Ground truth a video

Hey guys, I already have my detection algorithm and now I am going to evaluate them. So anyone could help me how to write a GUI about ground truth a video by hand? If ant codes are available, that will be perfect for me. Thanks a lot!!

Answer 1

[Image Analyst]

This is a nice framework to get you started quickly: http://www.mathworks.com/matlabcentral/fileexchange/24224-magic-matlab-generic-imaging-component

"Description
This GUI will help the novice user get up to speed very quickly on using GUI-based applications. Everything is laid out in a very simple Step 1, Step 2, Step 3, etc. layout. It is a very good starting point for a typical image analysis application. This application uses GUIDE to do the user interface design, and has most of the basic controls such as buttons, listboxes, checkboxes, radio buttons, scrollbars, etc. It allows the user to select a folder of images, select one or more images and display them, to select a series of options, and to individually or batch process one or more images. The user can optionally apply a mask (region of interest) to the image so that only the area within the mask will be analyzed. The results are optionally sent to Excel. In this demo, I ...................."

[Mianli]

Thanks for your suggestions. This tool is quite closing to the one I want,but it cannot label more than one object in each frame. So do u have other suggestions, please?

[Image Analyst]

I'm not sure what you're talking about. Did you replace the boilerplate demo code in the AnalyzeSingleImage with your own script? It sounds like not. You have to customize it. It's just a framework. You have to fill in some of the code, like label the radio buttons with some label that is meaningful to you, etc. MAGIC just takes care of the grunt-work, overhead stuff. You have to replace those with your code. Like I said, hack out all the contents of AnalyzeSingleImage and replace with your code that you developed.

[Mianli]

OK, I will try it. Thanks for your help :)

Answer 2

[Dima Lisin]

Please try the Training Image Labeler app in the Computer Vision System Toolbox.