Dataset of Salient Closed Boundary Tracking

This dataset contains nine video sequences captured by a webcam for salient closed boundary tracking evaluation, as shown in Fig. 1.

PARAMETERS:
FPS: 30
Frame size: 640(width) x 480(height)
Total frames: 9598

Frame numbers of each sequence:
1.BookStand.avi: 899
2.Bowl.avi: 1079
3.GarbageBin.avi: 924
4.MarkCup.avi: 859
5.MarkCupContour.avi: 1226
6.MarkCupPourWater.avi: 971
7.NonplanarBowl.avi: 1454
8.ToolBox.avi: 1135
9.TransparentCup.avi: 1051

For each frame, there are two kinds of ground truth:

(1) POLYGON based ground truth in the folder "GroundTruthByPolygons".

The <xxx_TrackingGroundTruth.txt> are organized as follows:
points_number x1 y1 x2 y2 x3 y3 ... ... xn yn

Xuebin Qin, Shida He, Camilo Perez Quintero, Abhineet Singh, Masood Dehghan and Martin Jagersand. "Real-time salient closed boundary tracking via line segments perceptual grouping." accepted in the IEEE/RSJ International Conference on Intelligent Robots and Systems (IROS), September 2017.
The code can be found https://github.com/NathanUA/SlientClosedBoundaryTracking.

(2) BOUNDARY based ground truth in the folder "GroundTruthByBoundaries".

Accurate and smooth one-pixel-width boundaries are represented by .png binary images.

Xuebin Qin, Shida He, Zichen Zhang, Masood Dehghan and Martin Jagersand. "Real-time salient closed boundary tracking using perceptual grouping and shape priors." accepted in the 28th British Machine Vision Conference (BMVC), September 2017.

Contacts

Xuebin Qin
Ph.D. student Department of Computing Science
University of Alberta
Edmonton, AB, Canada, T6G 2E8
Email: xuebin@ualberta.ca
Home page: https://webdocs.cs.ualberta.ca/~xuebin/