Hello everybody,
I need to do a classification on a dataset of some images. I just have images and need to make a dataset of some features. I wrote the code of feature extraction in matlab but I don't know how to create a dataset using the results. I want to use ID3 decision tree as my classifier in WEKA.
Thank you so much for your help.
you can create an array with each row corresponding to a vectorized feature (converting the 2d feature vector into one or more 1D feature vectors) for each image. number of rows will be the number of data points (images),
you can create an array with each row corresponding to a vectorized feature (converting the 2d feature vector into one or more 1D feature vectors) for each image. number of rows will be the number of data points (images),
You can just create a matrix in which each row represents the image in one dimension (may be extracted features) and each column contains the total number of images.
You could also create a dataset in which each row corresponds to a sample image and each column to a different attribute computed with your matlab program. Then you add the attribute names as the first row of your dataset. Finally, you can read the dataset from Weka as a csv (comma separated) file and run a classification algorithm.
hi, You must go to uci site. Its address is as follows:
http://archive.ics.uci.edu/ml/datasets.html
On this site about the types of data sets has been described.An example of this is the following datasets.I assure you that this site is very useful.If you have questions ask you
Hi, Please see the paper entitle "Illumination invariant segmentation of vegetation for time series wheat images based on decision tree model" by W. Guo and et al., 2013
Hi, It seems you have a code on matlab for feature extraction but you are not sure about how to use the images for extraction of features.
I would suggest you to run each image on the feature extraction code to get the feature vector of each image i.e. If your feature extraction code gives you say 8 values (may be float or int type) for a single image then write those 8 values in a single line using , as a separator between them. See below for example:
Let my feature extraction code gives 8 float type values for a single image.
0.2454,0.6581,0.6578,0.3487,0.6987,0.5894,0.0124,0.2687
Now to create a feature dataset just give a identity number to your image say "image_1" for the first image and so on.
so now the feature vector of the dataset will be
0.2454,0.6581,0.6578,0.3487,0.6987,0.5894,0.0124,0.2687,image_1
now do this for each image to get the dataset ready for WEKA.
It will look like this :
0.2454,0.6581,0.6578,0.3487,0.6987,0.5894,0.0124,0.2687,image_1
0.4786,0.2458,0.6358,0.4785,0.0021,0.0245,0.1247,0.3577,image_2
0.3247,0.3657,0.1247,0.2145,0.0035,0.0004,0.0542,0.0324,image_3
-0.1896,0.1749,0.4173,0.3977,-0.0898,-0.1780,0.4107,0.4160,image_4
-0.1099,-0.1108,0.0220,-0.0276,-0.1044,-0.0373,-0.0127,-0.0992,image_5
-0.0732,0.3977,0.0420,-0.0389,-0.1409,-0.2450,-0.3245,0.1457,image_6
after the feature extraction is applied on 6 images.
To create a WEKA compatible feature extraction file i.e. *.arff file you need to create a empty file using notepad or any text editor and rename it as your desired file name with .arff extension.
Then write the feature values in the following format to create your own WEKA compatible arff file.
Given below is the example of the feature dataset for WEKA arrff file for the above example of feature vectors that I have given.
____________________________________________________________
@RELATION any_desired_name
@ATTRIBUTE any_name_of_the_feature_values_1
@ATTRIBUTE any_name_of_the_feature_values_2
@ATTRIBUTE any_name_of_the_feature_values_3
@ATTRIBUTE any_name_of_the_feature_values_4
@ATTRIBUTE any_name_of_the_feature_values_5
@ATTRIBUTE any_name_of_the_feature_values_6
@ATTRIBUTE any_name_of_the_feature_values_7
@ATTRIBUTE any_name_of_the_feature_values_8
@ATTRIBUTE class {image_1, image_2, image_3, image_4, image_5, image_6}
@DATA
0.2454,0.6581,0.6578,0.3487,0.6987,0.5894,0.0124,0.2687,image_1
0.4786,0.2458,0.6358,0.4785,0.0021,0.0245,0.1247,0.3577,image_2
0.3247,0.3657,0.1247,0.2145,0.0035,0.0004,0.0542,0.0324,image_3
-0.1896,0.1749,0.4173,0.3977,-0.0898,-0.1780,0.4107,0.4160,image_4
-0.1099,-0.1108,0.0220,-0.0276,-0.1044,-0.0373,-0.0127,-0.0992,image_5
-0.0732,0.3977,0.0420,-0.0389,-0.1409,-0.2450,-0.3245,0.1457,image_6
_____________________________________________________________
Do not use the lines in your arff file.
In the above example the float values are the feature values and here the feature dimension is 8 and the number of classes is 6 for 6 images.
Create your own using the format either manually or you can write a piece of code to automatically create the file using this format.
I would suggest write a code which can do all this that I have suggested here. It will help you if you have a large database of images.
after the creation of feature dataset in arff file format run it in WEKA.
Also you can open a example arff file of the weka software in any text editing software (Notepad, Notepad++, gedit etc) to understand the format better. The example feature files (*.arff) are located in the data folder of the WEKA software folder.
- Badges
- Science topic