Ask Your Question

Revision history [back]

click to hide/show revision 1
initial version

LandscapeClassificationMatrix

I am running Windows Vista and the latest SimpleCV version. Now I am running MLTestSuite with my own dataset. Since I started to work through more insistently with SimpleCV this issue arises:

I am training my dataset according to SVMClassifier in MLTestSuite: - with three paths and three classes e.g.: beach, forest, waterbody ...

print('Train')
classifierSVMP = SVMClassifier(extractors,props)
classifierSVMP.train(path,classes, savedata = "/Users/Arenzky/Desktop/testtab/",disp=display,subset= n) #train

Then I have a test dataset. The test dataset has paths to images, but which have no defined classes. They do no belong to a specific feature, such as forest or waterbody...e.g folder1, folder2,...

QUESTION 1:

Lets say we have 15 images from the test path. How many images (from folder1,2 or 3)can be classified as beach, how many as waterbody and how many as forest based on my trained dataset?

QUESTION 2:

    print('Test')
[pos,neg,confuse] = classifierSVMP.test(path,classes,savedata = "/Users/Arenzky/Desktop/testtab/", disp=display,subset=n)
files = []

Further, to my knowledge the "confuse" displays a so-called confusion matrix. Is there a way to have a matrix combining classes fom the train dataset and from test dataset such as:

/       beach forest water_body (train)
folder1

folder2

folder3
(test)

I am not sure if I am on the right track, but the MLTestSuite seemed to me the most feasible Template to work with. Can anybody give me some help or any tip?

thanks Daniel

LandscapeClassificationMatrix

I am running Windows Vista and the latest SimpleCV version. Now I am running MLTestSuite with my own dataset. Since I started to work through more insistently with SimpleCV this issue arises:

I am training my dataset according to SVMClassifier in MLTestSuite: - with three paths and three classes e.g.: beach, forest, waterbody ...

print('Train')
classifierSVMP = SVMClassifier(extractors,props)
classifierSVMP.train(path,classes, savedata = "/Users/Arenzky/Desktop/testtab/",disp=display,subset= n) #train

Then I have a test dataset. The test dataset has paths to images, but which have no defined classes. They do no belong to a specific feature, such as forest or waterbody...e.g folder1, folder2,...

QUESTION 1:

Lets say we have 15 images from the test path. How many images (from folder1,2 or 3)can be classified as beach, how many as waterbody and how many as forest based on my trained dataset?

QUESTION 2:

    print('Test')
[pos,neg,confuse] = classifierSVMP.test(path,classes,savedata = "/Users/Arenzky/Desktop/testtab/", disp=display,subset=n)
files = []

Further, to my knowledge the "confuse" displays a so-called confusion matrix. Is there a way to have a matrix combining classes fom the train dataset and from test dataset such as:

/       beach forest water_body (train)
folder1

folder2

folder3
(test)

I am not sure if I am on the right track, but the MLTestSuite seemed to me the most feasible Template to work with. Can anybody give me some help or any tip?

thanks Daniel

LandscapeClassificationMatrix

I am running Windows Vista and the latest SimpleCV version. Now I am running MLTestSuite with my own dataset. Since I started to work through more insistently with SimpleCV this issue arises:

I am training my dataset according to SVMClassifier in MLTestSuite: - with three paths and three classes e.g.: beach, forest, waterbody ...

print('Train')
classifierSVMP = SVMClassifier(extractors,props)
classifierSVMP.train(path,classes, savedata = "/Users/Arenzky/Desktop/testtab/",disp=display,subset= n) #train

Then I have a test dataset. The test dataset has paths to images, but which have no defined classes. They do no belong to a specific feature, such as forest or waterbody...e.g folder1, folder2,...

QUESTION 1:

Lets say we have 15 images from the test path. How many images (from folder1,2 or 3)can be classified as beach, how many as waterbody and how many as forest based on my trained dataset?

QUESTION 2:

    print('Test')
[pos,neg,confuse] = classifierSVMP.test(path,classes,savedata = "/Users/Arenzky/Desktop/testtab/", disp=display,subset=n)
files = []

Further, to my knowledge the "confuse" displays a so-called confusion matrix. Is there a way to have a matrix combining classes fom the train dataset and from test dataset such as:

/       beach forest water_body (train)
folder1

folder2

folder3
(test)

I am not sure if I am on the right track, but the MLTestSuite seemed to me the most feasible Template to work with. Can anybody give me some help or any tip?

thanks Daniel