TLD dataset
Zdenek Kalal
zdenek.kalal@gmail.com
17.12.2010
The TLD dataset is a collection of 10 video sequences and an evaluation protocol for comparison of object tracking
algorithms. Figure 1 shows snapshots from the sequences and table 1 lists their properties. The sequences were
manually annotated with bounding boxes. More than 50% of occlusion or more than 90 degrees of out-of-plane
rotation was annotated as “not visible”.
If you use this dataset, please cite our CVPR’10 paper:
@article{Kalal2010,
author = {Kalal, Z and Matas, J and Mikolajczyk, K},
journal = {Conference on Computer Vision and Pattern Recognition},
title = {{P-N Learning: Bootstrapping Binary Classifiers by Structural
Constraints}},
year = {2010}
}
Table 1: Properties of sequences
Fig.1: Snapshots from the sequences with the objects marked by the bounding box
