Visual Tracking Datasets and Groundtruth
Overview
This website provides a data set for visual tracking. The data set contains 12 image sequences that depict a wide range of challenges in visual tracking, including: illumination effects, scene clutter, changes in target appearance, occlusions, and abrupt changes in velocity. Approximately half of the videos were acquired by our lab while the others were obtained from other sources. The majority of the sequences are available with manually constructed ground truth that identifies the target region within the video. As was the case with the data, some of the ground truth was constructed by us and some was obtained from elsewhere.
For details of our technical approach to visual tracking using a pixelwise spatiotemporal oriented energy representation, see our project page.
Videos Without Ground Truth
The following videos were captured in our lab with a Canon HF10 consumer camcorder at a capture rate of 30 fps. The original source of these videos is Cannons et al., ECCV, 2010. These three videos are available for download.
Book 1
Sample frames from the Book 1 video sequence. Original Source: Cannons et al., ECCV, 2010.
Book 2
Sample frames from the Book 2 video sequence. Original Source: Cannons et al., ECCV, 2010.
Book 3
Sample frames from the Book 3 video sequence. Original Source: Cannons et al., ECCV, 2010.
Videos With Ground Truth
Occluded Face 2
Sample frames from the Occluded Face 2 video sequence. Original Source: Babenko et al., CVPR, 2009. Video and ground truth available for download.
Sylvester
Sample frames from the Sylvester video sequence. Original Source: Ross et al., IJCV, 2008. Video and ground truth available for download.
Tiger 2
Sample frames from the Tiger 2 video sequence. Original Source: Babenko et al., CVPR, 2009. Video and ground truth available for download.
Ming
Sample frames from the Ming video sequence. Original Source: Ross et al., IJCV, 2008 (video) and Cannons et al., ECCV, 2010 (ground truth). Video available for download. Ground truth available for download.
Pop Machines
Sample frames from the Pop Machines video sequence. Original Source: Cannons et al., ECCV, 2010. Video and ground truth available for download.
PETS
Sample frames from the PETS video sequence. Original Source: PETS2001. Video and ground truth available for download.
Illumination
Sample frames from the Illumination video sequence. Original Source: Cannons and Wildes, ACCV 2007. Video and ground truth available for download.
Woman
Sample frames from the Woman video sequence. Original Source: Adam et al., CVPR 2006. Video available for download. Ground truth available for download.
Car11
Sample frames from the Car11 video sequence. Original Source: Ross et al., IJCV, 2008. Video and ground truth available for download.
Related Publications
- A. Adam, E. Rivlin and I. Shimshoni, Robust Fragments-Based Tracking using the Integral Histogram, CVPR, 2006.
- B. Babenko, M.H. Yang, and S. Belongie, Visual Tracking with Online Multiple Instance Learning, CVPR, pp.983-990, 2009.
- K.J. Cannons, A Review of Visual Tracking, York University Technical Report CSE-2008-07, September 16, 2008.
- K. Cannons, J.M. Gryn and R.P. Wildes, Visual Tracking using a Pixelwise Spatiotemporal Oriented Energy Representation, European Conference on Computer Vision (ECCV), 2010, pp. 511-524
- K.J. Cannons and R.P. Wildes, Spatiotemporal Oriented Energy Features for Visual Tracking, ACCV, pp.532-543, 2007.
- D.A. Ross, J. Lim, R.S. Lin, and M.H. Yang, Incremental Learning for Robust Visual Tracking, IJCV 77, pp.125-141, 2008.
- PETS (2001), http://www.cvg.rdg.ac.uk/slides/pets.html
|
