A Background Estimation dataset

Content


This section presents a description of each category challenges with frame samples and ground-truth images . Ground-truth images have been generated manually by selecting a unique frame when the background is entirely visible in a temporal instant or by mixing different areas of different temporal instants.

The generated dataset contains 4 categories or challenges key to evaluate the Background Estimation with 10 sequences containing the challenge: : Clutter: Low framerate: We have grouped all the test sequences into three categories according to a subjective estimation of the analysis complexity considering:

  • Baseline, containing simple sequences with low object density and no stationary objects, to evaluate the BE task in simple scenarios.

  • Clutter, containing sequences with high foreground motion and continuous background occlusions, situations where BE complexity is increased.

  • Low framerate, containing simple sequences recorded with low framerate to evaluate the impact of motion velocity in the BE task.

  • Static objects, containing stationary objects in the scene for more than and less than 50% and 100% of the video sequences duration, respectively.

In the following images the scenarios of all sequences from each category are presented through the ground-truth backgrounds.