Results - 1

Using Dollar's Code and our new dataset we did some experiments. And here are some of the results.

Note that in this stage all images used for training are from the static cameras of the Azadi dataset. But testing has been made on many frames including off-dataset frames.

For comparision I've included Results from the G.Thomas's paper that we have implemented ourselves.

Frame Original Image Dollar's [1] Thomas' [2]
1 - IRIB - PTZ
2 - Azadi - Static
3 - IRIB - PTZ
4 - Azadi - Static
5 - IRIB - PTZ
6 - Azadi - Static
7 - IRIB - PTZ
8 - Azadi - Static
9 - Azadi - Static
10 - Azadi - Static
11 - Azadi - Static
12 - Azadi - Static
13 - Azadi - Static
14 - Azadi - Static
15 - Azadi - Static
16 - Azadi - Static
17 - Azadi - Static
18 - Azadi - Static

References:

[1]: P. Dollár and C. Zitnick, Structured Forests for Fast Edge Detection, International Conference on Computer Vision 2013

[2]: G. Thomas, Real-time camera tracking using sports pitch markings, Journal of Real-Time Image Processing, 2007