NonBuiltUp
* Image category 3: Slum
* Selecting feature point locations using the Detector method.
* Extracting SURF features from the selected feature point locations.
** detectSURFFeatures is used to detect key points for feature extraction.

* Extracting features from 1260 images...done. Extracted 19446 features.

* Keeping 80 percent of the strongest features from each category.

* Balancing the number of features across all image categories to improve clustering.
** Image category 2 has the least number of strongest features: 2074.
** Using the strongest 2074 features from each of the other image categories.

* Using K-Means clustering to create a 10 word visual vocabulary.
* Number of features          : 6222
* Number of clusters (K)      : 10

* Initializing cluster centers...100.00%.
* Clustering...completed 30/100 iterations (~0.03 seconds/iteration)...converged in 30 iterations.

* Finished creating Bag-Of-Features

-----------------------------------------------------------------
Training an image categiry classifier

Training an image category classifier for 3 categories.
--------------------------------------------------------
* Category 1: BuiltUp
* Category 2: NonBuiltUp
* Category 3: Slum

* Encoding features for 1260 images...done.

* Finished training the category classifier. Use evaluate to test the classifier on a test set.

-----------------------------------------------------------------
Evaluating perfomance on the Training set

Evaluating image category classifier for 3 categories.
-------------------------------------------------------

* Category 1: BuiltUp
* Category 2: NonBuiltUp
* Category 3: Slum

* Evaluating 1260 images...done.

* Finished evaluating all the test sets.

* The confusion matrix for this test set is:

                          PREDICTED
KNOWN         | BuiltUp   NonBuiltUp   Slum   
----------------------------------------------
BuiltUp       | 0.47      0.27         0.26   
NonBuiltUp    | 0.15      0.71         0.15   
Slum          | 0.14      0.05         0.81   

* Average Accuracy is 0.66.

                  accuracy    sensitivity    specificity    precision    recall    Fscore 
                  ________    ___________    ___________    _________    ______    _______

    BuiltUp       72.698      46.905         85.595          61.95       46.905    0.53388
    NonBuiltUp    79.603      70.714         84.048          68.91       70.714      0.698
    Slum          80.079      80.952         79.643         66.536       80.952     0.7304

Evaluating perfomance on the Training set
-----------------------------------------------------------------
Evaluating perfomance on the Test set

Evaluating image category classifier for 3 categories.
-------------------------------------------------------

* Category 1: BuiltUp
* Category 2: NonBuiltUp
* Category 3: Slum

* Evaluating 315 images...done.

* Finished evaluating all the test sets.

* The confusion matrix for this test set is:

                          PREDICTED
KNOWN         | BuiltUp   NonBuiltUp   Slum   
----------------------------------------------
BuiltUp       | 0.55      0.21         0.24   
NonBuiltUp    | 0.12      0.71         0.16   
Slum          | 0.16      0.08         0.76   

* Average Accuracy is 0.68.

                  accuracy    sensitivity    specificity    precision    recall    Fscore 
                  ________    ___________    ___________    _________    ______    _______

    BuiltUp       75.556      55.238         85.714         65.909       55.238    0.60104
    NonBuiltUp    80.952      71.429         85.714         71.429       71.429    0.71429
    Slum           78.73       76.19             80         65.574        76.19    0.70485

Saving perfomance on the Test set
-----------------------------------------------------------------
Creating features (BoVW) using vocabulary size 20 and SUFR point locations Detector for the Train datastore keeping the 80% strongest features

Creating Bag-Of-Features.
-------------------------
* Image category 1: BuiltUp
* Image category 2: NonBuiltUp
* Image category 3: Slum
* Selecting feature point locations using the Detector method.
* Extracting SURF features from the selected feature point locations.
** detectSURFFeatures is used to detect key points for feature extraction.

* Extracting features from 1260 images...done. Extracted 19446 features.

* Keeping 80 percent of the strongest features from each category.

* Balancing the number of features across all image categories to improve clustering.
** Image category 2 has the least number of strongest features: 2074.
** Using the strongest 2074 features from each of the other image categories.

* Using K-Means clustering to create a 20 word visual vocabulary.
* Number of features          : 6222
* Number of clusters (K)      : 20

* Initializing cluster centers...100.00%.
* Clustering...completed 20/100 iterations (~0.04 seconds/iteration)...converged in 20 iterations.

* Finished creating Bag-Of-Features

-----------------------------------------------------------------
Training an image categiry classifier

Training an image category classifier for 3 categories.
--------------------------------------------------------
* Category 1: BuiltUp
* Category 2: NonBuiltUp
* Category 3: Slum

* Encoding features for 1260 images...done.

* Finished training the category classifier. Use evaluate to test the classifier on a test set.

-----------------------------------------------------------------
Evaluating perfomance on the Training set

Evaluating image category classifier for 3 categories.
-------------------------------------------------------

* Category 1: BuiltUp
* Category 2: NonBuiltUp
* Category 3: Slum

* Evaluating 1260 images...done.

* Finished evaluating all the test sets.

* The confusion matrix for this test set is:

                          PREDICTED
KNOWN         | BuiltUp   NonBuiltUp   Slum   
----------------------------------------------
BuiltUp       | 0.53      0.29         0.18   
NonBuiltUp    | 0.14      0.75         0.12   
Slum          | 0.13      0.06         0.81   

* Average Accuracy is 0.70.

                  accuracy    sensitivity    specificity    precision    recall    Fscore 
                  ________    ___________    ___________    _________    ______    _______

    BuiltUp       75.556      53.333         86.667         66.667       53.333    0.59259
    NonBuiltUp    79.921      74.762           82.5         68.113       74.762    0.71283
    Slum          83.889      80.952         85.357         73.434       80.952     0.7701

Evaluating perfomance on the Training set
-----------------------------------------------------------------
Evaluating perfomance on the Test set

Evaluating image category classifier for 3 categories.
-------------------------------------------------------

* Category 1: BuiltUp
* Category 2: NonBuiltUp
* Category 3: Slum

* Evaluating 315 images...done.

* Finished evaluating all the test sets.

* The confusion matrix for this test set is:

                          PREDICTED
KNOWN         | BuiltUp   NonBuiltUp   Slum   
----------------------------------------------
BuiltUp       | 0.53      0.27         0.20   
NonBuiltUp    | 0.12      0.76         0.11   
Slum          | 0.20      0.05         0.75   

* Average Accuracy is 0.68.

                  accuracy    sensitivity    specificity    precision    recall    Fscore 
                  ________    ___________    ___________    _________    ______    _______

    BuiltUp       73.651      53.333          83.81         62.222       53.333    0.57436
    NonBuiltUp    81.587       76.19         84.286         70.796        76.19    0.73394
    Slum           81.27      75.238         84.286         70.536       75.238    0.72811

Saving perfomance on the Test set
-----------------------------------------------------------------
Creating features (BoVW) using vocabulary size 50 and SUFR point locations Detector for the Train datastore keeping the 80% strongest features

Creating Bag-Of-Features.
-------------------------
* Image category 1: BuiltUp
* Image category 2: NonBuiltUp
* Image category 3: Slum
* Selecting feature point locations using the Detector method.
* Extracting SURF features from the selected feature point locations.
** detectSURFFeatures is used to detect key points for feature extraction.

* Extracting features from 1260 images...done. Extracted 19446 features.

* Keeping 80 percent of the strongest features from each category.

* Balancing the number of features across all image categories to improve clustering.
** Image category 2 has the least number of strongest features: 2074.
** Using the strongest 2074 features from each of the other image categories.

* Using K-Means clustering to create a 50 word visual vocabulary.
* Number of features          : 6222
* Number of clusters (K)      : 50

* Initializing cluster centers...100.00%.
* Clustering...completed 19/100 iterations (~0.04 seconds/iteration)...converged in 19 iterations.

* Finished creating Bag-Of-Features

-----------------------------------------------------------------
Training an image categiry classifier

Training an image category classifier for 3 categories.
--------------------------------------------------------
* Category 1: BuiltUp
* Category 2: NonBuiltUp
* Category 3: Slum

* Encoding features for 1260 images...done.

* Finished training the category classifier. Use evaluate to test the classifier on a test set.

-----------------------------------------------------------------
Evaluating perfomance on the Training set

Evaluating image category classifier for 3 categories.
-------------------------------------------------------

* Category 1: BuiltUp
* Category 2: NonBuiltUp
* Category 3: Slum

* Evaluating 1260 images...done.

* Finished evaluating all the test sets.

* The confusion matrix for this test set is:

                          PREDICTED
KNOWN         | BuiltUp   NonBuiltUp   Slum   
----------------------------------------------
BuiltUp       | 0.58      0.26         0.16   
NonBuiltUp    | 0.10      0.83         0.07   
Slum          | 0.11      0.06         0.83   

* Average Accuracy is 0.75.

                  accuracy    sensitivity    specificity    precision    recall    Fscore 
                  ________    ___________    ___________    _________    ______    _______

    BuiltUp       78.889      58.095         89.286         73.054       58.095    0.64721
    NonBuiltUp    83.571      82.619         84.048         72.141       82.619    0.77026
    Slum          86.905      83.333          88.69         78.652       83.333    0.80925

Evaluating perfomance on the Training set
-----------------------------------------------------------------
Evaluating perfomance on the Test set

Evaluating image category classifier for 3 categories.
-------------------------------------------------------

* Category 1: BuiltUp
* Category 2: NonBuiltUp
* Category 3: Slum

* Evaluating 315 images...done.

* Finished evaluating all the test sets.

* The confusion matrix for this test set is:

                          PREDICTED
KNOWN         | BuiltUp   NonBuiltUp   Slum   
----------------------------------------------
BuiltUp       | 0.54      0.28         0.18   
NonBuiltUp    | 0.09      0.76         0.15   
Slum          | 0.16      0.08         0.76   

* Average Accuracy is 0.69.

                  accuracy    sensitivity    specificity    precision    recall    Fscore 
                  ________    ___________    ___________    _________    ______    _______

    BuiltUp       76.508      54.286         87.619         68.675       54.286    0.60638
    NonBuiltUp    80.317       76.19         82.381         68.376        76.19    0.72072
    Slum          80.952       76.19         83.333         69.565        76.19    0.72727

Saving perfomance on the Test set
-----------------------------------------------------------------
**********************************************************************************

[elboyran]

Creating datastore for dataset 3:px200m30...

imds = 

  ImageDatastore with properties:

       Files: {
              ' ...\ROI1\BuiltUp\Bangalore_ROI1_tile_sr1001er1200sc2101ec2300.tif';
              ' ...\ROI1\BuiltUp\Bangalore_ROI1_tile_sr1001er1200sc2201ec2400.tif';
              ' ...\ROI1\BuiltUp\Bangalore_ROI1_tile_sr1001er1200sc2301ec2500.tif'
               ... and 13119 more
              }
      Labels: [BuiltUp; BuiltUp; BuiltUp ... and 13119 more categorical]
    ReadSize: 1
     ReadFcn: @readDatastoreImage

      Label       Count
    __________    _____

    BuiltUp       9151 
    NonBuiltUp    3807 
    Slum           164 

ans =

  3×2 table

      Label       Count
    __________    _____

    BuiltUp       164  
    NonBuiltUp    164  
    Slum          164  

-----------------------------------------------------------------
Splititng the datastore into Train and Test datastores with fractionTrain: 80%
Training datastore: 
      Label       Count
    __________    _____

    BuiltUp       131  
    NonBuiltUp    131  
    Slum          131  

Testing datastore: 
      Label       Count
    __________    _____

    BuiltUp       33   
    NonBuiltUp    33   
    Slum          33   

-----------------------------------------------------------------
Creating features (BoVW) using vocabulary size 10 and SUFR point locations Detector for the Train datastore keeping the 80% strongest features

Creating Bag-Of-Features.
-------------------------
* Image category 1: BuiltUp
* Image category 2: NonBuiltUp
* Image category 3: Slum
* Selecting feature point locations using the Detector method.
* Extracting SURF features from the selected feature point locations.
** detectSURFFeatures is used to detect key points for feature extraction.

* Extracting features from 393 images...done. Extracted 16946 features.

* Keeping 80 percent of the strongest features from each category.

* Balancing the number of features across all image categories to improve clustering.
** Image category 2 has the least number of strongest features: 1534.
** Using the strongest 1534 features from each of the other image categories.

* Using K-Means clustering to create a 10 word visual vocabulary.
* Number of features          : 4602
* Number of clusters (K)      : 10

* Initializing cluster centers...100.00%.
* Clustering...completed 25/100 iterations (~0.03 seconds/iteration)...converged in 25 iterations.

* Finished creating Bag-Of-Features

-----------------------------------------------------------------
Training an image categiry classifier

Training an image category classifier for 3 categories.
--------------------------------------------------------
* Category 1: BuiltUp
* Category 2: NonBuiltUp
* Category 3: Slum

* Encoding features for 393 images...done.

* Finished training the category classifier. Use evaluate to test the classifier on a test set.

-----------------------------------------------------------------
Evaluating perfomance on the Training set

Evaluating image category classifier for 3 categories.
-------------------------------------------------------

* Category 1: BuiltUp
* Category 2: NonBuiltUp
* Category 3: Slum

* Evaluating 393 images...done.

* Finished evaluating all the test sets.

* The confusion matrix for this test set is:

                          PREDICTED
KNOWN         | BuiltUp   NonBuiltUp   Slum   
----------------------------------------------
BuiltUp       | 0.73      0.11         0.15   
NonBuiltUp    | 0.19      0.66         0.15   
Slum          | 0.15      0.00         0.85   

* Average Accuracy is 0.75.

                  accuracy    sensitivity    specificity    precision    recall    Fscore 
                  ________    ___________    ___________    _________    ______    _______

    BuiltUp       79.898      73.282         83.206         68.571       73.282    0.70849
    NonBuiltUp    84.987      66.412         94.275         85.294       66.412    0.74678
    Slum          85.242      85.496         85.115         74.172       85.496    0.79433

Evaluating perfomance on the Training set
-----------------------------------------------------------------
Evaluating perfomance on the Test set

Evaluating image category classifier for 3 categories.
-------------------------------------------------------

* Category 1: BuiltUp
* Category 2: NonBuiltUp
* Category 3: Slum

* Evaluating 99 images...done.

* Finished evaluating all the test sets.

* The confusion matrix for this test set is:

                          PREDICTED
KNOWN         | BuiltUp   NonBuiltUp   Slum   
----------------------------------------------
BuiltUp       | 0.76      0.18         0.06   
NonBuiltUp    | 0.15      0.73         0.12   
Slum          | 0.09      0.00         0.91   

* Average Accuracy is 0.80.

                  accuracy    sensitivity    specificity    precision    recall    Fscore 
                  ________    ___________    ___________    _________    ______    _______

    BuiltUp       83.838      75.758         87.879         75.758       75.758    0.75758
    NonBuiltUp    84.848      72.727         90.909             80       72.727     0.7619
    Slum          90.909      90.909         90.909         83.333       90.909    0.86957

Saving perfomance on the Test set
-----------------------------------------------------------------
Creating features (BoVW) using vocabulary size 20 and SUFR point locations Detector for the Train datastore keeping the 80% strongest features

Creating Bag-Of-Features.
-------------------------
* Image category 1: BuiltUp
* Image category 2: NonBuiltUp
* Image category 3: Slum
* Selecting feature point locations using the Detector method.
* Extracting SURF features from the selected feature point locations.
** detectSURFFeatures is used to detect key points for feature extraction.

* Extracting features from 393 images...done. Extracted 16946 features.

* Keeping 80 percent of the strongest features from each category.

* Balancing the number of features across all image categories to improve clustering.
** Image category 2 has the least number of strongest features: 1534.
** Using the strongest 1534 features from each of the other image categories.

* Using K-Means clustering to create a 20 word visual vocabulary.
* Number of features          : 4602
* Number of clusters (K)      : 20

* Initializing cluster centers...100.00%.
* Clustering...completed 27/100 iterations (~0.03 seconds/iteration)...converged in 27 iterations.

* Finished creating Bag-Of-Features

-----------------------------------------------------------------
Training an image categiry classifier

Training an image category classifier for 3 categories.
--------------------------------------------------------
* Category 1: BuiltUp
* Category 2: NonBuiltUp
* Category 3: Slum

* Encoding features for 393 images...done.

* Finished training the category classifier. Use evaluate to test the classifier on a test set.

-----------------------------------------------------------------
Evaluating perfomance on the Training set

Evaluating image category classifier for 3 categories.
-------------------------------------------------------

* Category 1: BuiltUp
* Category 2: NonBuiltUp
* Category 3: Slum

* Evaluating 393 images...done.

* Finished evaluating all the test sets.

* The confusion matrix for this test set is:

                          PREDICTED
KNOWN         | BuiltUp   NonBuiltUp   Slum   
----------------------------------------------
BuiltUp       | 0.74      0.15         0.11   
NonBuiltUp    | 0.16      0.78         0.06   
Slum          | 0.15      0.01         0.85   

* Average Accuracy is 0.79.

                  accuracy    sensitivity    specificity    precision    recall    Fscore 
                  ________    ___________    ___________    _________    ______    _______

    BuiltUp        81.17      74.046         84.733         70.803       74.046    0.72388
    NonBuiltUp    87.532      77.863         92.366         83.607       77.863    0.80632
    Slum          89.059      84.733         91.221         82.836       84.733    0.83774

Evaluating perfomance on the Training set
-----------------------------------------------------------------
Evaluating perfomance on the Test set

Evaluating image category classifier for 3 categories.
-------------------------------------------------------

* Category 1: BuiltUp
* Category 2: NonBuiltUp
* Category 3: Slum

* Evaluating 99 images...done.

* Finished evaluating all the test sets.

* The confusion matrix for this test set is:

                          PREDICTED
KNOWN         | BuiltUp   NonBuiltUp   Slum   
----------------------------------------------
BuiltUp       | 0.64      0.18         0.18   
NonBuiltUp    | 0.18      0.76         0.06   
Slum          | 0.09      0.03         0.88   

* Average Accuracy is 0.76.

                  accuracy    sensitivity    specificity    precision    recall    Fscore 
                  ________    ___________    ___________    _________    ______    _______

    BuiltUp       78.788      63.636         86.364             70       63.636    0.66667
    NonBuiltUp    84.848      75.758         89.394         78.125       75.758    0.76923
    Slum          87.879      87.879         87.879         78.378       87.879    0.82857

Saving perfomance on the Test set
-----------------------------------------------------------------
Creating features (BoVW) using vocabulary size 50 and SUFR point locations Detector for the Train datastore keeping the 80% strongest features

Creating Bag-Of-Features.
-------------------------
* Image category 1: BuiltUp
* Image category 2: NonBuiltUp
* Image category 3: Slum
* Selecting feature point locations using the Detector method.
* Extracting SURF features from the selected feature point locations.
** detectSURFFeatures is used to detect key points for feature extraction.

* Extracting features from 393 images...done. Extracted 16946 features.

* Keeping 80 percent of the strongest features from each category.

* Balancing the number of features across all image categories to improve clustering.
** Image category 2 has the least number of strongest features: 1534.
** Using the strongest 1534 features from each of the other image categories.

* Using K-Means clustering to create a 50 word visual vocabulary.
* Number of features          : 4602
* Number of clusters (K)      : 50

* Initializing cluster centers...100.00%.
* Clustering...completed 21/100 iterations (~0.03 seconds/iteration)...converged in 21 iterations.

* Finished creating Bag-Of-Features

-----------------------------------------------------------------
Training an image categiry classifier

Training an image category classifier for 3 categories.
--------------------------------------------------------
* Category 1: BuiltUp
* Category 2: NonBuiltUp
* Category 3: Slum

* Encoding features for 393 images...done.

* Finished training the category classifier. Use evaluate to test the classifier on a test set.

-----------------------------------------------------------------
Evaluating perfomance on the Training set

Evaluating image category classifier for 3 categories.
-------------------------------------------------------

* Category 1: BuiltUp
* Category 2: NonBuiltUp
* Category 3: Slum

* Evaluating 393 images...done.

* Finished evaluating all the test sets.

* The confusion matrix for this test set is:

                          PREDICTED
KNOWN         | BuiltUp   NonBuiltUp   Slum   
----------------------------------------------
BuiltUp       | 0.78      0.13         0.09   
NonBuiltUp    | 0.10      0.82         0.08   
Slum          | 0.09      0.00         0.91   

* Average Accuracy is 0.84.

                  accuracy    sensitivity    specificity    precision    recall    Fscore 
                  ________    ___________    ___________    _________    ______    _______

    BuiltUp        86.26      77.863         90.458         80.315       77.863     0.7907
    NonBuiltUp    89.822      82.443         93.511           86.4       82.443    0.84375
    Slum          91.349       90.84         91.603         84.397        90.84      0.875

Evaluating perfomance on the Training set
-----------------------------------------------------------------
Evaluating perfomance on the Test set

Evaluating image category classifier for 3 categories.
-------------------------------------------------------

* Category 1: BuiltUp
* Category 2: NonBuiltUp
* Category 3: Slum

* Evaluating 99 images...done.

* Finished evaluating all the test sets.

* The confusion matrix for this test set is:

                          PREDICTED
KNOWN         | BuiltUp   NonBuiltUp   Slum   
----------------------------------------------
BuiltUp       | 0.67      0.21         0.12   
NonBuiltUp    | 0.09      0.79         0.12   
Slum          | 0.09      0.06         0.85   

* Average Accuracy is 0.77.

                  accuracy    sensitivity    specificity    precision    recall    Fscore 
                  ________    ___________    ___________    _________    ______    _______

    BuiltUp       82.828      66.667         90.909         78.571       66.667    0.72131
    NonBuiltUp    83.838      78.788         86.364         74.286       78.788    0.76471
    Slum          86.869      84.848         87.879         77.778       84.848    0.81159

Saving perfomance on the Test set
-----------------------------------------------------------------
**********************************************************************************

[elboyran]

Creating datastore for dataset 4:px268m40...

imds = 

  ImageDatastore with properties:

       Files: {
              ' ...\ROI1\BuiltUp\Bangalore_ROI1_tile_sr1073er1340sc1877ec2144.tif';
              ' ...\ROI1\BuiltUp\Bangalore_ROI1_tile_sr1073er1340sc2011ec2278.tif';
              ' ...\ROI1\BuiltUp\Bangalore_ROI1_tile_sr1073er1340sc2145ec2412.tif'
               ... and 6710 more
              }
      Labels: [BuiltUp; BuiltUp; BuiltUp ... and 6710 more categorical]
    ReadSize: 1
     ReadFcn: @readDatastoreImage

      Label       Count
    __________    _____

    BuiltUp       4805 
    NonBuiltUp    1848 
    Slum            60 

ans =

  3×2 table

      Label       Count
    __________    _____

    BuiltUp       60   
    NonBuiltUp    60   
    Slum          60   

-----------------------------------------------------------------
Splititng the datastore into Train and Test datastores with fractionTrain: 80%
Training datastore: 
      Label       Count
    __________    _____

    BuiltUp       48   
    NonBuiltUp    48   
    Slum          48   

Testing datastore: 
      Label       Count
    __________    _____

    BuiltUp       12   
    NonBuiltUp    12   
    Slum          12   

-----------------------------------------------------------------
Creating features (BoVW) using vocabulary size 10 and SUFR point locations Detector for the Train datastore keeping the 80% strongest features

Creating Bag-Of-Features.
-------------------------
* Image category 1: BuiltUp
* Image category 2: NonBuiltUp
* Image category 3: Slum
* Selecting feature point locations using the Detector method.
* Extracting SURF features from the selected feature point locations.
** detectSURFFeatures is used to detect key points for feature extraction.

* Extracting features from 144 images...done. Extracted 11882 features.

* Keeping 80 percent of the strongest features from each category.

* Balancing the number of features across all image categories to improve clustering.
** Image category 2 has the least number of strongest features: 1095.
** Using the strongest 1095 features from each of the other image categories.

* Using K-Means clustering to create a 10 word visual vocabulary.
* Number of features          : 3285
* Number of clusters (K)      : 10

* Initializing cluster centers...100.00%.
* Clustering...completed 18/100 iterations (~0.02 seconds/iteration)...converged in 18 iterations.

* Finished creating Bag-Of-Features

-----------------------------------------------------------------
Training an image categiry classifier

Training an image category classifier for 3 categories.
--------------------------------------------------------
* Category 1: BuiltUp
* Category 2: NonBuiltUp
* Category 3: Slum

* Encoding features for 144 images...done.

* Finished training the category classifier. Use evaluate to test the classifier on a test set.

-----------------------------------------------------------------
Evaluating perfomance on the Training set

Evaluating image category classifier for 3 categories.
-------------------------------------------------------

* Category 1: BuiltUp
* Category 2: NonBuiltUp
* Category 3: Slum

* Evaluating 144 images...done.

* Finished evaluating all the test sets.

* The confusion matrix for this test set is:

                          PREDICTED
KNOWN         | BuiltUp   NonBuiltUp   Slum   
----------------------------------------------
BuiltUp       | 0.71      0.15         0.15   
NonBuiltUp    | 0.23      0.54         0.23   
Slum          | 0.13      0.00         0.88   

* Average Accuracy is 0.71.

                  accuracy    sensitivity    specificity    precision    recall    Fscore 
                  ________    ___________    ___________    _________    ______    _______

    BuiltUp       78.472      70.833         82.292         66.667       70.833    0.68687
    NonBuiltUp    79.861      54.167         92.708         78.788       54.167    0.64198
    Slum          83.333        87.5          81.25             70         87.5    0.77778

Evaluating perfomance on the Training set
-----------------------------------------------------------------
Evaluating perfomance on the Test set

Evaluating image category classifier for 3 categories.
-------------------------------------------------------

* Category 1: BuiltUp
* Category 2: NonBuiltUp
* Category 3: Slum

* Evaluating 36 images...done.

* Finished evaluating all the test sets.

* The confusion matrix for this test set is:

                          PREDICTED
KNOWN         | BuiltUp   NonBuiltUp   Slum   
----------------------------------------------
BuiltUp       | 1.00      0.00         0.00   
NonBuiltUp    | 0.33      0.42         0.25   
Slum          | 0.25      0.00         0.75   

* Average Accuracy is 0.72.

                  accuracy    sensitivity    specificity    precision    recall    Fscore 
                  ________    ___________    ___________    _________    ______    _______

    BuiltUp       80.556         100         70.833         63.158          100    0.77419
    NonBuiltUp    80.556      41.667            100            100       41.667    0.58824
    Slum          83.333          75           87.5             75           75       0.75

Saving perfomance on the Test set
-----------------------------------------------------------------
Creating features (BoVW) using vocabulary size 20 and SUFR point locations Detector for the Train datastore keeping the 80% strongest features

Creating Bag-Of-Features.
-------------------------
* Image category 1: BuiltUp
* Image category 2: NonBuiltUp
* Image category 3: Slum
* Selecting feature point locations using the Detector method.
* Extracting SURF features from the selected feature point locations.
** detectSURFFeatures is used to detect key points for feature extraction.

* Extracting features from 144 images...done. Extracted 11882 features.

* Keeping 80 percent of the strongest features from each category.

* Balancing the number of features across all image categories to improve clustering.
** Image category 2 has the least number of strongest features: 1095.
** Using the strongest 1095 features from each of the other image categories.

* Using K-Means clustering to create a 20 word visual vocabulary.
* Number of features          : 3285
* Number of clusters (K)      : 20

* Initializing cluster centers...100.00%.
* Clustering...completed 21/100 iterations (~0.02 seconds/iteration)...converged in 21 iterations.

* Finished creating Bag-Of-Features

-----------------------------------------------------------------
Training an image categiry classifier

Training an image category classifier for 3 categories.
--------------------------------------------------------
* Category 1: BuiltUp
* Category 2: NonBuiltUp
* Category 3: Slum

* Encoding features for 144 images...done.

* Finished training the category classifier. Use evaluate to test the classifier on a test set.

-----------------------------------------------------------------
Evaluating perfomance on the Training set

Evaluating image category classifier for 3 categories.
-------------------------------------------------------

* Category 1: BuiltUp
* Category 2: NonBuiltUp
* Category 3: Slum

* Evaluating 144 images...done.

* Finished evaluating all the test sets.

* The confusion matrix for this test set is:

                          PREDICTED
KNOWN         | BuiltUp   NonBuiltUp   Slum   
----------------------------------------------
BuiltUp       | 0.73      0.21         0.06   
NonBuiltUp    | 0.17      0.73         0.10   
Slum          | 0.10      0.00         0.90   

* Average Accuracy is 0.78.

                  accuracy    sensitivity    specificity    precision    recall    Fscore 
                  ________    ___________    ___________    _________    ______    _______

    BuiltUp       81.944      72.917         86.458         72.917       72.917    0.72917
    NonBuiltUp    84.028      72.917         89.583         77.778       72.917    0.75269
    Slum          90.972      89.583         91.667         84.314       89.583    0.86869

Evaluating perfomance on the Training set
-----------------------------------------------------------------
Evaluating perfomance on the Test set

Evaluating image category classifier for 3 categories.
-------------------------------------------------------

* Category 1: BuiltUp
* Category 2: NonBuiltUp
* Category 3: Slum

* Evaluating 36 images...done.

* Finished evaluating all the test sets.

* The confusion matrix for this test set is:

                          PREDICTED
KNOWN         | BuiltUp   NonBuiltUp   Slum   
----------------------------------------------
BuiltUp       | 1.00      0.00         0.00   
NonBuiltUp    | 0.17      0.75         0.08   
Slum          | 0.17      0.00         0.83   

* Average Accuracy is 0.86.

                  accuracy    sensitivity    specificity    precision    recall    Fscore 
                  ________    ___________    ___________    _________    ______    _______

    BuiltUp       88.889         100         83.333             75          100    0.85714
    NonBuiltUp    91.667          75            100            100           75    0.85714
    Slum          91.667      83.333         95.833         90.909       83.333    0.86957

Saving perfomance on the Test set
-----------------------------------------------------------------
Creating features (BoVW) using vocabulary size 50 and SUFR point locations Detector for the Train datastore keeping the 80% strongest features

Creating Bag-Of-Features.
-------------------------
* Image category 1: BuiltUp
* Image category 2: NonBuiltUp
* Image category 3: Slum
* Selecting feature point locations using the Detector method.
* Extracting SURF features from the selected feature point locations.
** detectSURFFeatures is used to detect key points for feature extraction.

* Extracting features from 144 images...done. Extracted 11882 features.

* Keeping 80 percent of the strongest features from each category.

* Balancing the number of features across all image categories to improve clustering.
** Image category 2 has the least number of strongest features: 1095.
** Using the strongest 1095 features from each of the other image categories.

* Using K-Means clustering to create a 50 word visual vocabulary.
* Number of features          : 3285
* Number of clusters (K)      : 50

* Initializing cluster centers...100.00%.
* Clustering...completed 26/100 iterations (~0.02 seconds/iteration)...converged in 26 iterations.

* Finished creating Bag-Of-Features

-----------------------------------------------------------------
Training an image categiry classifier

Training an image category classifier for 3 categories.
--------------------------------------------------------
* Category 1: BuiltUp
* Category 2: NonBuiltUp
* Category 3: Slum

* Encoding features for 144 images...done.

* Finished training the category classifier. Use evaluate to test the classifier on a test set.

-----------------------------------------------------------------
Evaluating perfomance on the Training set

Evaluating image category classifier for 3 categories.
-------------------------------------------------------

* Category 1: BuiltUp
* Category 2: NonBuiltUp
* Category 3: Slum

* Evaluating 144 images...done.

* Finished evaluating all the test sets.

* The confusion matrix for this test set is:

                          PREDICTED
KNOWN         | BuiltUp   NonBuiltUp   Slum   
----------------------------------------------
BuiltUp       | 0.75      0.19         0.06   
NonBuiltUp    | 0.10      0.90         0.00   
Slum          | 0.10      0.00         0.90   

* Average Accuracy is 0.85.

                  accuracy    sensitivity    specificity    precision    recall    Fscore 
                  ________    ___________    ___________    _________    ______    _______

    BuiltUp       84.722          75         89.583         78.261           75    0.76596
    NonBuiltUp    90.278      89.583         90.625         82.692       89.583       0.86
    Slum          94.444      89.583         96.875         93.478       89.583    0.91489

Evaluating perfomance on the Training set
-----------------------------------------------------------------
Evaluating perfomance on the Test set

Evaluating image category classifier for 3 categories.
-------------------------------------------------------

* Category 1: BuiltUp
* Category 2: NonBuiltUp
* Category 3: Slum

* Evaluating 36 images...done.

* Finished evaluating all the test sets.

* The confusion matrix for this test set is:

                          PREDICTED
KNOWN         | BuiltUp   NonBuiltUp   Slum   
----------------------------------------------
BuiltUp       | 1.00      0.00         0.00   
NonBuiltUp    | 0.17      0.83         0.00   
Slum          | 0.08      0.00         0.92   

* Average Accuracy is 0.92.

                  accuracy    sensitivity    specificity    precision    recall    Fscore 
                  ________    ___________    ___________    _________    ______    _______

    BuiltUp       91.667         100         87.5            80             100    0.88889
    NonBuiltUp    94.444      83.333          100           100          83.333    0.90909
    Slum          97.222      91.667          100           100          91.667    0.95652

Saving perfomance on the Test set
-----------------------------------------------------------------
**********************************************************************************

[elboyran]

Creating datastore for dataset 5:px334m50...

imds = 

  ImageDatastore with properties:

       Files: {
              ' ...\ROI1\BuiltUp\Bangalore_ROI1_tile_sr1003er1336sc2005ec2338.tif';
              ' ...\ROI1\BuiltUp\Bangalore_ROI1_tile_sr1003er1336sc2172ec2505.tif';
              ' ...\ROI1\BuiltUp\Bangalore_ROI1_tile_sr1003er1336sc2339ec2672.tif'
               ... and 4017 more
              }
      Labels: [BuiltUp; BuiltUp; BuiltUp ... and 4017 more categorical]
    ReadSize: 1
     ReadFcn: @readDatastoreImage

      Label       Count
    __________    _____

    BuiltUp       2902 
    NonBuiltUp    1088 
    Slum            30 

ans =

  3×2 table

      Label       Count
    __________    _____

    BuiltUp       30   
    NonBuiltUp    30   
    Slum          30   

-----------------------------------------------------------------
Splititng the datastore into Train and Test datastores with fractionTrain: 80%
Training datastore: 
      Label       Count
    __________    _____

    BuiltUp       24   
    NonBuiltUp    24   
    Slum          24   

Testing datastore: 
      Label       Count
    __________    _____

    BuiltUp       6    
    NonBuiltUp    6    
    Slum          6    

-----------------------------------------------------------------
Creating features (BoVW) using vocabulary size 10 and SUFR point locations Detector for the Train datastore keeping the 80% strongest features

Creating Bag-Of-Features.
-------------------------
* Image category 1: BuiltUp
* Image category 2: NonBuiltUp
* Image category 3: Slum
* Selecting feature point locations using the Detector method.
* Extracting SURF features from the selected feature point locations.
** detectSURFFeatures is used to detect key points for feature extraction.

* Extracting features from 72 images...done. Extracted 11853 features.

* Keeping 80 percent of the strongest features from each category.

* Balancing the number of features across all image categories to improve clustering.
** Image category 2 has the least number of strongest features: 1193.
** Using the strongest 1193 features from each of the other image categories.

* Using K-Means clustering to create a 10 word visual vocabulary.
* Number of features          : 3579
* Number of clusters (K)      : 10

* Initializing cluster centers...100.00%.
* Clustering...completed 18/100 iterations (~0.02 seconds/iteration)...converged in 18 iterations.

* Finished creating Bag-Of-Features

-----------------------------------------------------------------
Training an image categiry classifier

Training an image category classifier for 3 categories.
--------------------------------------------------------
* Category 1: BuiltUp
* Category 2: NonBuiltUp
* Category 3: Slum

* Encoding features for 72 images...done.

* Finished training the category classifier. Use evaluate to test the classifier on a test set.

-----------------------------------------------------------------
Evaluating perfomance on the Training set

Evaluating image category classifier for 3 categories.
-------------------------------------------------------

* Category 1: BuiltUp
* Category 2: NonBuiltUp
* Category 3: Slum

* Evaluating 72 images...done.

* Finished evaluating all the test sets.

* The confusion matrix for this test set is:

                          PREDICTED
KNOWN         | BuiltUp   NonBuiltUp   Slum   
----------------------------------------------
BuiltUp       | 0.92      0.08         0.00   
NonBuiltUp    | 0.21      0.67         0.13   
Slum          | 0.17      0.00         0.83   

* Average Accuracy is 0.81.

                  accuracy    sensitivity    specificity    precision    recall    Fscore 
                  ________    ___________    ___________    _________    ______    _______

    BuiltUp       84.722      91.667          81.25         70.968       91.667        0.8
    NonBuiltUp    86.111      66.667         95.833         88.889       66.667     0.7619
    Slum          90.278      83.333          93.75         86.957       83.333    0.85106

Evaluating perfomance on the Training set
-----------------------------------------------------------------
Evaluating perfomance on the Test set

Evaluating image category classifier for 3 categories.
-------------------------------------------------------

* Category 1: BuiltUp
* Category 2: NonBuiltUp
* Category 3: Slum

* Evaluating 18 images...done.

* Finished evaluating all the test sets.

* The confusion matrix for this test set is:

                          PREDICTED
KNOWN         | BuiltUp   NonBuiltUp   Slum   
----------------------------------------------
BuiltUp       | 0.50      0.17         0.33   
NonBuiltUp    | 0.00      0.50         0.50   
Slum          | 0.00      0.00         1.00   

* Average Accuracy is 0.67.

                  accuracy    sensitivity    specificity    precision    recall    Fscore 
                  ________    ___________    ___________    _________    ______    _______

    BuiltUp       83.333       50               100            100        50       0.66667
    NonBuiltUp    77.778       50            91.667             75        50           0.6
    Slum          72.222      100            58.333         54.545       100       0.70588

Saving perfomance on the Test set
-----------------------------------------------------------------
Creating features (BoVW) using vocabulary size 20 and SUFR point locations Detector for the Train datastore keeping the 80% strongest features

Creating Bag-Of-Features.
-------------------------
* Image category 1: BuiltUp
* Image category 2: NonBuiltUp
* Image category 3: Slum
* Selecting feature point locations using the Detector method.
* Extracting SURF features from the selected feature point locations.
** detectSURFFeatures is used to detect key points for feature extraction.

* Extracting features from 72 images...done. Extracted 11853 features.

* Keeping 80 percent of the strongest features from each category.

* Balancing the number of features across all image categories to improve clustering.
** Image category 2 has the least number of strongest features: 1193.
** Using the strongest 1193 features from each of the other image categories.

* Using K-Means clustering to create a 20 word visual vocabulary.
* Number of features          : 3579
* Number of clusters (K)      : 20

* Initializing cluster centers...100.00%.
* Clustering...completed 22/100 iterations (~0.02 seconds/iteration)...converged in 22 iterations.

* Finished creating Bag-Of-Features

-----------------------------------------------------------------
Training an image categiry classifier

Training an image category classifier for 3 categories.
--------------------------------------------------------
* Category 1: BuiltUp
* Category 2: NonBuiltUp
* Category 3: Slum

* Encoding features for 72 images...done.

* Finished training the category classifier. Use evaluate to test the classifier on a test set.

-----------------------------------------------------------------
Evaluating perfomance on the Training set

Evaluating image category classifier for 3 categories.
-------------------------------------------------------

* Category 1: BuiltUp
* Category 2: NonBuiltUp
* Category 3: Slum

* Evaluating 72 images...done.

* Finished evaluating all the test sets.

* The confusion matrix for this test set is:

                          PREDICTED
KNOWN         | BuiltUp   NonBuiltUp   Slum   
----------------------------------------------
BuiltUp       | 0.92      0.08         0.00   
NonBuiltUp    | 0.13      0.83         0.04   
Slum          | 0.13      0.00         0.88   

* Average Accuracy is 0.88.

                  accuracy    sensitivity    specificity    precision    recall    Fscore 
                  ________    ___________    ___________    _________    ______    _______

    BuiltUp       88.889      91.667           87.5         78.571       91.667    0.84615
    NonBuiltUp    91.667      83.333         95.833         90.909       83.333    0.86957
    Slum          94.444        87.5         97.917         95.455         87.5    0.91304

Evaluating perfomance on the Training set
-----------------------------------------------------------------
Evaluating perfomance on the Test set

Evaluating image category classifier for 3 categories.
-------------------------------------------------------

* Category 1: BuiltUp
* Category 2: NonBuiltUp
* Category 3: Slum

* Evaluating 18 images...done.

* Finished evaluating all the test sets.

* The confusion matrix for this test set is:

                          PREDICTED
KNOWN         | BuiltUp   NonBuiltUp   Slum   
----------------------------------------------
BuiltUp       | 0.67      0.17         0.17   
NonBuiltUp    | 0.17      0.83         0.00   
Slum          | 0.00      0.00         1.00   

* Average Accuracy is 0.83.

                  accuracy    sensitivity    specificity    precision    recall    Fscore 
                  ________    ___________    ___________    _________    ______    _______

    BuiltUp       83.333      66.667         91.667             80       66.667    0.72727
    NonBuiltUp    88.889      83.333         91.667         83.333       83.333    0.83333
    Slum          94.444         100         91.667         85.714          100    0.92308

Saving perfomance on the Test set
-----------------------------------------------------------------
Creating features (BoVW) using vocabulary size 50 and SUFR point locations Detector for the Train datastore keeping the 80% strongest features

Creating Bag-Of-Features.
-------------------------
* Image category 1: BuiltUp
* Image category 2: NonBuiltUp
* Image category 3: Slum
* Selecting feature point locations using the Detector method.
* Extracting SURF features from the selected feature point locations.
** detectSURFFeatures is used to detect key points for feature extraction.

* Extracting features from 72 images...done. Extracted 11853 features.

* Keeping 80 percent of the strongest features from each category.

* Balancing the number of features across all image categories to improve clustering.
** Image category 2 has the least number of strongest features: 1193.
** Using the strongest 1193 features from each of the other image categories.

* Using K-Means clustering to create a 50 word visual vocabulary.
* Number of features          : 3579
* Number of clusters (K)      : 50

* Initializing cluster centers...100.00%.
* Clustering...completed 20/100 iterations (~0.03 seconds/iteration)...converged in 20 iterations.

* Finished creating Bag-Of-Features

-----------------------------------------------------------------
Training an image categiry classifier

Training an image category classifier for 3 categories.
--------------------------------------------------------
* Category 1: BuiltUp
* Category 2: NonBuiltUp
* Category 3: Slum

* Encoding features for 72 images...done.

* Finished training the category classifier. Use evaluate to test the classifier on a test set.

-----------------------------------------------------------------
Evaluating perfomance on the Training set

Evaluating image category classifier for 3 categories.
-------------------------------------------------------

* Category 1: BuiltUp
* Category 2: NonBuiltUp
* Category 3: Slum

* Evaluating 72 images...done.

* Finished evaluating all the test sets.

* The confusion matrix for this test set is:

                          PREDICTED
KNOWN         | BuiltUp   NonBuiltUp   Slum   
----------------------------------------------
BuiltUp       | 0.92      0.08         0.00   
NonBuiltUp    | 0.08      0.92         0.00   
Slum          | 0.13      0.00         0.88   

* Average Accuracy is 0.90.

                  accuracy    sensitivity    specificity    precision    recall    Fscore 
                  ________    ___________    ___________    _________    ______    _______

    BuiltUp       90.278      91.667         89.583         81.481       91.667    0.86275
    NonBuiltUp    94.444      91.667         95.833         91.667       91.667    0.91667
    Slum          95.833        87.5            100            100         87.5    0.93333

Evaluating perfomance on the Training set
-----------------------------------------------------------------
Evaluating perfomance on the Test set

Evaluating image category classifier for 3 categories.
-------------------------------------------------------

* Category 1: BuiltUp
* Category 2: NonBuiltUp
* Category 3: Slum

* Evaluating 18 images...done.

* Finished evaluating all the test sets.

* The confusion matrix for this test set is:

                          PREDICTED
KNOWN         | BuiltUp   NonBuiltUp   Slum   
----------------------------------------------
BuiltUp       | 0.50      0.33         0.17   
NonBuiltUp    | 0.17      0.83         0.00   
Slum          | 0.00      0.00         1.00   

* Average Accuracy is 0.78.

                  accuracy    sensitivity    specificity    precision    recall    Fscore 
                  ________    ___________    ___________    _________    ______    _______

    BuiltUp       77.778          50         91.667             75           50        0.6
    NonBuiltUp    83.333      83.333         83.333         71.429       83.333    0.76923
    Slum          94.444         100         91.667         85.714          100    0.92308

Saving perfomance on the Test set
-----------------------------------------------------------------
**********************************************************************************

[elboyran]

Creating datastore for dataset 6:px400m60...

imds = 

  ImageDatastore with properties:

       Files: {
              ' ...\ROI1\BuiltUp\Bangalore_ROI1_tile_sr1001er1400sc2001ec2400.tif';
              ' ...\ROI1\BuiltUp\Bangalore_ROI1_tile_sr1001er1400sc2201ec2600.tif';
              ' ...\ROI1\BuiltUp\Bangalore_ROI1_tile_sr1001er1400sc2401ec2800.tif'
               ... and 2648 more
              }
      Labels: [BuiltUp; BuiltUp; BuiltUp ... and 2648 more categorical]
    ReadSize: 1
     ReadFcn: @readDatastoreImage

      Label       Count
    __________    _____

    BuiltUp       1937 
    NonBuiltUp     703 
    Slum            11 

ans =

  3×2 table

      Label       Count
    __________    _____

    BuiltUp       11   
    NonBuiltUp    11   
    Slum          11   

-----------------------------------------------------------------
Splititng the datastore into Train and Test datastores with fractionTrain: 80%
Training datastore: 
      Label       Count
    __________    _____

    BuiltUp       9    
    NonBuiltUp    9    
    Slum          9    

Testing datastore: 
      Label       Count
    __________    _____

    BuiltUp       2    
    NonBuiltUp    2    
    Slum          2    

-----------------------------------------------------------------
Creating features (BoVW) using vocabulary size 10 and SUFR point locations Detector for the Train datastore keeping the 80% strongest features

Creating Bag-Of-Features.
-------------------------
* Image category 1: BuiltUp
* Image category 2: NonBuiltUp
* Image category 3: Slum
* Selecting feature point locations using the Detector method.
* Extracting SURF features from the selected feature point locations.
** detectSURFFeatures is used to detect key points for feature extraction.

* Extracting features from 27 images...done. Extracted 6215 features.

* Keeping 80 percent of the strongest features from each category.

* Balancing the number of features across all image categories to improve clustering.
** Image category 2 has the least number of strongest features: 841.
** Using the strongest 841 features from each of the other image categories.

* Using K-Means clustering to create a 10 word visual vocabulary.
* Number of features          : 2523
* Number of clusters (K)      : 10

* Initializing cluster centers...100.00%.
* Clustering...completed 21/100 iterations (~0.01 seconds/iteration)...converged in 21 iterations.

* Finished creating Bag-Of-Features

-----------------------------------------------------------------
Training an image categiry classifier

Training an image category classifier for 3 categories.
--------------------------------------------------------
* Category 1: BuiltUp
* Category 2: NonBuiltUp
* Category 3: Slum

* Encoding features for 27 images...done.

* Finished training the category classifier. Use evaluate to test the classifier on a test set.

-----------------------------------------------------------------
Evaluating perfomance on the Training set

Evaluating image category classifier for 3 categories.
-------------------------------------------------------

* Category 1: BuiltUp
* Category 2: NonBuiltUp
* Category 3: Slum

* Evaluating 27 images...done.

* Finished evaluating all the test sets.

* The confusion matrix for this test set is:

                          PREDICTED
KNOWN         | BuiltUp   NonBuiltUp   Slum   
----------------------------------------------
BuiltUp       | 0.78      0.11         0.11   
NonBuiltUp    | 0.22      0.67         0.11   
Slum          | 0.11      0.00         0.89   

* Average Accuracy is 0.78.

                  accuracy    sensitivity    specificity    precision    recall    Fscore 
                  ________    ___________    ___________    _________    ______    _______

    BuiltUp       81.481      77.778         83.333             70       77.778    0.73684
    NonBuiltUp    85.185      66.667         94.444         85.714       66.667       0.75
    Slum          88.889      88.889         88.889             80       88.889    0.84211

Evaluating perfomance on the Training set
-----------------------------------------------------------------
Evaluating perfomance on the Test set

Evaluating image category classifier for 3 categories.
-------------------------------------------------------

* Category 1: BuiltUp
* Category 2: NonBuiltUp
* Category 3: Slum

* Evaluating 6 images...done.

* Finished evaluating all the test sets.

* The confusion matrix for this test set is:

                          PREDICTED
KNOWN         | BuiltUp   NonBuiltUp   Slum   
----------------------------------------------
BuiltUp       | 1.00      0.00         0.00   
NonBuiltUp    | 0.00      0.50         0.50   
Slum          | 0.50      0.00         0.50   

* Average Accuracy is 0.67.

                  accuracy    sensitivity    specificity    precision    recall    Fscore 
                  ________    ___________    ___________    _________    ______    _______

    BuiltUp       83.333      100             75            66.667       100           0.8
    NonBuiltUp    83.333       50            100               100        50       0.66667
    Slum          66.667       50             75                50        50           0.5

Saving perfomance on the Test set
-----------------------------------------------------------------
Creating features (BoVW) using vocabulary size 20 and SUFR point locations Detector for the Train datastore keeping the 80% strongest features

Creating Bag-Of-Features.
-------------------------
* Image category 1: BuiltUp
* Image category 2: NonBuiltUp
* Image category 3: Slum
* Selecting feature point locations using the Detector method.
* Extracting SURF features from the selected feature point locations.
** detectSURFFeatures is used to detect key points for feature extraction.

* Extracting features from 27 images...done. Extracted 6215 features.

* Keeping 80 percent of the strongest features from each category.

* Balancing the number of features across all image categories to improve clustering.
** Image category 2 has the least number of strongest features: 841.
** Using the strongest 841 features from each of the other image categories.

* Using K-Means clustering to create a 20 word visual vocabulary.
* Number of features          : 2523
* Number of clusters (K)      : 20

* Initializing cluster centers...100.00%.
* Clustering...completed 24/100 iterations (~0.02 seconds/iteration)...converged in 24 iterations.

* Finished creating Bag-Of-Features

-----------------------------------------------------------------
Training an image categiry classifier

Training an image category classifier for 3 categories.
--------------------------------------------------------
* Category 1: BuiltUp
* Category 2: NonBuiltUp
* Category 3: Slum

* Encoding features for 27 images...done.

* Finished training the category classifier. Use evaluate to test the classifier on a test set.

-----------------------------------------------------------------
Evaluating perfomance on the Training set

Evaluating image category classifier for 3 categories.
-------------------------------------------------------

* Category 1: BuiltUp
* Category 2: NonBuiltUp
* Category 3: Slum

* Evaluating 27 images...done.

* Finished evaluating all the test sets.

* The confusion matrix for this test set is:

                          PREDICTED
KNOWN         | BuiltUp   NonBuiltUp   Slum   
----------------------------------------------
BuiltUp       | 0.78      0.11         0.11   
NonBuiltUp    | 0.11      0.67         0.22   
Slum          | 0.11      0.00         0.89   

* Average Accuracy is 0.78.

                  accuracy    sensitivity    specificity    precision    recall    Fscore 
                  ________    ___________    ___________    _________    ______    _______

    BuiltUp       85.185      77.778         88.889         77.778       77.778    0.77778
    NonBuiltUp    85.185      66.667         94.444         85.714       66.667       0.75
    Slum          85.185      88.889         83.333         72.727       88.889        0.8

Evaluating perfomance on the Training set
-----------------------------------------------------------------
Evaluating perfomance on the Test set

Evaluating image category classifier for 3 categories.
-------------------------------------------------------

* Category 1: BuiltUp
* Category 2: NonBuiltUp
* Category 3: Slum

* Evaluating 6 images...done.

* Finished evaluating all the test sets.

* The confusion matrix for this test set is:

                          PREDICTED
KNOWN         | BuiltUp   NonBuiltUp   Slum   
----------------------------------------------
BuiltUp       | 1.00      0.00         0.00   
NonBuiltUp    | 0.00      0.00         1.00   
Slum          | 0.50      0.00         0.50   

* Average Accuracy is 0.50.

                  accuracy    sensitivity    specificity    precision    recall    Fscore
                  ________    ___________    ___________    _________    ______    ______

    BuiltUp       83.333      100             75            66.667       100       0.8   
    NonBuiltUp    66.667        0            100               NaN         0         0   
    Slum              50       50             50            33.333        50       0.4   

Saving perfomance on the Test set
-----------------------------------------------------------------
Creating features (BoVW) using vocabulary size 50 and SUFR point locations Detector for the Train datastore keeping the 80% strongest features

Creating Bag-Of-Features.
-------------------------
* Image category 1: BuiltUp
* Image category 2: NonBuiltUp
* Image category 3: Slum
* Selecting feature point locations using the Detector method.
* Extracting SURF features from the selected feature point locations.
** detectSURFFeatures is used to detect key points for feature extraction.

* Extracting features from 27 images...done. Extracted 6215 features.

* Keeping 80 percent of the strongest features from each category.

* Balancing the number of features across all image categories to improve clustering.
** Image category 2 has the least number of strongest features: 841.
** Using the strongest 841 features from each of the other image categories.

* Using K-Means clustering to create a 50 word visual vocabulary.
* Number of features          : 2523
* Number of clusters (K)      : 50

* Initializing cluster centers...100.00%.
* Clustering...completed 21/100 iterations (~0.02 seconds/iteration)...converged in 21 iterations.

* Finished creating Bag-Of-Features

-----------------------------------------------------------------
Training an image categiry classifier

Training an image category classifier for 3 categories.
--------------------------------------------------------
* Category 1: BuiltUp
* Category 2: NonBuiltUp
* Category 3: Slum

* Encoding features for 27 images...done.

* Finished training the category classifier. Use evaluate to test the classifier on a test set.

-----------------------------------------------------------------
Evaluating perfomance on the Training set

Evaluating image category classifier for 3 categories.
-------------------------------------------------------

* Category 1: BuiltUp
* Category 2: NonBuiltUp
* Category 3: Slum

* Evaluating 27 images...done.

* Finished evaluating all the test sets.

* The confusion matrix for this test set is:

                          PREDICTED
KNOWN         | BuiltUp   NonBuiltUp   Slum   
----------------------------------------------
BuiltUp       | 0.89      0.11         0.00   
NonBuiltUp    | 0.11      0.89         0.00   
Slum          | 0.11      0.00         0.89   

* Average Accuracy is 0.89.

                  accuracy    sensitivity    specificity    precision    recall    Fscore 
                  ________    ___________    ___________    _________    ______    _______

    BuiltUp       88.889      88.889         88.889             80       88.889    0.84211
    NonBuiltUp    92.593      88.889         94.444         88.889       88.889    0.88889
    Slum          96.296      88.889            100            100       88.889    0.94118

Evaluating perfomance on the Training set
-----------------------------------------------------------------
Evaluating perfomance on the Test set

Evaluating image category classifier for 3 categories.
-------------------------------------------------------

* Category 1: BuiltUp
* Category 2: NonBuiltUp
* Category 3: Slum

* Evaluating 6 images...done.

* Finished evaluating all the test sets.

* The confusion matrix for this test set is:

                          PREDICTED
KNOWN         | BuiltUp   NonBuiltUp   Slum   
----------------------------------------------
BuiltUp       | 1.00      0.00         0.00   
NonBuiltUp    | 0.00      0.50         0.50   
Slum          | 0.50      0.00         0.50   

* Average Accuracy is 0.67.

                  accuracy    sensitivity    specificity    precision    recall    Fscore 
                  ________    ___________    ___________    _________    ______    _______

    BuiltUp       83.333      100             75            66.667       100           0.8
    NonBuiltUp    83.333       50            100               100        50       0.66667
    Slum          66.667       50             75                50        50           0.5

Saving perfomance on the Test set
-----------------------------------------------------------------
**********************************************************************************