need code which created 'stopSigns.mat'

Using Computer Vision system toolbox i'm trying to detect bottles in an image. For that I want to train Computer Vision System toolbox with "trainCascadeObjectDetector" function. I also analysed the example program "Train a Stop Sign Detector"

But the problem is i'am stuck at one point, that is, for my application how to create and load positive samples data from "bottles.mat" file, which is same as "stopSigns.mat"

Can any-one crack the issue, or is there any alternate way to train Computer Vision System Toolbox without using .mat file for positive samples data input.

[mustafa]

You can create a GUI for bounding boxes via 'imrect' or use @Brett Shoelson's great GUI on this link to create .mat file for object recognition.

Answer 1

[Anand]

You don't really need a mat file. You need ground truth data for training. So, you need a set of images with bottles in them and bounding boxes for the bottles in the images.

The mat file just contains an array of structs specifying the name of the image file and the bounding box of the bottle in that image.

Read this: http://www.mathworks.com/help/vision/ref/traincascadeobjectdetector.html#inputarg_positiveinstances

The mat file is just used to pass the positive instances input argument to the trainCascadeObjectDetector function.

Hope that helps!

[shreedhar bc]

Sir I have 100 images in which I have bottles at different place. Each image contain one bottle. Now How to define bounding box for bottle at different place in all 100 different Images ?

[Anand]

Like mustafa suggested, you can use this file exchange submission to help with that:

cascadeTrain GUI.

You'll need to individually draw bounding boxes for each of the training images.

Answer 2

[Esam]

Try this code to create ur .mat file as I did. u'll need to describe each positive sample :

data = [];

field1 ='imageFilename';

value1 = {'Car/T01.bmp','Car/T02.bmp','Car/T03.bmp','Car/T04.bmp','Car/T05.bmp'};

field2 ='objectBoundingBoxes';

value2 = {[21,25,100,150] ,[17,10,168,122] ,[20,13,150,182] ,[9,15,162,100] ,[17,3,130,107]};

data = struct(field1,value1,field2,value2);

save ('cars.mat','data');

[Datti Nagadhara Harini]

how do u get those values to prepare a boundary box around the object